Volume 19 Issue 4 by IFAMA

International Food and Agribusiness Management Review

Official Journal of the International Food and Agribusiness Management Association

Volume 19 Issue 4 2016

EVERY GENERATION NEEDS ITS LEADERS. As a farmer-owned cooperative, we help lead the agricultural industry by promoting the most innovative, responsible farming practices available today. And we want to make sure that the next generation of leaders is well trained for tomorrow. That’s why we proudly support youth leadership programs like FFA. It’s also why we’ve invested in scholarships and countless land-grant university initiatives across the country. At CHS, we know every generation needs its leaders. And every potential leader needs a little help becoming one. Visit chsinc.com/stewardship to learn more about our ongoing commitment to leadership.

International Food and Agribusiness Management Review

Editorial Staff Executive Editor

Gerhard Schiefer, University of Bonn, Germany

Regional Managing Editors Asia, Australia, and New Zealand

Kim Bryceson, University of Queensland, Australia Jeff Jia,University of Exeter, United Kingdom Nicola M. Shadbolt, Massey University, New Zealand

Europe

Pegah Amani, Technical Institute of Sweden, Sweden Vera Bitsch, Technical University of Munich, Germany Alessio Cavicchi, University of Macerata, Italy Klaus Frohberg, University of Bonn, Germany Diogo Souza Monteiro, University of Kent, United Kingdom Jacques Trienekens, Wageningen University, The Netherlands

North America

Ram Acharya, New Mexico State University, USA Vincent R. Amanor-Boadu, Kansas State University, USA Yuliya Bolotova, Clemson University, USA Michael Gunderson, Purdue University, USA Mark Hansen, Brigham Young University, USA R. Brent Ross, Michigan State University, USA Aleksan Shanoyan, Kansas State University, USA David Van Fleet, Arizona State University, USA

South America

Aziz da Silva Júnior, Federal University of Vicosa, Brazil

Africa

Ajuruchukwu Obi, University of Fort Hare, South Africa

Editorial Board

Filippo Arfini, Universita’ di Parma, Italy Stefano Boccaletti, Universita’ Cattolica, Italy Michael Boehlje, Purdue University, USA Dennis Conley, University of Nebraska - Lincoln, USA Francis Declerck, ESSEC Business School, France Hamish Gow, Massey University, New Zealand Jukka Kola, University of Helsinki, Finland Jay Lillywhite, New Mexico State University, USA

Woody Maijers, INHOLLAND University, The Netherlands Marcos Fava Neves, FEA / USP / PENSA, Brazil Onno Omta, Wageningen University, The Netherlands Hernán Palau, Buenos Aires University, Argentina Christopher Peterson, Michigan State University, USA Thomas Reardon, Michigan State University, USA Mary Shelman, Harvard Business School, USA Johan van Rooyen, University of Stellenbosch, South Africa

The IFAMR (ISSN #: 1559-2448) is published quarterly and the archived library is available at http://www.ifama.org. For copyright and publishing information, please contact: Marijn van der Gaag, Administrative Editor Wageningen Academic Publishers • P.O. Box 220 6700 AE Wageningen • The Netherlands • Tel: +31 317 476511 Fax: +31 317 453417 • Email: ifamr@wageningenacademic.com • Web: http://www.wageningenacademic.com/loi/ifamr

International Food and Agribusiness Management Review Volume 19 Issue 4, 2016

EDITORâ&#x20AC;&#x2122;S NOTE Dear Colleagues, Welcome to the final issue of 2016 and the first issue under the stewardship of a new publisher and editor. The journal is looking back at a year with the successful publication of regular and special issues with Peter Goldsmith as executive and Kathryn White as administrative editors. The publications demonstrate the broad interest in the management, organization, and development of the food sector and its enterprises. The challenge of sustainability of the food sector has initiated an increased interest in the need for cooperation and interaction among all actors in the food sector including agriculture, industry, trade, retail, consumers, and policy. IFAMR will continue to support this broad view and to assure that the journal provides a most relevant platform for discussions of challenges that the food sector is facing in meeting its responsibility for feeding the world in increasingly complex and fragile economic, social and natural environments. Wageningen Academic Publishers as the new publisher of IFAMR with Marijn van der Gaag as Administrative Editor and Prof. Gerhard Schiefer as the new Executive Editor are committed to this goal and are looking forward to building on the successful activities of the former editors and to serving the community with a high level journal for its publication interests. The editors are supported by a dedicated group of Managing Editors from around the world. Without them the journal could not handle the many submissions and reviews a high-level journal has to contend with. During the past year some of the Managing Editors left their engagement due to retirements, new positions and similar developments. We gratefully acknowledge the services by Murray McGregor of the University of South Australia, Derek Baker of the University of New England (Australia), and Joao Martines-Filho of the Universidade de Sao Paulo. We were also saddened by the news that long-time friend, colleague and former Managing Editor, David Sparling, from the University of Western Ontario passed away in late July after a 22-month battle with brain cancer. A number of colleagues from various countries are volunteering to engage as Managing Editors for some time to come. We welcome Kim Bryceson of the University of Queensland, Jeff Jia of the University of Exeter, Pegah Amani of the Technical Research Institute of Sweden, Klaus Frohberg of the University of Bonn, and Aziz da Silva of the Federal University of Vicosa, Brazil. The journal is in its 19th year but still in dynamic development. We are open for proposals for changes and further improvements. Just link up with us. Looking forward to a fruitful cooperation with all of you Gerhard Schiefer, Executive Editor, IFAMR ÂŠ 2016 International Food and Agribusiness Management Association (IFAMA). All rights reserved.

iii

International Food and Agribusiness Management Review Volume 19 Issue 4, 2016

TABLE OF CONTENTS 1. 2.

GLIMPSE 2.0: a framework to feed the world

Aidan J. Connolly, Luiz R. Sodre, and Kate Phillips-Connolly

Members’ attitudes towards cooperatives and their perception of agency problems

Do coffee cooperatives benefit farmers? An exploration of heterogeneous impact of coffee cooperative membership in Southwest Ethiopia

Performance of small and medium-sized food and agribusiness enterprises: evidence from Indian firms

The relevance of business practices in linking smallholders and large agro-businesses in Sub-Sahara Africa

The influence of value chain integration on performance: an empirical study of the malt barley value chain in Ethiopia

Dairy farm households, processor linkages and household income: the case of dairy hub linkages in East Africa

Karin Hakelius and Helena Hansson

Zekarias Shumeta and Marijke D’Haese

Jabir Ali

Linda Kleemann

Mulugeta D. Watabaji, Adrienn Molnar, Manoj K. Dora, and Xavier Gellynck

Elizaphan J.O. Rao, Immaculate Omondi, Aziz A. Karimov, and Isabelle Baltenweck

Fertilizer freight rate disparity in Brazil: a regional approach

Lilian M. de Lima, Lilian de Pelegrini Elias, José V. Caixeta-Filho, and Jamile de Campos Coleti

109

Consumer perceptions of climate change and willingness to pay for mandatory implementation of low carbon labels: the case of South Korea 129

Hyeyoung Kim, Lisa A. House, and Tae-Kyun Kim

International Food and Agribusiness Management Review iii

Table of contents

Volume 19, Issue 4, 2016

10. Economic feasibility of tobacco leaves for biofuel production and high value squalene

145

11. Ethanol and sugarcane expansion in Brazil: what is fueling the ethanol industry?

163

Aleksandre Maisashvili, Henry L. Bryant, and James W. Richardson

Ana Claudia Santâ&#x20AC;&#x2122;Anna, Aleksan Shanoyan, Jason Scott Bergtold, Marcellus M. Caldas, and Gabriel Granco

International Food and Agribusiness Management Review iv

OPEN ACCESS International Food and Agribusiness Management Review Volume 19 Issue 4, 2016; DOI: 10.22434/IFAMR2015.0202 Received: 8 November 2015 / Accepted: 7 October 2016

GLIMPSE 2.0: a framework to feed the world RESEARCH ARTICLE Aidan J. Connolly a, Luiz R. Sodreb, and Kate Phillips-Connollyc aChief

Innovation Officer and Vice-President, Corporate Accounts, Alltech, 3031 Catnip Hill Road, Nicholasville, KY 40356, USA

bCo-founder

and CEO, Perfarm, Rua Tefé, 292, São Paulo, SP, 01251-050, Brazil

cPhD,

Institute for International Integration Studies, Trinity College Dublin, University of Dublin, College Green, Dublin 2, Ireland, United Kingdom

Abstract Five years ago a new acronym GLIMPSE was proposed in the International Food and Agribusiness Management Review to summarize the seven barriers faced by agriculture in its quest to feed the world, based on interviews of 25 agribusiness experts. Through an iterative, grounded theory methodology the original research that led to the GLIMPSE framework was validated, deepened and expanded. The new research made minor revisions to the original GLIMPSE, but confirmed it as an effective framework to explain to an interested public how agriculture can tackle the planet’s nutritional requirements if certain constraints are addressed. Specifically, international policy makers, governments, non-governmental organization, charities, industry organizations, integrated food companies and farmers often struggle to explain the complex challenges agribusiness faces, and in this respect the GLIMPSE framework allows all stakeholders to describe the main challenges agriculture faces on its journey to feed almost 10 billion people by 2050. Keywords: future of food, agriculture, agribusiness, sustainability, feeding the world JEL code: Q10 Corresponding author: aconnolly@alltech.com

Connolly et al.

Volume 19, Issue 4, 2016

1. Introduction The collective businesses involved in agricultural production commonly referred to as the agribusiness sector (Davis and Goldberg, 1957), face a number of serious challenges in feeding a global population that is increasing exponentially in size and income. At the same time, the expectations and attitudes towards agribusiness are changing. There has been little research either to identify areas where agribusiness can make a difference in meeting these challenges or to understand changing attitudes towards agribusiness. One of the most comprehensive efforts was the GLIMPSE framework of 2012 (Connolly et al., 2012), which sought to identify both the obstacles and opportunities in agribusiness. GLIMPSE has sparked considerable discussion over the four years since its publication. Recently both external input and continued research by the authors has generated sufficient new information that it warrants a review of the GLIMPSE framework. In this paper we extend the research, update and expand the data, and offer some modifications that reflect the additional contributions from academic experts, industry leaders, professionals and consumers at large. The paper begins with a brief background section, noting the ‘wicked problem’ of population growth, the changing expectations facing the sector, and the limited literature available. The next section outlines the methodology used, including a review of 1.3 million websites, followed by a summary of the results. A discussion on the findings for each element of the acronym GLIMPSE is then provided, and the paper concludes with a brief discussion of the revised GLIMPSE model, its uses and implications. The paper provides managers with useful tools for assessing ways in which consumer attitudes are changing so that they can orient their business towards growing to feed the 9 billion people projected to be living in the world by 2050. Background The original impetus for GLIMPSE was to try and identify the obstacles to feeding the rapidly growing population, and then to see where there were opportunities for agribusiness to contribute to that effort. Since then world population growth projections have been revised sharply upward again: it is now estimated that the world will have 2.5 billion more people to feed in 35 years, and 11 billion by 2100. As the benefits of the Green Revolution of the last century level out, and the amount of arable land is limited, finding ways to increase food production to feed everybody is a significant challenge. Equally challenging is the growth of the ‘consuming class’1: more than 1.2 billion people were added to that group in the past decade, and another 1.8 billion are expected to join in the next decade (Dobbs et al., 2015). Much of this growth will come from rising incomes, notably in developing countries such as Brazil, Russia, India and China. Gross domestic product growth in 20 emerging economies is expected to nearly triple by 2050 (PwC, 2015). As income rises, demand for protein and other types of non-grain foods increases substantially. And, just as consumers in developing regions are changing their expectations about food, so are consumers in more developed markets. Attitudes about food, and importantly, agribusiness are changing sharply. In part that reflects a growing disconnect: as populations move away from rural areas and become connected to agriculture only through the food that they eat, their focus moves to how they relate to food and food production. Increasingly, these issues concentrate around safety, nutrition and the environment. Food safety concerns include disease outbreaks (e.g. bovine spongiform encephalopathy/mad cow disease, avian influenza), food contaminations (e.g. Salmonella, Escherichia coli) and food adulteration. Nutritional concerns include the presence of additives, carcinogens, fats, salts, sugars, etc. Environmental issues relate to the impact on the environment or the animals of modern farming methods. Through various media platforms consumers are

1 Individuals

who have over $10 per day to buy discretionary items.

International Food and Agribusiness Management Review

Connolly et al.

Volume 19, Issue 4, 2016

increasingly aware of these issues, and they use these same platforms to both monitor what is happening and express their views. For these consumers, despite occasional spikes in food prices (e.g. the 2007-2008 world food price crisis), the cost of food is no longer a major part of their household budget: in more developed countries the proportion of the household budget spent on food has fallen from about 45% in 1900 to 6-15% in 2009 (Bill & Melinda Gates Foundation, 2012). They can afford to focus on preferences rather than needs. The challenge for agribusiness is to feed a rapidly growing population whose food preferences and expectations of agribusiness are changing rapidly, yet there is surprisingly little research on the subject. There has been some work on specific agribusiness challenges, such as a study that asked agribusiness professionals around the world to rank the probability of various challenge scenarios happening; it concluded that global warming is seen as the greatest future challenge (Lakner and Baker, 2014) A useful, but regionally limited survey of CEOs in Africa asked for challenges constraining agribusiness expansion in Africa; the conclusion was that scarcity of resources, access to technology and climate change were the major challenges for the near future (PwC, 2015). Other work has addressed climate change (Vervoort et al., 2014); a scenario-guided analysis of food security challenges due to climate change (Wheeler and Von Braun, 2013); environmental sustainability; and the use of natural resources. There have also been literature reviews seeking a more cohesive analysis of these challenges, such as a review of the challenges in achieving sustainable agricultural production by 2050, including action recommendations. McKenzie and Williams (2015) and Boehlje et al. (2011) attempted to categorize and evaluate the challenges faced by agriculture using frameworks, and suggested that the three major issues are growing risk and uncertainty; developing and adopting new technologies; and rapid market responsiveness to changes in industry structure. A meta-analysis of a set of Food and Agriculture Organization foresight studies asserts that the serious challenges are not about producing more food but on managing the social and political issues that drive food insecurity (Bourgeois, 2014). None of this work is comprehensive, and what recommendations are offered are of use only to small segments of the food chain. Thus, the decision was made to revisit GLIMPSE, substantially increasing the depth and scope of the research to test the validity and relevance of the work. Extensive new research was conducted to ensure that the views of participants throughout the food chain about the challenges facing them were captured.

2. Methodology The research was conducted in three phases. Phase 1 consisted of in-depth open-ended interviews with experts in agriculture, which were followed in Phase 2 with a survey by industry leaders ranking the challenges they face. Finally, in Phase 3, consumer views and attitudes were explored through an analysis of social media content. In total, the views of some 600 academic experts and senior level executives, as well as more than a million social media posts published in the past three years were collected and analyzed. The first two phases followed the same methodology as the original GLIMPSE work, but with larger sample sizes and specific attention to regional, functional, sector distribution, while the third turned to a newly available, and still underutilized, tool. Phase 1: interviews Phase 1 consisted of 59 in-depth interviews with academic experts and industry leaders from 23 different countries, from a range of backgrounds. The participants were chosen to ensure a valid response rate, broad regional coverage (to allow for differentiation of regional differences), and industry backgrounds. Participants from the government sector included a former United States Secretary of Agriculture and a recent European Union Commissioner on Health and Food Safety, while participants from the academic arena included professors from universities including Harvard, Purdue, UC Davis, University of Sunshine Coast (Australia) International Food and Agribusiness Management Review

Connolly et al.

Volume 19, Issue 4, 2016

and University College Dublin. From the private sector there were agribusiness partners from some of the top four global strategic consulting firms; managing directors of investment and development banks in the United Kingdom, Brazil and South Africa; C-suite executives of numerous agribusinesses from different countries (including Canada, Peru, Ukraine, Egypt, India and China); and current and former presidents of national associations (e.g. the National Turkey Federation, the American Animal Science Society, etc.). Interviews were typically 25 to 40 minutes long, and were designed to allow for spontaneous responses through open-ended questions. As in the original GLIMPSE research, respondents were asked for their opinion on the biggest challenges facing agribusiness in feeding a growing, and increasingly affluent, global population. Additional directed, but still open-ended questions based on the original GLIMPSE findings were also asked. The follow-up questions were used to test whether the original GLIMPSE findings were both collectively exhaustive and mutually exclusive. Using a grounded theory approach (Glaser and Strauss, 1967), the collected data was evaluated using a coding analysis process (Bryman, 2012). This produced a consolidated grouping of challenges that were then tested in Phase 2. Phase 2: survey The findings from Phase 1 were tested through a survey of 527 experts. Given the list of 22 challenges derived from the expert interviews, the respondents were asked to select the five biggest challenges and rank order them. The challenges were introduced in a randomized order to avoid any anchoring from the first list of challenges. Respondents were also given the option of adding their own challenge(s). To allow for segmentation analysis, demographic questions were also included. The survey was offered in English, Spanish, Portuguese and Chinese. Invitations were sent to industry leaders and academicians, who were identified through the International Food and Agribusiness Management Association network, as well as attendants from an international industry symposium and attendants at a leadership level industry gathering. Executives comprised 25% of the respondent pool; 23% were managers, 17% agribusiness owners, and the remaining were independent, retired professionals, and others. The respondents were distributed nearly evenly across industry activity groups, with about a third in academia, research or consulting; 23% in farm inputs; 19% in primary production, and the remaining quarter in downstream activities, government, nongovernmental organizations (NGOs), financial markets, or the press. To ensure that respondents used the same terms of reference when identifying their place in the agribusiness value chain, an illustrated value chain (with examples) was provided. The respondent sample was also evenly distributed across years of experience in agribusiness and by the size of firm with which they were associated. Phase 3: social media The first two phases involved expert opinion drawn from the agribusiness sector. The third phase sought to capture public perception of agribusiness in general and the challenges of feeding the world in particular. Analyzing social media content to support business decisions is a relatively new, and underutilized, research tool. Social media captures information in an unprompted manner, allowing responses to reveal what is truly top of mind to consumers. The focus of Phase 3 was on identifying and evaluating discussions about the challenges of agribusiness, and to look for trends and patterns across the data. Using an artificial intelligence (AI) system known as Crimson Hexagon, the frequency of particular words and topics were automatically identified and classified. Using publicly available sources, content from Twitter, Facebook, blogs, forums and others social media platforms posted during a three-year period (2012-2015) were analyzed. Primary keywords such as â&#x20AC;&#x2DC;food productionâ&#x20AC;&#x2122; International Food and Agribusiness Management Review

Connolly et al.

Volume 19, Issue 4, 2016

or ‘agribusiness’ were used to identify the industry subject to discussions; secondary or auxiliary keywords, such as ‘challenge’ or ‘barrier’ were used to identify themes and topics within industry-related discussions. Posts containing ‘http’ were excluded from the search to keep the focus on discussions, not referrals to third party websites and/or advertisements.

3. Results Phase 1: interviews A careful data coding process (including searching for common words or phrases that could be applied to a broader range of subjects) yielded two main groups of challenges: those related to optimal use of production factors to improve input efficiency and output productivity; and those challenges related to coordination between agribusiness and key stakeholders. There was a consensus that regional low productivity is a result of constraints on factors of production that are readily available elsewhere in the economy or in the world. In particular, respondents noted technology, labor, and capital (fixed or financial) as being top challenges. These factors can often be transferred from one area to another. While land and natural resources, including water and climate, are not easily transferable from one place to another, they often are not the most limiting factor in low productivity. Moreover, even in places where land or natural resources is indeed the limiting factor, the respondents felt that the larger challenges are still related to the use of other factors of production, and that the overall increase in productivity can only be achieved by the increase in productivity of technology, labor or capital. Production output in any industry is constrained by the most limiting factor of production. The challenges not directly linked to factors of production relate to coordination between agribusiness and key stakeholders, whether the stakeholders were those involved in food processing or the larger community (consumers, governments, creditors, investors, etc.). Again, there was a strong consensus among interviewees, in this instance regarding the challenge of aligning the efforts and priorities of these disparate stakeholders towards ensuring that there is enough food, at the right time and place, for the growing global population. The 22 challenges identified in Phase 1 are consistent with those from other studies. For example, all of the challenges found in the PwC Agribusinesses Insights Survey 2014/2015 are represented within the challenges identified in this study. Similarly, the challenges most likely to affect agribusiness through 2030 included water scarcity, increased demand for individualized nutrition and local foods, and increased consumption in developing countries (Lakner and Baker, 2014) are also found in this study. Phase 2: survey A survey including shared closed groups containing agribusiness experts in professional social media networks, Alltech customers and IFAMA members yielded responses from 527 agribusiness professionals, with the responses relatively evenly distributed across the defined demographic groupings. The findings were analyzed both by the respondent characteristics and by the challenge group. Although there were variations in the frequency, all 22 challenges derived from Phase 1 were ranked among the top 5 challenges, validating the GLIMPSE findings. Moreover, just 2% of the respondents provided unique additional challenges, which were added to the original dataset for analysis in Phase 3. These additional challenges account for only 0.5% of the total, which can be seen as validating the theoretical saturation described in Phase 1, confirming that the range of experts’ opinions gathered were indeed representative of the industry, and that the coding and categorization process was representative of the data set.

International Food and Agribusiness Management Review

Connolly et al.

Volume 19, Issue 4, 2016

Interestingly, the challenges relating to the use of production factors and those relating to interactions with stakeholders were noted nearly equally. Production related challenges (those under Infrastructure, People, Science and Environment) accounted for 49% and stakeholder related challenges (Government, Loss and Markets) for 51%. The relative rankings are discussed further, below. The relative frequency with which population segments listed each challenge was analyzed, providing useful insights in the way that different segments of the agribusiness value chain perceive the challenge, as well as a useful way to both confirm the broad validity of the challenge list and to eliminate any unintended bias in the process. The data was analyzed against five variables: region, years of experience in agribusiness, size of the firm, activity (where the respondent is in the value chain), and the nature of the respondents role along with the relative weighting assigned to the challenge by the respondent. ■■ Region Regional analysis confirmed that the full range of challenges is felt globally, with results evenly distributed across regions. The highest level of responses was for the Consumer Marketplace, which was ranked in the top 5 by all regions, and was the most mentioned overall. However, while it was seen as important in all regions, it was not ranked as one of the most serious challenges. The most variability was found in Investment and Infrastructure. There were strong variations in regional importance, ranging from a low of 3.35% in North America to a high of 12.68% in Africa. Environmental challenges, on the other hand, had the least regional variability, with all regions similarly. Oceania, Africa and Latin America were notably more concerned than average with Science & Innovation challenges (15.42, 10.56 and 11.16%, vs an overall average of 4.26%). Other regional variations were found between occupations: professionals in Asia were notably more concerned about the challenges relating to human capital than professionals in Europe and Latin America (14.44%, vs 7.45 and 7.30%, respectively), while professionals in Europe, and to a lesser extent Oceania, ranked Food Losses as more of a challenge than other regions (9.36 and 3.74%, respectively). On the other hand, professionals in North America and South America were the most likely to cite Government & Policies challenges (24.89 and 24.46%, respectively). ■■ Experience Overall there was less variability in the frequency of responses as analyzed against years of experience in agribusiness. The two notable exceptions involved respondents with less than 10 years of experience, and those with more than 40 years of experience. The less experienced respondents see Consumer Markets as more of a challenge than those with more experience (28.95 vs 22.02% for respondents with over 50 years of experience). Professionals with more than 40 years of experience in agribusiness see Government & Policies as a somewhat bigger challenge (22.10%) than the newer people to the industry (16.96%). ■■ Size Analysis based on company size also demonstrated little variability, with the exception of those professionals from large companies (over 5,000 employees), who saw the Environment and Food Losses categories relatively more important than respondents from other segments (20.83 vs 19.61% average and 9.60 vs 7.35% average, respectively), and saw the Government & Policies category as less critical a challenge (16.76 vs 18.71% average).

International Food and Agribusiness Management Review

Connolly et al.

Volume 19, Issue 4, 2016

■■ Activity Analysis of responses by activity within the agribusiness value chain produced the largest variations between respondents. Although the findings must be considered with some caution, due to some segment sample sizes being rather small, the responses do appear to correlate with the respondent’s position in the value chain. Entities closer to end consumers (such as Food Retail and Press) were much more likely to rank Consumer Markets as a significant challenge than the industry at large (25.88 vs 1.89%). Similarly the Food Retail, and Trading & Processors segments are more concerned about Food Loss than the industry as a whole (16 and 11.58 vs 7.11%, respectively). At the other end of the value chain, NGOs are more concerned with production factors (especially scarcity and optimal usage), particularly in three categories: People (15.56 vs 9.84% average), Science & Innovation (15.56 vs 9.28% average) and Environment (24.44 vs 19.11% average). Government professionals see the Government & Policies category as being the most important obstacle (33.33 vs 19.81% average), but do not consider Food Losses or issues relating to Consumer Markets as critical obstacles (20.00 vs 25.88% average). Self-employed professionals were less concerned about the challenges involving People (6.31 vs 10.01% average) and Science & Innovation (9.01 vs 13.54% average), but more concerned about Environment (26.13 vs 19.61% average) and Food Losses (9.91 vs 7.35% average). Phase 3: social media The AI application retrieved 1,395,652 posts meeting the search criteria, primarily from blogs and forums, as well as Facebook and Twitter. Over 250,000 posts had an identifiable location, the majority of which were in English speaking countries (64% from the United States), which was expected as the keywords used were all in English. The largest proportion of responses came from California (7%) and Texas (6%), followed by New York (5%), Florida and the District of Columbia (4% each). ■■ Analysis of content The major data analysis examined the frequency with which specific words are used. Obviously, the most frequent words are the keywords used within the search criteria, and these are excluded from the analysis. The associations between the remaining words relating to the challenges were be analyzed for patterns. About a third of the most frequently used words can be easily associated to GLIMPSE categories, including ‘water’ (Environment), ‘government’ (Government & Policies) and ‘health’ (Consumer Markets). Some words such as ‘industry,’ ‘business,’ ‘company,’ and ‘management’ transcend many categories. To get a better understanding of the context, the full posts from which the words were drawn were sampled. From these samples it appears that the posts relate to the challenges of doing business in the private sector as a whole; these challenges can and do inhibit the overall inability to rise to the challenge of feeding the world, but are not specific to that issue. Some of these elements are already included within the challenge categories, (particularly Infrastructure & Investment, People and Science & Innovation). However, as the objective of GLIMPSE is to evaluate the external challenges to the industry, the general challenges of being in business are not put forward as a separate challenge category. There are no other obvious categories for the remaining words that do not clearly link to a GLIMPSE category. Thus, the associations confirm the comprehensiveness of the GLIMPSE framework, both in the areas captured and by the lack of any other required categories. Moving from analyzing single words to analyzing clusters of words adds more depth to the analysis. The AI sorting program is first trained to recognize related words by having data samples manually categorized; the system then uses an algorithm to aggregate the remaining data based on content similarities from the sample. For this work, 350 posts were manually classified according to a set of criteria drawn from the GLIMPSE

International Food and Agribusiness Management Review

Connolly et al.

Volume 19, Issue 4, 2016

categories. Posts containing the key words but not closely related to the object of the study were excluded from analysis. The relationships of words that frequently appear together in posts are represented in word clouds, or interconnected bubbles (Figure 1). When observing these clusters, GLIMPSE categories can be clearly identified in several of them. Analyzing the data based on the forum from which it was drawn also produced variations. In particular, the word clouds (Figure 2) drawn from Twitter and Facebook produced more content words related to Consumer Markets and People, while those related to Government & Policies and Science & Innovation were rarely found. Note that the present analysis does not take into consideration the number of views or level of engagement of posts (likes, shares, etc.) but only their content. Overall, there were 814,299 relevant posts identified within the three-year period examined; the rest were excluded. Observing the word clouds from each of the categories there is a clear correlation between the most frequent words and category theme, demonstrating that the application did a satisfactory job of categorizing the posts. As expected, given the inter-relationships between various GLIMPSE categories, some words appear in multiple word clouds. Over the three-year period, Government & Policies was the category with the highest number of related posts (20%), followed by New Technology/Biotechnology (17%). Losses has been classified along with â&#x20AC;&#x2DC;Othersâ&#x20AC;&#x2122; as the low rate found during the manual categorization process made it difficult to determine a clear criteria for recognition training by the system (Figure 3). Finally, there was evidence of change over time in the trends and patterns identified, with particular growth shown in posts relating to the People and Science & Innovation categories.

Figure 1. Cluster (7 October 2012 â&#x20AC;&#x201C; 7 September 2015; sample of 10,000 posts). ) (G = Government & Policies; E = Environment; P = People; M = Consumer Markets; S = Science & Innovation)

International Food and Agribusiness Management Review

Connolly et al.

Volume 19, Issue 4, 2016

Figure 2. Word cloud without keywords for different content sources (7 October 2012 â&#x20AC;&#x201C; 7 September 2015; sample of 10,000 posts each).

Figure 3. Breakdown (%) of challenge categories in 2012/2013 and 2013/2014 (Phase 3: social media). Overall, the social media analysis supports the findings and conclusions from Phases 1 and 2. Most of the GLIMPSE categories are present in the posts the study identified. Moreover, analyses of the most frequent words indicate that the GLIMPSE challenges are reasonably comprehensive. When the results are compared with those from Phase 2 the only notable difference is the Food Losses category, which did not have a particularly strong presence. This may be attributed to either low awareness of the issue among the general public, or that it is seen as a subsidiary challenge (Figure 4).

International Food and Agribusiness Management Review

Connolly et al.

Volume 19, Issue 4, 2016

E Environment 18% S Science & innovation 13% P People 10%

E Environment 16%

G Government & policies 20% L Food losses & others 7% I Investments & infrastructure 8%

M Consumer markets 24%

S Science & innovation 23%

Phase 2

P People 14%

G Government & policies 20%

M Consumer markets 17%

L Food losses & others 4% I Investments & infrastructure 6%

Phase 3

Figure 4. Comparison of the frequency of challenge categories under GLIMPSE obtained in Phase 2 (survey) and Phase 3 (social media).

4. Discussion The findings from the three phases of new research were evaluated in the context of the original GLIMPSE acronym, testing the alignment between the new data and the GLIMPSE framework, expanding the understanding of each of the challenges, and integrating the new consumer element. In reviewing the findings, it was apparent that the overlap between Government and Politics & Policies was such that it made sense to combine. It also became clear that people, in particular the availability of appropriately skilled workers is emerging as a significant challenge. Accordingly, the GLIMPSE acronym has been amended to reflect these findings. The discussion below uses the revised framework (Table 1). Government and policies Although Government & Policies covers many areas, there were four aspects that particularly stood out: corruption and self-interest; environmental regulations; trade-distorting subsidies or quotas; and poor quality government infrastructure. Table 1. New and original (2012) GLIMPSE frameworks. New GLIMPSE framework, aligned with 2012 categories G Government & policies

L Losses

Infrastructure & investments

Markets: consumers

People

Science & innovation

Environment

Infrastructure

Markets

Politics & policies

Science & innovation

Environment

2012 GLIMPSE categories Government

Losses in the food and ingredient supply chain

International Food and Agribusiness Management Review

Connolly et al.

Volume 19, Issue 4, 2016

■■ Corruption and self interest Professionals working with governments (5.00%), along with those from academia, research (5.86%) and ‘others’ (5.15%) expressed the most concern with corruption and self-interest politics, while agribusiness experts generally did not spontaneously bring up corruption. Regionality is significant here: the majority of interviewees consider corruption to be embedded in their societies. Professionals in Africa and Asia indicated the most direct concern with the impact of corruption (~7% put it among the top 5 challenges), while the professionals from Latin America indicated less concern about corruption than the global average (4.05 vs 5.07% average). Despite the current corruption scandals in Brazil, a managing director from an investment bank in Brazil noted that corruption was inherent in the society as a whole and thus was not seen as a particular, direct, barrier to agribusiness. The perception is that corruption and self-interest is omnipresent and inextricably linked to doing business. This acceptance of corruption as inevitable is itself a constraint on developing solutions. ■■ Environmental regulatory changes Climate change was the challenge most frequently mentioned in the open-ended question in Phase 2. Agribusinesses have to prepare for regulations arising from concerns about climate change, as well as changing regulations involving other environmental elements (such as waste disposal and water management). Both region and role affected the responses regarding these challenges. Across all regions, regulatory concerns were the second most commonly cited (6.26% average), but regionally regulatory changes were particularly noted as a challenge by professionals in Latin America (9.87%). By role, professionals from trading and processors and very large companies (over 10,000) (4.21 and 4.26%, respectively) were more sanguine about regulatory changes than government respondents and farm producers, who considered environmental regulations a significant challenge (11.6 and 9.39, respectively, vs 6.90% average). As a respondent from a major Irish feed brand noted, they know that ‘Governments will play a role in regulatory changes [to] offset emissions by changing production methods. [...] 10% of emission in EU comes from agriculture’. ■■ Subsidies and quotas Professionals from Oceania (4.21%) and Latin America (3.86%) saw the issue of subsidies as more problematic than those from Europe (1.49%) and North America (2.17%), which seems to correlate with the history of subsidies between the regions. Professionals working in the government attributed a much higher level of importance to the Government & Policies challenge within GLIMPSE (5.00 vs 2.52% average). There were numerous quotes about the levels of subsidies in other markets, including from an agricultural company, an Irish farm input retail firm (who noted that approximately 80% of farmers ‘lose money operationally and survive on government subsidies’), the American Feed Association (China’s subsidies for inefficient agriculture) and so on. A leader of Harvard’s agribusiness program argues that ‘Countries want to be self-sufficient to better respond to rise in prices’. But the consensus among a significant majority of the experts interviewed is that protectionism fosters inefficiency and diminishes worldwide food output. ‘It is politics more than efficient resource allocation,’ noted a partner of a strategic consulting firm. ■■ Poor government infrastructure The importance of poor government infrastructure (that is politically unstable governments, poor law enforcement, and excessively high bureaucracy) seems to be affected more by the respondents’ role within the agribusiness community than by the region that they are from. Regionally, the level of concern is fairly evenly distributed, with an average around 6%. However, government professionals saw the challenges of poor government infrastructure as a significant challenge, and were significantly more likely than the average respondent to cite it as a challenge for agribusiness (11.67 vs 6.90% average). An advisor to the Ministry of Agriculture in Ukraine noted that an ‘unstable and poor legislation system’ is their number one International Food and Agribusiness Management Review

Connolly et al.

Volume 19, Issue 4, 2016

challenge in agribusiness. A frequently cited concern is a lack of rule of law, and that enforcement of both laws and regulations especially in rural areas is at best uneven, frequently chaotic, and often absent altogether. Examples offered included not following vaccination protocols and illegal deforestation. Failure to enforce regulation is matched as a challenge by excessive regulation. As an executive of a large food company in Thailand noted ‘regulations delay the release of new products’ and that government should be more active in solving the issue. Producers worry about the delays in getting agricultural products to processors or market. Loss and waste Food loss (that is, waste that occurs in the upstream, or production part of the value chain), and food waste (which occurs in the downstream, or retail and consumption part of the value chain) are estimated to result in the loss of 24% of the calories contained in all the food that is produced (Lipinski et al., 2013). Although food loss is more of an issue in less developed regions, and food waste in more developed regions, there is both loss and waste globally. ■■ Food loss: upstream In developing regions, food loss occurs primarily through the production and logistics process, and is linked to a number of GLIMPSE factors, notably Infrastructure, Technology, and Government. A lack of workforce skills was also noted by as a significant number of respondents. Food loss also contributes to issues in the Environment, particularly where the waste or pollution of natural resources affects the environment. Respondents suggested that improvements in the coordination of agents such as governments, consumers, or factors of production, would lead to the reduction of waste in the supply chain. Regionally, food loss along the supply chain is considered a greater challenge in Africa (5.63 vs 3.59% average), Europe (4.68%) and Latin America (4.72%). Similarly, when segmenting by activity within agribusiness, Food Retail Trading & Processor (4.74%) and Academics & Research (4.69%) demonstrate more concern about food loss. When examining the data by role, self-employed professionals and those from larger companies (over 5,000 employees) (4.50 and 4.17%, respectively, vs 3.41% average) are most concerned about food waste. The social media analysis indicates that there is limited awareness of food waste and losses among the general public. A Harvard agribusiness expert argues that economics suggest that waste will be reduced as resources become scarce, but there is no evidence that point has been reached yet. ■■ Food waste: downstream Food waste at the consumption and retail level is more prevalent in developed regions. There is a clear regional divide with regard to the perceived importance of food waste at the consumer or retail level: professionals in Europe (4.68%) and North America (4.07%) rank it as a much more serious challenge than those in Africa (1.41%) and Oceania (1.40%). Unsurprisingly, those working in food retail (10.00, vs 3.79% average) and trading and processors (6.84%) are more concerned about food waste at the consumer level as well. While the majority of the interviewees consider food waste at the consumption level a challenge, some do not relate it directly to agribusiness. In 2014 the European Commission set a goal of reducing food waste in Europe by 30% by 2025 (EC, 2014a,b), but is already working on a more ambitious target to ‘transform Europe into a more competitive resource-efficient economy ... including waste’ (EC, 2015).

International Food and Agribusiness Management Review

Connolly et al.

Volume 19, Issue 4, 2016

A partner in a strategic consulting firm sees food waste as an opportunity: solving it would save enough food to eliminate hunger – if the logistics of getting the food from the places where it is in to where there are shortages could be resolved. The managing director of a Dutch horticultural trading company pointed to closed loop cycle systems and the need for new technologies that will allow the use of food waste. Infrastructure and investments Infrastructure & investment addresses the capital necessary for agricultural production, from capital (fixed and financial) for all stages of the value chain, to the infrastructure needed for production, storage and transportation. ■■ Infrastructure The respondents identified Infrastructure as one of the most limiting factors of food production and distribution across several regions. Professionals from Latin America and Africa were particularly concerned, citing poor infrastructure as their second biggest challenge (6.87 and 6.34%, respectively, vs 4.51% average). Respondents from larger companies (over 5,000 employees), food retail and the press (4.95, 4.0 and 3.68%, respectively vs 3.42% average) also see poor infrastructure as more of a challenge than those from other segments. On the African continent, getting goods to market can be difficult for farmers, as only a minority of the rural population lives within two kilometers of an all-weather road: 32% in Kenya, 31% in Angola, 26% in Malawi, 24% in Tanzania, 18% in Mali and just over 10% in Ethiopia (Juma, 2012). A number of respondents specifically noted the lack of cold storage and transportation infrastructure among the major challenges for agriculture delivering enough food to feed the world in the future. ■■ Investments The respondents noted a number of investment and finance issues. Lack of access to finance or investors was ascribed to the lack of credit lines for agribusiness and the lack of financial sophistication of players within the sector, while government instability was noted as making investors concerned about long-term gains. For both the private and public sector, the ability to access existing or new lines of credits is a particular challenge, especially in developing regions. A director involved in African business schools notes that ‘capital investment financing is much more of a problem than working capital financing,’ though he notes that both limit industry growth. There is significant regional variation in the responses concerning finance: among professionals in Africa the frequency of finance being cited as a challenge was almost twice the global average (6.34 vs 3.51%), and it was cited as the most significant challenge by 10% of the respondents from Africa. Unsurprisingly, the respondents working in financial markets also consider it significantly more important than respondents from other industry segments (6.78 vs 2.70% average). Markets Market related responses, especially those referring to consumers, were amongst the most frequently cited. Though individually they were not generally ranked as the most challenging issues, combined they represent a substantial challenge. These challenges can be grouped into three categories: capacity, industry structure, and consumer expectations.

International Food and Agribusiness Management Review

Connolly et al.

Volume 19, Issue 4, 2016

■■ Industry capacity A partner in an Indian consulting firm noted that the younger generations in India are not as concerned with religious prohibitions on eating meat. This is consistent with the global trend: as countries get wealthier they tend towards more protein based diets. In turn, protein-based diets require more factors of production, whether technology, natural resources, or capital (including human capital). As a challenge, the capacity of the industry to produce enough protein requires a fresh look at how current and potential protein sources can be developed to feed the world. Consumer preferences will matter: as an Irish agricultural entrepreneur asked, ‘Would the world be prepared to eat protein from insects that have been fed on food waste?’ The head of a large Spanish producer pointed out that the practicalities of producing meat in the volume that the market requires, in ways that consumers demand (for example, treated humanely, with traceability and without antibiotics) requires the development of better technologies and environmentally friendly practices. Nonetheless, although the respondents recognize it as a challenge, they seem to have confidence in the capacity of the industry to produce enough protein. When adjusted for regions, the shift in consumer preference towards more protein was the lowest ranked challenge in this study. Curiously, one of the least concerned sectors was professionals in Asia, where much of the growth in demand is expected (0.63 vs 2.21% average). The self-employed and professionals in small firms respondents didn’t note it as a challenge at all, while professionals working in the financial markets and food retail were the most concerned about the challenge (5.08 and 4.00% respectively, vs 2.72% average). ■■ Industry structure The structural problems to growth within the agricultural sector are both external (access to consumer markets) and internal (poor coordination within the industry). Access to consumer markets was a particular concern for respondents from Asia and Africa (4.18 and 5.63%, respectively, vs 2.77% average). Respondents noted that fragmented markets and lack of transparency lead to inefficient pricing; multiple intermediaries in many markets reduces efficiency and squeeze margins for primary producers; and the lack of a standardized pricing system makes it difficult for smaller producers to compete successfully. A respondent from a development bank in South Africa also noted that inconsistent quality, a lack of volume and other limitations make marketplace access difficult. Some experts, including the partner of a top consulting firm, argue that the advent of urbanization and technology, particularly smart phones, has diminished the magnitude of these challenges, and a director of an industry association notes that new retailing structures are arising in cities in emerging countries, so that over the next two decades opportunities should improve for farmers. However, getting better access to the consumer marketplace is only half the equation: poor coordination within the industry, between producers, processor, and retailers also hampers growth. Moreover, fluctuations and changes in consumers demand can aggravate the challenge. The lack of coordination is an issue both with players operating in one economic activity of the industry and those working across different activities (e.g. coordinating producers with processors with food retail). In both cases there is a zero-sum approach to transactional relationships rather than cooperation for mutual benefit. There was considerable disagreement among respondents as to where in the value chain the problem lies. Some producer and processor respondents noted that food retailers responding to consumer expectations have substantially increased requirements from their supply chain (upstream accountability and documentation, ingredient changes, packaging, etc.) but are not willing to adjust prices. Large processors in both Italy and Sri Lanka noted that these demands by retailers are reducing their margins to unsustainable levels.

International Food and Agribusiness Management Review

Connolly et al.

Volume 19, Issue 4, 2016

Other participants argued that it is large transnational companies that are not connected with other players (particularly producers) in the industry who have required suppliers to squeeze their margins below sustainable levels. Still others attribute the problem to the producers, saying that the producers are not well connected to one another and frequently are not capable of responding effectively to supply chain requirements. Although one food processor argued that companies are only competing for the same consumers and this is the natural dynamics of a free and competitive market, a European investment advisor pointed out that ‘contradictory information driven by different lobby groups may be detrimental to the industry as whole’. This disparity of views is emblematic of the lack of coordination of players within agribusiness. When the data is adjusted for region, it was the second highest ranked challenge (6.25% average), with the least variability across different regions (2.27%). ■■ Consumer expectations Consumer expectations of agribusiness in general and food preferences in particular have been changing rapidly, along with the platforms for expressing those views. A demand for more natural and healthy food options, concurrent with consumer opposition to genetically modified organisms (GMOs) and other forms of biotech innovation has created considerable communication challenges for agribusiness. CRISPR-Cas9 may offer some alternatives to classic transgenic options (Waltz, 2016). The demand for more natural food options can be seen in a recent Food Label Survey (Consumer Reports National Research Center, 2015) in the US that found that consumers say they search for foods that are locally grown (66%), natural (59%), free from artificial growth hormones (50%), pesticide free (49%), or organic (49%). However, most of the respondents agreed that, given the currently available technology, the planet does not have the resources to produce enough food to feed everyone using only non-GMO, natural and organic production techniques. Some noted that economics will play a role in preventing these requirements from becoming stricter, as food produced meeting all of these requirements will become more and more expensive. Although the population segments requiring food to be non-GMO, organic and natural are small, they tend to be more educated, wealthier and more voluble, thus more likely to affect policy making. Some of the experts noted that if policies shift further in those directions, it will make feeding the world of the future –especially the poorer segments – even more challenging. Professionals in North American, Latin American and Africa were more concerned about this challenge (5.52, 5.15 and 4.93%, respectively, compared to 4.15% average), along with professionals with fewer years of experience in agribusiness (6.14 vs 4.28). Finally, retail food and press professionals were more concerned with changing consumer expectations (10.00 and 6.67%, respectively, vs 4.81% average). On the other hand, NGO professionals were notably less concerned (2.22%). Many of the approaches that agribusiness is using to try and improve the availability of food, including biotechnology innovations in GMOs, feed additives, clones, etc. have been met with active resistance in the market place. Professional respondents from North America were markedly more concerned with consumer acceptance of biotechnology than those from other regions (9.14 vs 5.22% average) and it was the most frequently mentioned concern, with more than 10% ranking it their top concern. On the other hand, none of the professional respondents from Africa noted acceptance of biotechnology as a primary challenge, and less than 1% mentioned it all. Professionals working with farm inputs, academics, and researchers (who are responsible for much of the new biotechnology of the industry), are somewhat more concerned than average (7.58 and 7.87% respectively, vs 5.77% average) with how consumers accept or resist/delay new technology.

International Food and Agribusiness Management Review

Connolly et al.

Volume 19, Issue 4, 2016

The opposition to GMO foods amongst consumers is widespread. A recent survey (Pew Research Center, 2015) published in collaboration with the American Association for the Advancement of Science (AAAS) concluded that the 57% of the general public says genetically modified (GM) foods are unsafe to eat while only 37% claim it is safe. In contrast, 88% of AAAS scientists assert GMOs are safe to eat. At least one third of the interviewed experts, from around the world, spontaneously mentioned opposition to GMOs among the top 3 challenges. An expert with experience in Ghana noted that there is a ‘huge anti-GMO’ movement in Africa, and an expert from U.C. Davis noted that varieties of GM cassava that would improve harvest rates substantially are available in Africa, but have not been approved by regulatory authorities because of fears over safety. A director from an investment bank suggested that ‘if the first GMO products directly benefited consumers, with higher nutritional values, instead of farmers, with higher yields and lower costs, the outcome of public opinion might have been different’. As it is, consumer groups have managed to dominate the conversation and agribusiness has failed to deliver its own message clearly. To some extent this reflects agribusiness’ low level of credibility with consumers, who see some well-known names in the industry as having a conflict of interest, and motivated solely by profit, with little regard for possible risks, and benefits to consumers as secondary priority. Some respondents object that many of the loudest consumer voices are poorly informed and share inaccurate or misleading information. Whatever the truth of these claims, it is clear that communication between consumers and agribusiness has not been successful. One of the interviewees, a food retail owner and television show host, invited a large biotechnology company to their televised food show to participate in a debate about GMO foods, but the firm declined to participate. Respondents from an industry association, a major US food retailer, a large Latin American food processor and a French cooperative all noted that the agribusiness sector has not done a good job communicating with consumers regarding the use and constraints around factors of production such as new technologies, resource, available infrastructure and others. Other interviewees, including a director of an international institute, and a former CEO of an international food NGO, argue that the industry has done a good job communicating with consumers but that given the lack of credibility within the industry there is not much that can be done to affect consumer attitudes. The advent of CRISPR-Cas9 may offer new options for agriculture, the ability to edit genes in a manner which is perceived as consumer friendly and similar to the mechanisms used by the organism in nature to correct gene deficiencies (Connolly, 2016). CRISPR offers producers the potential to achieve similar or better responses to GMO technologies without the regulatory hurdles (Waltz, 2016). Farm input and press professionals are more concerned about the challenge of consumer communications (5.27 vs 3.89% average), while NGOs and government professionals are less concerned about inefficient communication (2.22 and 1.67%, respectively). When segmenting for company size, professionals from companies with fewer than 200 employees identify communication as more of a challenge than those from companies with more than 5,000 employees or the self-employed (5.01, 3.85 and 1.80%, respectively, vs 4.14% average). Governments are realizing that communications with consumers is an important issue and are trying initiatives to bridge the gap, such as the ‘know your farmer, know your food’ program for supporting local food systems to create jobs and boost economic growth created by the USDA in 2009 (USDA, 2016). People The biggest difference between the first and second GLIMPSE studies is the emphasis on people. In the original study adequate qualified workers were noted as a potential problem for the future, but were not seen as a particularly significant challenge. In just four years it has become an important challenge. There are International Food and Agribusiness Management Review

Connolly et al.

Volume 19, Issue 4, 2016

two related aspects: a shortage of an adequate, qualified workforce, and the difficulty in attracting talented young people to agribusiness. ■■ Lack of qualified workforce The majority of respondents view the challenge of having a qualified workforce as significant: approximately 20% of the respondents put the challenge of getting qualified workers in their top five challenges (though rarely as the top challenge). Variability across regions is low (1.5%). A researcher from a South African agricultural council noted ‘there are a lot of people who are willing to work on farms but who are not qualified’. An executive from a large Brazilian cooperative concurred saying, ‘There is a significant contingent of people (in agriculture) demanding training’. The president of a US poultry company argues that farming requires a broad set of skills, and that set of skills is becoming more specialized as new technologies are transforming all segments of the agribusiness value chain. A study from Rezende et al. (2009) found that the lack of qualification in the workplace derives from informality, seasonality and high turnover of employees. ■■ Attracting young people to agribusiness The head of an industry association, and owner of a large poultry operation, commented that it has never been so difficult to hire people to work in agriculture. He noted that the greatest challenge in a recent new project was to find 800 people to work in it, because ‘people don’t want to work in agriculture’. Some experts argue that agriculture and food production are going through a ‘talent renaissance’ with the increase in use of technology and the growing awareness of the sector by population at large. They believe that the rise of automation in agriculture will require more jobs indoors than in fields, attracting a greater number of new and young people than in the past. However, although a US professor has seen a sharp increase in college applications to agriculture-related degrees, there is still a significant shortage of qualified people to fill jobs in agriculture in North America. There is an agreement among experts that agribusiness still demands more talent than what is available. People with less than 10 years of experience in agriculture find the challenge of attracting new talent substantially more important than those with more than 50 years (7.16 and 3.67%, respectively, vs 5.72% average). Overall it was the seventh highest ranked challenge, with more than 20% of respondents mentioning it. While more than one in eight professionals working in farm input companies ranked the challenge as most significant, self-employed professionals did not see the challenge as relevant. A subset of this challenge is business succession. Many farmers and agribusiness owners are facing difficulties in the succession of their business as the younger generations are not as attracted to the industry. According to Rabobank’s analysis (2015), the average age of farmers in the USA increased from 45 in 1974 to 58 in 2007. The same study found that in the USA there are seven times as many farmers over age 75 as there are under age 25. Australia is seeing similar demographic changes: the average age of farmers in Australia increased from 44 to 56 over the past 30 years. Experts from South Africa, Ireland, Brazil, the USA and several other countries spontaneously mentioned succession as a significant challenge of the industry. A counter argument, presented by the head of a Chinese biotechnology company, is that succession issues will be resolved by changes in ownership structures. Science and innovation Although the category of Science and Innovation covers a broad spectrum, for the respondents the challenges presented can be generally categorized as challenges of access to needed technology, resistance to innovation, and needed innovations.

International Food and Agribusiness Management Review

Connolly et al.

Volume 19, Issue 4, 2016

■■ Access to technology Poor productivity was blamed in part on a lack of access to technology, particularly in Asia and Africa (6.90 and 5.63%, respectively, vs 4.32% average). Reducing the gap between high performers and low performers is key to increasing food production. Low performers typically have land, natural resources and sometimes people, but are less likely to have access to technologies such as crop or animal genetics, nutrition and management, irrigation, automation, and many other resources. The lack of access can be attributed to insufficient expertise (the producer may not even be aware of technological resources), local availability, an adequate ecosystem, finance, government limitations and others. Experts from Peru, South Africa, Ukraine and India commented on the lack of access to technology in their countries. A partner from a strategic consulting firm commented that India, China, Niger, Ethiopia, and Ukraine, among others are all producing at 25 to 30% of the developed world yields. Another study concluded that the average irrigated and rain-fed maize yield in China has about half the yield achieved using technology typically used in other countries (Meng et al., 2013). ■■ Resistance to innovation The range of views about the importance of this challenge was particularly high. Many of the respondents considered the challenge irrelevant while others saw it as a significant issue. One banking executive in Brazil noted that expenditure in food production research globally is about the same today as it was in 1970, with no adjustment for inflation. The president of an animal feed company in the USA commented, ‘Some farmers in the Midwest have not been farming for 20 years. They farmed once and repeated themselves for another 19 years’. The director of a development bank in South Africa argued that if technology is economically available, agribusinesses will adopt it. Others believe that current levels of technology and adoption are acceptable. The views of respondents who believe that the current rate of new technology evolution is not sufficient to support the needed growth in agricultural production and productivity were compared to those of respondents who believe farmers are not willing to adopt new technology or do so in a slower than optimal manner. People with less than 10 years' experience in the industry believe that adequacy of technology is less of an issue than resistance, while those with more than 50 years of experience see the issue as being more about the adequacy of the technology (2.78 and 3.76%, respectively). Of the professional respondents, those associated with NGOs were notably more concerned about the adequacy of with technology production and adoption (6.67 vs 3.43% average). ■■ Needed innovation Disease outbreaks and pest infestations are chronic challenges in agribusiness that have been exacerbated by both agricultural practices and the globalization of the food chain. Avian Influenza is a striking example of the challenge: the H5N1 strain was first identified in Asia in 2003 and by 2005 it had reached Europe and the Middle East, followed by Africa by 2006. In one week in 2015, outbreaks were reported in Taiwan, Vietnam, Indonesia, Ghana, Ivory Coast, Nigeria, Palestinian Territories, Mexico, and the USA (National Wildlife Health Center, 2015). The outbreak in the USA killed more than 48 million birds, representing 10% of total national turkeys, and 40% of Iowa’s laying hens (USDA, 2015). Examples of other high profile outbreaks include bovine spongiform encephalopathy (BSE, or mad cow disease), pig virus, E. coli, and foot and mouth disease. One respondent, a member of the World Agricultural Forum and former executive of a major poultry processor, cited these sorts of biosecurity threats as the number one treat to agribusiness in the future. Many of these outbreaks result from current production methods and technologies, as well as expanded trade of living animals. Effective control and response mechanisms depend on implementation by people International Food and Agribusiness Management Review

Connolly et al.

Volume 19, Issue 4, 2016

throughout the food value chain, and regulatory reaction to a single incident can change the larger operating environment. The president of a large poultry operation noted that Salmonella contamination of one producer’s birds lead the US government to change policies for the entire country. Crop protection, particularly rapidly evolving resistance to pest control products is another issue. A Brazilian executive noted that there have been few innovations in crop protection. Overall, biosecurity was one of the top 5 biggest challenges cited by respondents, particularly those from Oceania, the USA and Canada (7.48 and 7.42, vs 6.17% average). Professionals working in trading companies (7.89%) and consulting (8.13%) also consider it relatively important (5.95% average). Environment Scarcity and depletion of natural resources (especially, but not only, land or water) as well as changing regulations and population patterns all present challenges to the agribusiness sector. ■■ Water The World Economic Forum’s Global Risks 2015 Report (WEF, 2015) lists water crises as the top global risk in terms of impact, and eighth in terms of likelihood of severe crises occurring. Water scarcity is, by a substantial margin, the most frequent challenge listed and discussed by study respondents. One third of the interviewees in Phase 1 spontaneously mentioned water as being one of the greatest challenges for agribusiness. 20% of the respondents listed scarcity of fresh water as the number one challenge, and nearly half considered it among the top 5 challenges. However, there is some regionality in the responses. Professionals from North America and Oceania are relatively more concerned with the challenge (12.04 and 11.21%, respectively, vs a global average of 9.10%). Indeed, a third of the professionals with experience in Oceania listed water as the number one challenge. On the other hand, professionals from Asia and Latin America (6.69 and 6.44%, respectively) were comparatively less concerned with water scarcity. In contrast to the concerns about innovation, it was the more experienced respondents (those with more than 40 years of experience) who were more concerned about water, than those with less experience (under 10 years) (11.31 and 7.89%, respectively, vs a 10.38% average). NGOs, financial markets and press professionals were more concerned about water than other sectors. Although most participants agree that technology will be the solution to overcome water scarcity, they do not believe that solutions will be developed in the near future. A partner from a consulting firm gave the example of desalinization facilities in Australia that remain unused due to high operating costs. Other interviewees suggested that water pricing is a viable short-term solution, and not just at the producer end. Consumers could be charged for the water cost of foods, which would in theory increase demand for more efficient production methods. ■■ Land For professionals involved in primary production, land availability is second only to water scarcity as a challenge, with professionals newer to the industry the most concerned. Overall, availability of additional land for food production was ranked as the third most serious challenge. Professionals from Asia and Africa were more concerned than those from Latin America. Land availability challenges in Asia are generally related to population, while those in Africa are more often related to land ownership and land reform. One respondent from South Africa noted that land is mostly in

International Food and Agribusiness Management Review

Connolly et al.

Volume 19, Issue 4, 2016

the hands of people who are not looking into further developing agricultural productivity. These owners have been granted the access to the land and do not have any particular incentive to increase their productivity. According to the Sustainable Europe Research Institute (2013) more than 75% of the land on earth (excluding Greenland and Antarctica) is already being used by humans, including urban areas (1%), cropland (12%), forestry (27%) and grazing land (36%). About half of the remaining land is unproductive land. Unless agricultural productivity increases, to meet the growing demand for food approximately 6 million additional hectares of land need to be converted to agricultural production every year until at least 2030 (ELD, 2013). However, respondents noted that conversion of land to agriculture faces a number of obstacles, including concerns about deforestation and release of carbon. ■■ Other environmental challenges Respondents cited scarcity of fertilizers, energy costs, poor waste management that contaminates and/or depletes other natural resources amongst other environmental challenges. An Australian expert in seafood retail noted that availability of salt water reserves for seafood farming is a challenge. Few respondents from Phase 2 (2.96%) specified ‘other’ challenges, but professionals from Europe were more likely to specify concerns than those from other regions (4.26 vs 3.00% average), as were respondents with fewer than 20 years in the field (3.37 vs 2.40% average).

5. GLIMPSE: revisited and revised The updated research largely validated the original GLIMPSE work, but found areas where updating or expansion was appropriate, as well as some opportunities for realignment. Drawing on the results from Phase 1 some adjustments were made to the category definitions. The most significant difference from the original GLIMPSE framework revolve around challenges related to people: the ‘lack of qualified or educated workforce’ and ‘new and young talents not being attracted by the industry’. Human capital was implicitly encompassed under various other challenges in the original study, but the present research prompted the creation of a new People category. The original GLIMPSE framework treated Policies and Government as separate categories, but the findings from the current work confirmed how interdependent these two categories are, so they have been consolidated into one category. The second significant adjustment is the specific inclusion of Consumers into the Markets element, reflecting the increasing importance of consumer interaction. Consumers’ preferences, expectations and requirements are changing, and the flow of information between them and participants in the food chain is stronger and more public than in the past. Other minor adjustments are the inclusion of Investments to Infrastructure under the letter I, to account for the fixed or financial capital needed for improvements in production, and the specific inclusion of food loss, reflecting the magnitude of the loss as a proportion of agricultural production is affected. These adjustments bring the GLIMPSE framework into alignment with the findings from the current research and incorporate all of the major challenge categories identified through the research.

6. Conclusions Through an iterative, grounded theory methodology the original research that led to the GLIMPSE framework was validated, deepened and expanded. The key challenges were identified though open-ended interviews with industry experts, then confirmed and classified across industry segments, with interview and survey input from almost 600 agribusiness professionals from 53 countries. The challenges identified and refined in International Food and Agribusiness Management Review

Connolly et al.

Volume 19, Issue 4, 2016

these steps were then tested for relevancy and accuracy through the analysis of more than one million social media posts. The findings from this work were then reviewed against the original GLIMPSE framework, and some adjustments made to the framework to reflect both the changes over time and the findings of the larger study. The current research improves the characterizations of the GLIMPSE categories and confirms that the challenges encompassed by the acronym are comprehensive. It also confirms the complexity of the challenges: the challenges are so interlinked that they cannot be analyzed on a standalone basis. The framework offers industry participants a useful tool for identifying those areas over which they have influence, thus helping to reduce uncertainty and measure risk. The challenge of feeding a population that is rapidly increasing both in numbers and income is complex and contradictory. It will require aligning factors of production, agribusiness and society. Four of the GLIMPSE challenges relate to production factors: Infrastructure & Investments, People, Science & Innovation, and Environment. The other three challenges reflect the challenges of coordination between agribusiness and key stockholders such as governments and consumers. GLIMPSE offers a framework for identifying and understanding these challenges and provides the agribusiness sector with identification tools for addressing these problems.

References Bill and Melinda Gates Foundation. 2012. Annual Letter. Available at: http://tinyurl.com/jgm4hve. Boehlje, M., M. Roucan-Kaneb and S. Bröringc. 2011. Future Agribusiness challenges: strategic uncertainty, innovation and structural change. International Food and Agribusiness Management Review 14: 53-82. Bourgeois, R. 2014. Food (in)security: the new challenges ahead. document de travail ART-Dev, Art-Dev. Bryman, A. 2012. Social Research methods, 4th ed. Oxford University Press. Oxford, UK. Connolly, A.J. 2016. A CRISPR Opportunity: is this the end of transgenic GMOs in our food? Available at: http://tinyurl.com/gruhm92. Connolly, A.J. and K. Phillips-Connolly. 2012. Can agribusiness feed 3 billion new people ... and save the planet? A GLIMPSE™ into the future. International Food and Agribusiness Management Review 15: 139-152. Consumer Reports National Research Center. 2015. Natural food labels survey – 2015 nationally-representative phone survey. Available at: http://tinyurl.com/j7duxk2. Davis, J.H. and R.A. Goldberg. 1957. A concept of agribusiness. Division of Research, Graduate School of Business Administration, Harvard University, Boston, MA, USA. Dobbs, R., J. Manyika and J. Woetzel. 2015. No ordinary disruption: the four global forces breaking all the trends. Public Affairs Books, New York, NY, USA. Economics of Land Degradation Initiative (ELD). 2013. A global strategy for sustainable land management. The rewards of investing in sustainable land management. ELD, Bonn, Germany , p. 12. European Commission (EC). 2014a. Communication from the Commission to the European Parliament, the Council, The European Economic and Social Committee and the Committee of the Regions Towards a circular economy: a zero waste programme for Europe. /* COM/2014/0398 final/2 */. . Available at: http://tinyurl.com/hf4fme4. European Commission (EC). 2014b. Proposal for a Directive of the European Parliament and of the Council amending Directives 2000/53/EC on end-of-life vehicles, 2006/66/EC on batteries and accumulators and waste batteries and accumulators, and 2012/19/EU on waste electrical and electronic equipment. COM/2015/0593 final – 2015/0272 (COD). Available at: http://tinyurl.com/o6jlym4. European Commission (EC). 2015. Circular economy strategy. Available at: http://tinyurl.com/mrgoa47. Glaser, B.G. and A.L. Strauss. 1967. The discovery of grounded theory: strategies for qualitative research. Aldine Transaction, New Brunswick, NJ, USA. Juma, C. 2012. Poor Infrastructure is Africa’s soft underbelly. Forbes. Available at: http://tinyurl.com/zzj9utw.

International Food and Agribusiness Management Review

Connolly et al.

Volume 19, Issue 4, 2016

Lakner, Z G.A. and Baker. 2014. Struggling with uncertainty: the state of global agri-food sector in 2030. International Food and Agribusiness Management Review 17: 141-176. Lipinski, B., C. Hanson, J. Lomax, L. Kitinoja, R. Waite and T. Searchinger. 2013. Installment 2 of ‘creating a sustainable food future – reducing food loss and waste’. World Resources Institute, Washington, DC, USA. McKenzie, F.C. and J. Williams. 2015. Sustainable food production: constraints, challenges and choices by 2050. Journal of Food Security 7: 221-233. Meng, Q., P. Hou, L. Wu, X. Chen, Z. Cui and F. Zhang. 2013. Understanding production potentials and yield gaps in intensive maize production in China. Field Crops Research 143: 91-97. National Wildlife Health Center. 2016. Avian Influenza in poultry. Available at: http://tinyurl.com/ycbav7. Pew Research Center. 2015. Public and scientists’ views on science and society. Available at: http://tinyurl. com/m7ktyng. PwC. 2015. Agribusinesses Insights Survey 2014/2015. Africa – are you in for the ride? Available at: http:// tinyurl.com/hy8b5zv. Rabobank. 2015. Succession: keeping the family farm alive. Available at: http://tinyurl.com/h9pk4d3. Rezende, G.C., L.R. Ferreira and A.C. Kreter. 2009. Labor legislation and its adverse impacts on transaction costs in Brazilian agriculture. Working Papers 06/2009. Columbia University – ILAS, New York, NY, USA. Sustainable Europe Research Institute (SERI). 2013. Land footprint Scenarios, p. 13. Available at: http://tinyurl.com/jt65s5v. United States Department of Agriculture (USDA). 2016. Every family needs a farmer. Available at: http:// tinyurl.com/haw5s8n. United States Department of Agriculture (USDA). 2015. Update on avian influenza findings: Poultry findings confirmed by USDA’s National veterinary services laboratories. Available at: http://tinyurl.com/ j5wu4q7. Vervoort, J.M., P.K. Thornton, P. Kristjanson, W. Forch, P.J. Ericksen, K. Kok, J.S.I. Ingram, M. Herrero, A. Palazzo, A.E.S. Helfgott, A. Wilkinson, P. Havlik, D. Mason-D’Croz, D. and C. Jost. 2014. Challenges to scenario-guided adaptive action on food security under climate change. Global Environmental Change 28: 383-394 Waltz, E. 2016. Gene-edited CRISPR mushroom escapes US regulation. Nature. Available at: http://tinyurl. com/jv56vq7. Wheeler, T. and J. Von Braun. 2013. Climate change impacts on global food security. Science 341: 508-513. World Economic Forum (WEF). 2015. Global Risks 2015. Available at: http://tinyurl.com/h77d8cr.

International Food and Agribusiness Management Review

OPEN ACCESS International Food and Agribusiness Management Review Volume 19 Issue 4, 2016; DOI: 10.22434/IFAMR2015.0219 Received: 21 December 2015 / Accepted: 5 October 2016

Members’ attitudes towards cooperatives and their perception of agency problems RESEARCH ARTICLE Karin Hakelius

and Helena Hanssonb

aAssistant

professor and bAssociate professor, Swedish University of Agricultural Sciences, Department of Economics, P.O. Box 7013, 750 07 Uppsala, Sweden

Abstract This study examines whether and how members’ perceptions of agency problems, in terms of the decision problem and the follow-up problem, shape their attitudes to agricultural cooperatives. The study is based on empirical data collected through a postal questionnaire sent to 2,250 Swedish farmers in 2013 (response rate ~40%). Exploratory factor analysis of a set of attitudinal measurement items was used to assess members’ attitudes to agricultural cooperatives. Seemingly unrelated regression analysis was used to identify the impact of members’ perceptions of agency problems on the attitude measures obtained from the exploratory factor analysis. The results suggest that perceived agency problems significantly explain members’ attitudes to their cooperatives. Therefore, working with these problems can be a way for directors of cooperatives to influence members’ attitudes and, in continuation, behaviors to these. This would be one way of developing more sustainable member-director relationships in these cooperatives. Keywords: agency problems, agricultural cooperative, attitude, governance, factor analysis JEL code: P13, Q13 Corresponding author: karin.hakelius@slu.se

Hakelius and Hansson

Volume 19, Issue 4, 2016

1. Introduction The food sector is changing profoundly in the Western world. Increasing internationalization of the food retail industry, increased competition, increased concentration in the food manufacturing sector, changes in consumer buying and consumption patterns, structural change among farms leading to fewer, larger units, and agricultural support for farms that is increasingly decoupled from actual production are some significant components of the ongoing change. Naturally, these changes also affect agricultural cooperatives, many of which are currently adjusting through mergers and acquisitions in order to become stronger market actors. When agricultural cooperatives become larger, the distance between members and directors in terms of possibilities to meet and enter into dialogue tends to increase. This creates unique problems, since cooperatives are governed by members through the directors, using a system of representative democracy, and not through a market in shares (e.g. Barbaud-Didier et al., 2012; Bijman et al., 2013, 2014; Chaddad and Iliopoulos, 2013; Cornforth, 2004; Liang and Hendrikse, 2013; Österberg and Nilsson, 2009). In order to handle the growing distance between members and directors, cooperatives typically introduce intermediary bodies with a representative function, such as councils and regional representatives. Furthermore, when agricultural cooperatives become larger, the heterogeneity among members is likely to increase, a development accelerated by the ongoing specialization at farm level (e.g. Ahearn et al., 2005; Balmann et al., 2006; Happe et al., 2008; Pietola et al., 2003). Regarding the developments concerning growing distance between members and directors and the increasingly heterogeneous membership, the scientific literature distinguishes two types of agency problems related to the governance of agricultural cooperatives: the decision problem and the follow-up problem. The decision problem emerges as it becomes increasingly difficult for directors to ‘read’ members’ wants and needs and work towards optimizing the long-term well-being of the cooperative and its members when the member group grows (e.g. Cook and Iliopoulos, 2000; Fulton and Giannakas, 2001; Hendrikse, 2007). Within the larger member group, the diversity of wants and needs expressed by members increase, as does the heterogeneity in terms of risk attitude in the member group. Over time, this development may lead to a situation where members feel that the directors are not taking their interests into consideration. The follow-up problem arises because with increasingly larger agricultural cooperatives, it becomes more difficult for members to monitor what directors do, resulting in decreased loyalty from members (Richards et al., 1998: 22). In many cases, when the cooperative grows, the organizational structure becomes more complex. In Sweden, as in many other countries, it is common to see a structure of a cooperative association functioning as a traditional cooperative association. The cooperative then has a number of investor-owned firm (IOF) subsidiaries in which production and marketing take place. These entities follow the logic of IOFs, aiming at profit levels determined by the board of the cooperative. Needless to say, monitoring such an organization is far more complicated than monitoring a small cooperative. The complexity lies mainly in the size itself, but another complicating factor is the difference in business goals in a cooperative compared with an IOF. In the IOF case, maximum profits are the objective and the main focus is to achieve this. In a cooperative, the focus is on generating the highest residual level for the members, meaning that additional considerations have to be added to doing well on the markets, for example better service and prices (Nilsson, 2001). Decreasing member loyalty is a problem for agricultural cooperatives and many researchers have even concluded that having loyal members is crucial for the success of the cooperative (Cook, 1994; Cechin et al., 2013; Fulton, 1999; Österberg and Nilsson, 2009). The foundations of loyalty to a cooperative are trust and commitment or, put differently, members’ positive attitudes to the governance system of agricultural cooperatives create loyalty and are of the utmost importance for development of the cooperative (Barraud-Didier et al., 2012; Cechin et al., 2013; James and Sykuta, 2005; Nilsson et al., 2009; Österberg and Nilsson, 2009). The incentive structure for a cooperative director is different from that of an IOF director, as explained by Cook (1994) and Hendrikse (2007). For example, the task of a cooperative director is to act as an agent, or representative, for the members, simultaneously taking into consideration market developments, and wisely International Food and Agribusiness Management Review

Hakelius and Hansson

Volume 19, Issue 4, 2016

govern the cooperative, in order to generate as large a dividend as possible for the members. In addition, the director has to maintain dialogue with members, listening to their wants and needs and explaining decisions made by the cooperative board. Hence, directors have to understand the dual role they hold, i.e. representing members and governing the cooperative, and try to combine these, leading to the best possible result. One way of describing this is to take into account the ‘people factor, such as member attitudes’ (Bhuyan, 2007: 276), focusing on creating trust among members (Borgen, 2001: 209), in addition to the economic factors. When members take part in governing the cooperative, their decision making in relation to the cooperative is central. Taking a behavioral perspective to understanding members’ actions in relation to the cooperative, the importance of members’ attitudes to the governance system of agricultural cooperatives is further emphasized. According to the Theory of Planned Behavior (Ajzen, 1991, 2002), attitudes and other subjective norms and perceived behavioral controls lead to a behavioral intention, which in turn leads to a certain behavior, or decision, under certain circumstances. Hence, from a behavioral perspective, attitudes are recognized as one type of determinant of behavior (e.g. Ajzen, 1991, 2002; Conner and Abraham, 2001; Fazio and Olson, 2003; Feist, 2012; Kaiser, 2006; Kaiser and Sheuthle, 2003; Siegel Levine and Straube, 2012). From this, it follows that members’ attitudes to the governance system of agricultural cooperatives are one type of determinant of their actions with respect to these cooperatives. In attempts to understand how members’ perceptions of agency problems may affect their actions with respect to agricultural cooperatives, understanding how their attitudes to these cooperatives are shaped by their perceptions of agency problems would be one important step. However, to the best of our knowledge, this has not been studied previously. The aim of this study was thus to identify whether and how members’ perceptions of agency problem, in terms of the decision problem and the follow-up problem, shape their attitudes to agricultural cooperatives. The study is based on a dataset obtained through a 2013 postal survey sent to a sample of farmers in Sweden who provided information about their attitudes to the governance system of their agricultural cooperatives and their perceptions of agency problems in these cooperatives. In a novel contribution to previous literature, this study thus provides insights into how members’ attitudes to the governance system of their agricultural cooperatives are shaped by their perceptions of agency problems. Such insights are important for understanding how positive or negative attitudes regarding the governance system of agricultural cooperatives are shaped. The findings presented here can thus be used by cooperative directors to develop the governance system and information channels in cooperatives, which can lead to decreased agency problems and possibly increased loyalty, created through increased trust and commitment among members. The paper continues with section 2 presenting the conceptual framework upon which the analysis is based. Data and methods used are presented in section 3 and the results in section 4. Section 5 comprises a discussion and some conclusions.

2. Conceptual framework A behavioral framework was applied to examine how farmers’ attitudes are related to their perceptions of agency problems generated due to the characteristics of the governance system of agricultural cooperatives. Attitudes in general and farmers’ attitudes to agricultural cooperatives Attitudes represent summary evaluations, in terms of liking, disliking, or indifference, of psychological objects (Ajzen, 1991; Kahneman and Sudgen, 2005). This means that attitudes constitute an individual’s idea about an object (Kahneman and Sudgen, 2005). Together with other variables, attitudes represent one type of determinant of human behavior (e.g. Ajzen, 1991, 2002; Conner and Abraham, 2001; Fazio and Olson, 2003; Feist, 2012; Kaiser, 2006; Kaiser and Sheuthle, 2003; Siegel Levine and Straube, 2012). Therefore, attitudes are also one type of antecedent of directors and members (in our case farmers) decision making (and behaviors) in businesses, and understanding determinants of these attitudes would be one step in understanding International Food and Agribusiness Management Review

Hakelius and Hansson

Volume 19, Issue 4, 2016

decision making. As attitudes are formed by people’s beliefs, there is a causal relationship from beliefs to attitudes and on to decision making (Ajzen, 1991). According to the Theory of Planned Behavior (Ajzen, 1991, 2002), the last step is mediated through behavioral intent, which is also formed by subjective norms and perceived behavioral controls. Within the area of risk attitudes, Weber et al. (2002) introduced the notion that attitudes may be unique to a specific domain. This means that people’s attitudes to a general object can be better understood if analyzed for the different domains covered by this general object. In the context of the governance system in agricultural cooperatives, domains of specific interest are trust and commitment and these are thus the focus of the attitude construct in this paper. ‘Trust’ is about feeling that one is not being exploited by others (James, 2001) and has been found to be crucial in a cooperative context, both in terms of trust between members and directors and in terms of trust among members (James and Sykuta, 2005, 2006). Existing definitions of trust include affective and cognitive dimensions (Hansen et al., 2002). In this study, we focus on the cognitive dimensions of trust, following findings by Hansen et al. (2002) that in cooperatives with complex businesses, this trust dimension is the more important of the two. ‘Commitment’ relates to satisfaction and loyalty (Fulton and Adamowicz, 1993; Gray and Kraenzie, 1998; Österberg and Nilsson, 2009) and has been defined as ‘the preference of co-op members to patronize a co-op even when the co-op’s price or service is not as good as that provided by an investor-owned firm’ (Fulton, 1999: 423). Agency problems In Agency Theory (e.g. Eisenhardt, 1989; Fama and Jensen, 1983; Laffont and Martimort, 2002), which is frequently used to analyze governance of organizations, the principal hands over the responsibility for performing a certain task to the agent. Once this happens, the principal loses the control over the situation and therefore controlling the agent becomes important. In an IOF setting, the shareholders act as the principal and the board as their agent (cf., Cook and Burress, 2013; Eisenhardt, 1989; Fama, 1980; Fama and Jensen, 1983; Hansmann, 1996). As long as the agent acts in a satisfactory way in the eyes of the principal, there is a high probability that the agency relationship will be beneficial to the principal, and hence continue. Should some shareholders not be satisfied with developments, they can sell their shares and the shares are valued on an open market, thereby offering a value to the shareholders. In the agricultural cooperative setting, however, the producer becomes a member of the cooperative, thereby entering into an agency relationship in which the directors (the agent) should govern the cooperative in the interest of the members (the principal). In the case of cooperatives of the Swedish type, dissatisfied members seldom have the possibility to exit the cooperative and find an alternative actor to trade with. From the directors’ perspective, decisions made follows a different logic than the one used by IOF directors, due to the difference in incentive structure mentioned above. This leads to a crucial relationship between the members and the directors, and therefore it is interesting to study the agency relationship between members and directors. As mentioned above, this arrangement leads to certain problems, labelled ‘agency problems’, which may decrease the output of the cooperative activity. In cases when non-member, or external, directors sit on a cooperative board, then these directors do not have the same strong connection to the members. This could, arguably, be compared to IOF directors. In the Swedish case, however, having external directors on cooperative boards is almost non-existent and therefore we do not include this aspect in our study. Agency problems are significant in agricultural cooperatives (Borgen, 2001; Feng and Hendrikse, 2012; Fulton and Larson, 2009; Richards et al., 1998) and are sometimes labelled ‘vaguely defined property rights’ problems (Cook, 1995: 1156, see also Cook and Iliopoulos, 2000; Nilsson et al., 2012), due to the specific way in which a cooperative is owned and managed (Bijman et al., 2013, 2014; Chaddad and Iliopoulos, 2013; Cornforth, 2004; Liang and Hendrikse, 2013; Österberg and Nilsson, 2009). These problems are believed to originate from specific features of the member-producer cooperative relationship, such as differences in International Food and Agribusiness Management Review

Hakelius and Hansson

Volume 19, Issue 4, 2016

planning horizons, investment portfolios and attitudes to the unallocated capital in cooperatives (e.g. Borgen, 2004; Nilsson and Svendsen, 2011; Novkovic, 2008). In this study, as mentioned above, we focus on the two agency problems that have their origin in the division of governance roles in a cooperative (between the principal/members and the agent/directors). It was also mentioned that the decision problem involves the directors’ problems in listening to and processing the requirements of the members, bearing in mind that decisions made by directors influence the profit levels at farm level, due to farms being partly vertically integrated with the cooperative. The follow-up problem relates to the difficulties the members face when trying to understand what is happening within the cooperative and on the markets where the cooperative acts. This problem is complicated by the fact that there is no market for tradable and appreciable residual rights to use as a tool to measure the success of the cooperative (e.g. Fama, 1980). Both these types of agency problems are associated with monitoring problems for the key actors in the cooperative collaboration, i.e. the members and directors (cf., Fulton and Giannakas, 2007; Nilsson and Svendsen, 2011; Richards et al., 1998). By integrating attitude research (e.g. Ajzen, 1991, 2002) with insights from the literature on agency problems (e.g. Feng and Hendrikse, 2012; Richards et al., 1998), the presumption in this study is that members’ perceptions of agency problems, in terms of the decision and follow-up problems, is one (or the major) antecedent of their beliefs about the cooperative. This will in turn members’ attitudes to the cooperative. Thus, the presumption is that there is a link between farmers’ perceptions of agency problems and their beliefs about the cooperative, and thereby to their attitudes to the cooperative governance system. Therefore, understanding members’ perceptions of agency problems would be one step in understanding how their attitudes to the governance system of agricultural cooperatives are formed, which in turn would be one determinant of their decisions (or behaviors) with respect to committing themselves to the governance system of the cooperatives.

3. Data and methods Study sample This study is based on empirical data collected by a 2013 postal questionnaire sent to a sample of 2,250 Swedish farmers. In the Swedish official statistics, a farm has to be 2 hectares or larger, and hence the sample was generated according to this limit. The 2013 postal questionnaire was part of a larger project, including comparisons to data collected in 1993. Therefore, the sampling procedures used in the original study from 1993 were followed, in which three strata were generated (2-20 ha, 21-50 ha, ≥51 ha) These farm size groups correspond to those used at the time in the official statistics. In Sweden, practically every farmer is a member of one or more agricultural cooperatives. Therefore, members were asked to summarize their opinions when filling in the survey. A final response rate of 40% was achieved after one reminder and thus a dataset of 900 farmers was assembled. Due to incomplete questionnaires not all responses could be used in the analyses. Table 1 presents some descriptive statistics on the dataset. The majority of respondents were male and the average farmer age was 54.3 years. Those farmers who had been directors had occupied that position for on average 5.8 years. Table 1. Basic characteristics of the dataset.1 Item

Average value

Standard deviation

proportion of males (%) average farmer age average number of years as director

71 54.3 5.8

– 15.01 9.40

Two hectares is the lower limit for a holding to be classified as a farm in Swedish agricultural statistics.

International Food and Agribusiness Management Review

Hakelius and Hansson

Volume 19, Issue 4, 2016

Measuring attitudes to agricultural cooperatives As attitudes exist in the mind of people (Kahneman and Sudgen, 2005), in this study and in other research on attitudes (e.g. Hakelius, 1996; Hansson and Lagerkvist, 2012; Pennings and Garcia, 2001; Pennings and Leuthold, 2000) they are considered latent constructs that can be measured using measurement indicators. To this end, a set of statements, most coming from the 1993-survey, presented in Hakelius (1996), to use as measurement indicators was devised for this research, focusing on the trust and commitment dimensions of the attitude construct, following the Theory of Reasoned Action (Ajzen, 1988; Ajzen and Fishbein, 1980). Hence, statements were formulated which aimed at capturing members’ beliefs about agricultural cooperatives in terms of trusting and committing to these, based on general ideas about why there are cooperatives and on the members’ own views through being a cooperative member. These statements are listed in Supplementary Table S1. Examples of statements categorized as trust are 7, 11 and 12, while examples of statements categorized as commitment are 5, 6 and 8. Statements were thus formulated based on theoretical insights into the dimensions that the attitude construct might comprise (trust and commitment) and into how attitudes may be measured as latent constructs. It should be noted, however, that the specific formulation of the statements was developed for data collection in the main project of Hakelius’ Ph.D.-thesis (1996) which this study forms part, and that scale development was taken from the main project. Thus, the measurement scale used in this paper was developed within the research project presented in Hakelius (1996). Analyses based on the project results, including a comparison between farmers’ attitudes to cooperative governance processes in 1993 and 2013, have been reported elsewhere by Hakelius and Hansson (2016). Answers to the statements were collected on a six-point Likert scale ranging from ‘agree completely’ (1) to ‘disagree completely’ (6). Before analyzing the data, the scale was reversed in order to facilitate interpretation of the data. Hence, in the analyses, data were coded to range from (1) ‘disagree completely’ to (6) ‘agree completely’. The main reasons for choosing a six-point Likert scale were that (a) it was considered sufficiently detailed for the purposes of the main project, and (b) an even-number scale was preferred in order to prevent respondents simply selecting the middle answer, while still providing a good number of possible answers to choose between. All intervening response options were anchored, in order to facilitate the respondents’ distinctions between these. Farmers’ attitudes to agricultural cooperative governance systems were assessed by finding the underlying latent structures in the data using exploratory common factor analysis (ECFA). ECFA builds on a reflective measurement model, where the latent construct is assumed to be reflected by its measurement indicators, in line with the theoretical understanding of an attitude construct (Hansson and Lagerkvist, 2012; Pennings and Garcia, 2001). In this study, ECFA was preferred over the confirmatory version, because there is no well-established measurement scale to measure the attitudes of interest. On running the ECFA, measurement indicators with factor loadings below 0.5 were considered non-significant (Hair et al., 2010). These were deleted in a step-wise manner, starting with the measurement indicator possessing the lowest communality. Measurement items loading significantly on two factors were also removed from further analysis. While a theoretical understanding about the attitude construct guided the choice of measurement items to use, the step-wise procedure described above allowed a reduction in the measurement items on a statistical basis, as well as testing which measurement items were significantly associated with the underlying construct. With a reflective measurement model, the underlying construct remains the same even though specific measurement items are removed (e.g. Jarvis et al., 2003). The choice of number of factors retained in the final factor solution was based on combined suggestions from Eigenvalues, scree plots, and the desire to obtain a solution that could be meaningfully interpreted. Oblique (oblimin) rotation, which allows factors to correlate and is thus considered to generate results that are theoretically more valid, was applied. The final factor solution was evaluated with respect to reliability by taking item-to-item and item-to-total correlations into consideration, as well as Cronbach’s alpha. In line with recommendations by Hair et al. (2010), cut-off values of 0.3, 0.5 and 0.6 were used.

International Food and Agribusiness Management Review

Hakelius and Hansson

Volume 19, Issue 4, 2016

Measuring agency problems Indicators of the agency problems examined were collected by asking farmers about their perceptions of possible increases in these agency problems during the preceding decade, i.e. between 2003 and 2013, and by asking them about the need for changes in governance system due to these changes in agency problems. In this way, five separate indicators were developed. These were all measured by asking farmers to indicate the extent to which they agree, on a 1-6 Likert scale, with statements relating to these problems. In particular, the following indicators and statements were used: ■■ Perception of increase in decision problems – measured by the statement ‘It has become more difficult for directors to act in the interest of the members, compared with 2003’. ■■ Perception of increase in follow-up problems – measured by the statement ‘It has become more difficult to follow up on what directors are doing, compared with 2003’. ■■ Perceptions among members concerning a need for a change in the governance system of the cooperative – measured by the statement ‘The system with elected directors involves too many problems and should therefore be changed to a new way of running agricultural cooperatives’. ■■ Perceptions among members concerning a need for more external directors on the board – measured by the statement ‘Members would benefit from external directors being engaged on the board to a larger extent than is the case today’. ■■ Perceptions among members concerning increased distance between the farmer’s own needs and the decisions made – measured by the statement ‘The decisions made by today’s agricultural cooperatives are further way from my ideals than was the situation in 2003’. Of these, the first two indicators directly represent the agency problems that were the focus of this study. The other three indicators represent aspects suggested in the literature as possible remedies for the follow-up and decision problems in cooperatives (Bøhren and Strøm, 2006; Fich and Shivdasani, 2006; Hermalin and Weisbach, 2003; Reynolds, 2003). The latter indicators can be thought of as indirect indicators of agency problems. The reason for their inclusion was that if farmers perceived that agency problems exist, they would also be likely to agree with these remedies to agency problems. The indicators of the agency problems were related to the measures of the attitudes obtained in the ECFA, with the seemingly unrelated regression model, in order to evaluate how perceptions of agency problems form the attitude construct to the agricultural cooperative. The data was analyzed using Stata 12 (StataCorp LP, College Station, TX, USA).

4. Results Factor analysis of attitude construct Descriptive statistics on measurement items used to measure farmers’ attitudes to cooperative governance structures can be found in Supplementary Table S1 to this paper. The results obtained when ECFA was applied to the measurement items listed in Supplementary Table S1 are presented in Table 2. This factor solution corresponds to that of farmers’ attitudes to cooperative governance processes in 2013 presented by Hakelius and Hansson (2016). Based on the same data as used in this study, those authors compared the development of farmers’ attitudes in 1993 to those in 2013. Solutions with two factors (based on suggestions by the Eigenvalues) and three factors (based on suggestions by the scree plot) were evaluated. The solution with two factors was considered easier to interpret, because the factors were more clearly distinguished. Measurement items with insignificant factor loadings were deleted in a step-wise process (factors 1 to 4) where that with the lowest communality was deleted first, until only measurement items with significant loadings remained. The first factor comprised significant measurement items exclusively of commitment type and is hence labelled ‘Commitment’. These measurement items related to the basic ideas behind cooperatives, e.g. that International Food and Agribusiness Management Review

Hakelius and Hansson

Volume 19, Issue 4, 2016

Table 2. Factor solution of the attitude construct (n=710).1 Measurement item C = commitment, T =trust

Factor 1 Factor 2 commitment trust

1. The idea behind cooperatives is a good one. (C) 2. I think that the agricultural cooperative movement is loyal to the cooperative ideals. (C) 3. If members take part in their cooperative member democracy, then they can influence the management of the enterprise. (C) 4. If I take part in the member democracy of my cooperative, then I can influence it in such a way that my own private economic situation improves. (C) 5. Those who are members of an agricultural cooperative ought to, as far as possible, participate in the democratic process. (C) 6. I think it would be interesting to become/that it is interesting to be an elected representative. (C) 7. Today, the board and the chief executive officer usually govern the cooperative in their own way, without caring about what the members think. (T) 8. If I commit myself to the association activities, the economic situation of all members will improve in the long run. (C) 9. If I take part in the cooperative’s democratic process, I will strengthen the special sense of belonging together within the cooperative. (C) 10. The idea that all members can influence their agricultural cooperative, through the democratic process, is basically good but impossible to carry out in reality. (T) 11. The individual members cannot influence the business decisions. Since it is the chief executive officer and the elected representatives who decide. (T) 12. As an elected representative of a cooperative, you soon lose perspective on the real world and in the end you only think about making the cooperative grow. (T) 13. If I participate in the democratic process in my agricultural cooperative, I may be part of influencing that cooperative. (C) 14. It is important to me that as many as possible participate in the democratic process in my agricultural cooperative. (C) 15. If a large proportion of members participate actively in the cooperative’s democratic process, the cooperative will operate better. (C) 16. If you are a member, you should participate both in the business decisions and in the democratic process. (C) Cronbach’s alpha Item-to-item correlation (range) Item-to-total correlation (range)

0.611

0.072

0.023

0.676

0.612

-0.075

0.705

-0.090

0.036

0.619

-0.024

0.830

-0.003

0.707

0.559

-0.316

0.767

0.064

0.729

-0.010

0.688

0.147

0.838 0.292-0.646 0.671-0.789

0.800 0.387-0.628 0.742-0.877

1 Due to missing values, only 710 out of the 900 questionnaires returned could be used in this part of the analysis; measurements in

italics were removed from the analysis due to insignificant factor loadings; the factor solution reported here has also been presented in Hakelius and Hansson (2016).

they are based on good ideals and that members should be committed to the cooperative. The second factor comprised significant measurement items exclusively of trust type and is hence labelled ‘Trust’. These measurement items related to farmers’ experiences of governance structures of cooperatives. In terms of reliability, both of these two factors had Cronbach’s alpha values well above the accepted cut-off value of 0.6 in exploratory settings (Hair et al., 2010), item-to-item correlations above, or very close to, the cut-off value of 0.3 (ibid.), and item-to-total correlations well above the cut-off value of 0.5 (ibid.). Based on this, the two factors were considered reliable measures of the attitude construct.

International Food and Agribusiness Management Review

Hakelius and Hansson

Volume 19, Issue 4, 2016

Impact of perceived agency problems on farmers’ attitudes to agricultural cooperatives Descriptive statistics on the indicators used to measure farmers’ perceptions of agency problems in Swedish agricultural cooperatives are presented in Table 3. Average values of the indicators relating to the perceived changes between 2003 and 2013 were >4, indicating that on average farmers tended to believe that agency problems increased between 2003 and 2013, at least moderately. The indicators relating to a need for a change in governance structure had average values of 3.53 and 3.79, suggesting that the farmers were on average indifferent with respect to these questions. The indicators of increased agency problems were related to the measures of the two dimensions of farmers’ attitude to agricultural cooperatives together with information about farmers’ age and the size of the farms measured in the number of hectares as control variables. Through this, we were able to assess the impact of increased agency problems on these attitudes in order to evaluate how perceptions of agency problems form attitudes. The results (Table 4) show that the attitude dimension ‘Trust’ in particular instrumental is for the indicators of agency problems. This means that formulation of this part of the attitude construct is dependent on the increase in, and existence of agency problems experienced by the farmers. It should be noted that a higher score on the attitude dimension ‘Trust’ means that the farmer had a more negative attitude dimension, given the way in which the measurement indicators were posed. The model relating to ‘Trust’ was able to explain 41.2% of the variation in this attitude dimension and all indicators of agency problems had statistically significant impacts on the attitude dimension measure. The model itself was also highly statistically significant (P<0.000). The impacts of all indicators of agency problems were positive except for externals, i.e. the idea that adding non-member directors to the board can improve decision making. This suggests that farmers scoring high on this attitude dimension believe that it has become more difficult to control and monitor what directors are doing. They believe that decisions are made further from what they would like compared with 2003, and that the current governance of agricultural cooperatives should be changed to a new system. Interestingly, there was a significantly negative impact of the indicator relating to externals, suggesting that even if farmers score high on the attitude dimension ‘Trust’ they do not believe that external board members would be a solution to the problems they obviously perceive. The model was able to explain the attitude dimension ‘Commitment’ to a lesser extent; only 12.1% of the total variation in this attitude was attributable to the indicators of agency problems, although the model itself was highly statistically significant (P<0.000). Only one of the indicators, governance, had a statistically Table 3. Descriptive statistics on indicators of increased agency problems in agricultural cooperatives in 2003 compared with 2013.1 Statement

Average St. dev. score

Median 1st - 3rd quantile

It has become more difficult for directors to act in the interest of the members, compared with 2003: decision. It has become more difficult to follow up on what directors are doing, compared with 2003: follow-up. The system with elected directors involves too many problems and should therefore be changed to a new way of running the agricultural cooperatives: governance. Members would benefit from external directors being engaged in the board to a larger extent than is the case today: externals. The decisions made by today’s agricultural cooperatives are further way from my ideals than was the situation in 2003: distance.

4.19

1.20

3-5

4.11

1.19

3-5

3.53

1.39

3-5

3.79

1.35

3-5

4.04

1.19

3-5

1 Indicators were measured on 1-6 Likert scale, where 1 indicates disagree totally and 6 indicates agree totally. Minimum value of all indicators was 1 and maximum value was 6.

International Food and Agribusiness Management Review

Hakelius and Hansson

Volume 19, Issue 4, 2016

Table 4. Regression results for impact of indicators of agency problems on attitudes to cooperatives. Regression results were derived from the seemingly unrelated regression technique.1

intercept decision follow-up governance externals distance age of farmer size of farm (ha) fit statistics

Dependent variable ‘Commitment’

Dependent variable ‘Trust’

estimated coefficient

P-value

4.069 0.000 0.063 0.214 -0.015 0.761 -0.243 0.000 0.037 0.225 0.043 0.295 0.006 0.029 0.001 0.376 χ2=82.96 (P-value 0.000) R2=0.1210

P-value

1.824 0.000 0.165 0.021 0.138 0.036 0.267 0.000 -0.070 0.013 0.140 0.000 -0.003 0.325 -0.000 0.505 χ2=423.44 (P-value 0.000) R2=0.412

Statistical inference is based on bootstrapped standard errors because the dependent variables were not normally distributed, n repetitions = 1000.

significant impact on the attitude construct. This impact was negative, implying that farmers scoring high on this attitude dimension disagree with the suggestion that agricultural cooperatives should be run in another way, which is plausible. As for the control variables, the age of the farmer was positively significantly related to the attitude dimension ‘Commitment’, indicating that older farmers are more committed to the cooperatives. However, the variable accounting for farm size was not significantly related to this attitude dimension, and none of the control variables was significantly related to the attitude dimension ‘Trust’. Taken together, findings indicate that there is a distrust among farmers concerning the possibility of implementing the traditional cooperative idea and that farmers want a different governance system than the one member-one vote system used today. Simultaneously, farmers recognize that it is more difficult for directors to ‘read’ members’ requirements, indicating that the decision problem is also present and has grown during the period. The dimensions of the attitude construct, especially that related to ‘Trust’, were significantly influenced by the beliefs about agency problems. Thus, there is evidence that attitudes to agricultural cooperatives are formed by farmers’ perceptions of agency problems. If attitudes are a determinant of behavior (and thus, in businesses, of decision making), this means that the existence of agency problems influences farmers’ decision making and, in continuation, their strategic behavior with respect to the cooperative.

5. Discussion and conclusions This study integrated insights concerning agency problems with attitude research with the aim of evaluating whether and how farmers’ attitudes to the governance system of agricultural cooperatives are shaped by their perceptions of agency problems. The agency problems considered here were the decision and followup problems, which basically stem from the same features: agricultural cooperatives have become large and more complex in terms of organizational structure and logic, and hence the member group is large and heterogeneous, with fewer directors at a greater distance from the members. Because attitudes are one type of determinant of human behavior (e.g. Ajzen, 1991, 2002; Conner and Abraham, 2001; Fazio and Olson, 2003; Feist, 2012; Kaiser, 2006; Kaiser and Sheuthle, 2003; Siegel Levine and Straube, 2012), and thus of decision making, identifying the antecedents of these attitude dimensions is important for understanding how agricultural cooperatives can be managed. The premise in this study was that farmers’ perceptions of agency problems are major determinants of farmers’ beliefs about agricultural cooperatives, and that they therefore also shape their attitude to cooperatives. Previous research has reported pronounced agency problems in cooperatives (e.g. Fulton and Larson, 2009; Richards et al., 1998), but to the best of our knowledge these have not previously been integrated with attitude research in order to determine whether and how attitudes to International Food and Agribusiness Management Review

Hakelius and Hansson

Volume 19, Issue 4, 2016

agricultural cooperatives are formed by perceptions of agency problems. By integrating insights from agency problems with attitude research, this study contributes novel information about how positive or negative attitudes to the cooperative governance system in agricultural cooperatives are shaped by farmers’ perceived agency problems. Such insights can be important for directors of agricultural cooperatives in developing governance systems and information channels in cooperatives. This can lead to decreased agency problems and increased loyalty, through increased trust and commitment among the members. This study is based on a large empirical dataset collected through a postal questionnaire sent to a sample of Swedish farmers in 2013, asking about farmers’ perceptions of agency problems in agricultural cooperatives. We found that members’ attitudes to agricultural cooperatives can be regarded as consisting of a ‘Commitment’ dimension and a ‘Trust’ dimension. We also found indications that members perceive agency problems to have increased between 2003 and 2013. For instance, members seemed to agree with statements that it has become more difficult for directors to act in the interest of the members compared with a decade ago; that over the last decade it has become more difficult to monitor what directors are doing, and that decisions now are further away from members’ ideas than previously. We also found that members’ attitudes to the governance system of agricultural cooperatives were statistically significantly formed by farmers’ perceptions of these agency problems. This was especially true for the attitude dimension ‘Trust’, i.e. attitudes based on the farmer’s experiences of the governance system of cooperatives, but to some extent also for the attitude dimension ‘Commitment’. Given that higher scores on the ‘Trust’ dimension of the attitude construct should be interpreted as the farmer having more a negative attitude dimension, the findings suggest that increased agency problems lead in particular to a decrease in trust. Furthermore, as the ‘Commitment’ dimension appears to be less affected by the perceived agency problems, the findings also suggest that farmers continue being committed to agricultural cooperative governance systems even when agency problems are perceived to increase. Taken together, these findings indicate that even in situations where the agency problems are perceived to increase, farmers continue to be committed to the governance system of their agricultural producer cooperatives, but their trust in the way these systems are working seems to decrease. The results thus indicate that farmers’ attitude construct is formed by their perceptions of agency problems. As attitudes are one type of determinant of behavior (e.g. Ajzen 1991, 2002), we therefore suggest that members’ behavior and decision making with respect to agricultural producer cooperatives are driven by their perceptions of agency problems, via the attitude construct. However, the link between perceptions of agency problems and attitudes and behavior with respect to agricultural cooperatives would need to be confirmed in future research. The findings presented here have policy implications for agricultural cooperatives. In particular, it appears to be important for these cooperatives to work to resolve the dissatisfaction indicated in the attitude dimension concerning the democratic process of agricultural cooperatives. Put differently, it is vital to work to reduce agency problems in order to influence members’ attitudes to agricultural cooperatives and thus their behavior to these. Following the idea of Barraud-Didier et al. (2012), decreasing loyalty may be turned around by focusing on improving members’ trust in the cooperative, possibly by using the tools of changing the governance structure and shortening the distance between members and directors. An interesting question for future research is how to exactly do this. Furthermore, working on changing the organizational structure and ownership structure of agricultural cooperative may prove to be beneficial for improving the commitment of both members and directors to engaging in the governance of agricultural cooperatives.

Supplementary material Supplementary material can be found online at https://doi.org/10.22434/IFAMR2015.0219. Table S1. Descriptive statistics of measurement items. International Food and Agribusiness Management Review

Hakelius and Hansson

Volume 19, Issue 4, 2016

References Ahearn, M.C., J. Yee and P. Korb. 2005. Effects of differing farm policies on farm structure and dynamics. American Journal of Agricultural Economics 87: 1182-1189. Ajzen I. 1988. Attitudes, personality and behavior. Milton Keynes: Open University Press, London, UK. Ajzen I. 1991. The theory of planned behavior. Organizational Behavior and Human Decision Processes 50: 179-211. Ajzen, I. 2002. Perceived behavioral control, self-efficacy, locus of control, and the theory of planned behavior. Journal of Applied Social Psychology 32: 665-683. Ajzen, I. and M. Fishbein. 1980. Understanding attitudes and predicting social behavior. englewood cliffs. Prentice-Hall, Upper Saddle River, NJ, USA. Balmann, A., K. Dautzenberg, K. Happe and K. Kellermann. 2006. On the dynamics of structural change in agriculture: internal frictions, policy threats and vertical integration. Outlook on Agriculture 35: 115-121. Barbaud-Didier, V., M.C. Henninger and A. El Akremi. 2012. The relationship between members’ trust and participation in the governance of cooperatives: the role of organizational commitment. International Food and Agribusiness Management Review 15: 1-24. Bhuyan, S. 2007. The ‘people’ factor in cooperatives: an analysis of members’ attitudes and behavior. Canadian Journal of Agricultural Economics 55: 275-298. Bijman, J., M. Hanish and G. van der Sangen. 2014. Shifting control? The changes of internal governance in agricultural cooperatives in the EU. Annals of Public and Cooperative Economics 85: 641-661. Bijman, J., G. Hendrikse and G. van der Sangen. 2013. Accommodating two words in one organization: changing board models in agricultural cooperatives. Managerial and Decision Economics 34: 204-217. Borgen, S.O. 2001. Identification as a trust-generating mechanism in cooperatives. Annals of Public and Cooperative Economics 72: 209-228. Borgen, S.O. 2004. Rethinking incentive problems in cooperative organizations. Journal of Socio-Economics 33: 383-393. Bøhren, Ø. and Ø. Strøm. 2006. Aligned, informed, and decisive: characteristics of value-creating boards. BI-Norwegian School of Management working paper. Cechin, A., J. Bijman, S. Pascucci and O. Omta. 2013. Decomposing the member relationship in agricultural cooperatives: implications for commitment. Agribusiness 29: 39-61. Chaddad, F. and C. Iliopoulos. 2013. Control rights, governance, and the costs of ownership in agricultural cooperatives. Agribusiness 29: 3-22. Conner, M. and C. Abraham. 2001. Conscientiousness and the theory of planned behavior: toward a more complete model of the antecedents of intentions and behavior. Personality and Social Psychology Bulletin 27: 1547-1561. Cook, M.L. 1994. The role of management behavior in agricultural cooperatives. Journal of Agricultural Cooperation 9: 42-58. Cook, M.L. 1995. The future of US agricultural cooperatives: a neo-institutional approach. Journal of Agricultural Economics 77: 1153-1159. Cook, M.L. and M. Burress. 2013. The impact of CEO tenure on cooperative governance. Managerial and Decision Economics 34: 218-229. Cook, M.L. and C. Iliopoulos. 2000. Ill-defined property rights in collective action: the case of US agricultural cooperatives. In: Institutions, contracts and organizations: perspectives from new institutional economics, edited by C. Ménard. Edward Elgar, Cheltenham, UK, pp. 335-348. Cornforth, C. 2004. The Governance of cooperatives and mutual associations: a paradox perspective. Annals of Public and Cooperative Economics 75: 11-32. Eisenhardt, K.M. 1989. Agency theory: an assessment and review. Academy of Management Review 14: 57-74. Fama, E. 1980. Agency problems and the theory of the firm. Journal of Political Economy 88: 288-307. Fama, E.F. and M.C. Jensen. 1983. Separation of ownership and control. Journal of Law and Economics 26: 301-325.

International Food and Agribusiness Management Review

Hakelius and Hansson

Volume 19, Issue 4, 2016

Fazio, R.H. and M.A. Olson. 2003. Attitudes: foundations, functions, and consequences. In The Sage Handbook of Social Psychology, edited by M.A. Hogg and J. Cooper. Sage, London, UK, pp. 140-160. Feist, G.J. 2012. Predicting interest in and attitudes toward science from personality and need for cognition. Personality and Individual Differences 52: 771-775. Feng, L. and G.W.J. Hendrikse. 2012. Chain interdependencies, measurement problems and efficient governance structure: cooperatives versus publicly listed firms. European Review of Agricultural Economics 39: 241-255. Fich, E.M. and A. Shivdasani. 2006. Are busy boards effective monitors? The Journal of Finance LXI: 689-724. Fulton, M. 1999. Cooperatives and member commitment. the role of cooperative entrepreneurship in the modern environment. Pellervo, Helsinki, Finland, pp. 418-437. Fulton, J.R. and W.L. Adamowicz. 1993. factors that influence the commitment of members to their cooperative organization. Journal of Agricultural Cooperation 8: 39-53. Fulton, M. and K. Giannakas. 2001. Organizational commitment in mixed oligopoly: agricultural cooperatives and investor-owned firms. American Journal of Agricultural Economics 5: 1258-1265. Fulton, M. and K. Giannakas. 2007. Agency and leadership in cooperatives. In: Vertical Markets and Cooperative Hierarchies, edited by K. Karantininis and J. Nilsson. Springer, Dordrecht, the Netherlands, pp. 93-113. Fulton, M. and K. Larson. 2009. The restructuring of the Saskatchewan wheat pool: overconfidence and agency. Journal of Cooperatives 23: 1-19. Gray, T.W. and C.A. Kraenzle. 1998. Member participation in agricultural cooperatives: a regression and scale analysis. USDA: RBS Research Report 165. Available at: http://tinyurl.com/hswnfdr. Hair, J.F., W.C. Black, B.J. Babin and R.E. Anderson. 2010. Multivariate data analysis – a global perspective 7th ed, Pearson. New Jersey, NJ, USA: Hakelius, K. 1996. Cooperative values. farmers’ cooperatives in the minds of the farmers. Ph.D. diss. #23. The Swedish University of Agricultural Sciences. Hakelius, K. and H. Hansson. 2016. Measuring changes in farmers’ attitudes to agricultural cooperatives: evidence from Swedish agriculture 1993-2013. Agribusiness 32: 531-546. Hansen, M.H., J.L. Morrow Jr, and J.C. Batista. 2002. The impact of trust on cooperative membership retention, performance, and satisfaction: an exploratory study. International Food and Agribusiness Management Review 5: 41-59. Hansmann, H. 1996. The Ownership of enterprise. The Belknap Press of Harvard University Press, Cambridge, MA, USA. Hansson, H. and C.J. Lagerkvist. 2012. Measuring farmers’ preferences for risk: a domain-specific risk preference scale. Journal of Risk Research 15: 737-753. Happe, K, A. Balmann, K. Kellermann and C. Sahrbacher. 2008. Does structure matter? The impact of switching the agricultural policy regime on farm structures. Journal of Economic Behavior and Organization 67: 431-444. Hendrikse, G. 2007. Two vignettes regarding boards in cooperatives versus corporations – irrelevance and incentives. In: Vertical Markets and Cooperative Hierarchies, edited by K. Karantininis and J. Nilsson. Springer, Dordrecht, the Netherlands, pp. 137-150. Hermalin, B.E. and M.S. Weisbach. 2003. Boards of directors as an endogenously determined institution: a survey of the economic literature. Economic Policy Review 9: 7-26. James, H.S. 2001. The trust paradox: a survey of economic inquiries into the nature of trust and trustworthiness. Journal of Economic Behavior and Organization 47: 291-307. James, H.S. and M.E. Sykuta. 2005. Property right and organizational characteristics of producer-owned firms and organizational trust. Annals of Public and Cooperative Economics 74: 545-580. James, H.S. and M.E. Sykuta. 2006. Farmer trust in producer- and investor-owned firms: evidence from Missouri corn and soybean producers. Agribusiness 22: 135-153. Jarvis C.B,. S.B. Mackenzie and P.M. Podsakoff. 2003. A critical review of construct indicators and measurement model misspecification in marketing and consumer research. Journal of Consumer Research 30: 199-218. International Food and Agribusiness Management Review

Hakelius and Hansson

Volume 19, Issue 4, 2016

Kahneman, D. and R. Sugden. 2005. Experienced utility as a standard of policy evaluation. Environmental and Resource Economics 32: 161-181. Kaiser, F.G. 2006. A moral extension of the theory of planned behaviour: norms and anticipated feelings of regret in conservationism. Personality and Individual Differences 41: 71-81. Kaiser, F.G. and H. Sheuthle. 2003. Two challenges to moral extension of the theory of planned behaviour: moral norms and just world beliefs in conservationism. Personality and Individual Differences 35: 1033-1048. Laffont, J.J. and D. Martimort. 2002. The theory of incentives: the principal-agent model. Princeton University Press, Princeton, NJ, USA. Linag, Q. and G. Hendrikse. 2013. Cooperative CEO identity and efficient governance: Member or outside CEO? Agribusiness 29: 23-38. Nilsson, J. 2001. Organisational principles for co-operative firms. Scandinavian Journal of Management 17: 329-356. Nilsson, J., A. Kihlén and L. Norell. 2009. Are traditional cooperatives an endangered species? About shrinking satisfaction, involvement and trust. International Food and Agribusiness Management Review 12: 101-122. Nilsson, J and G.T. Svendsen. 2011. Free riding or trust? Why members (do not) monitor their co-operatives. Journal of Rural Cooperation 39: 131-150. Nilsson, J., G.L.H. Svendsen and G.T. Svendsen. 2012. Are large and complex agricultural cooperatives losing their social capital? Agribusiness 28: 187-204. Novkovic, S. 2008. Defining the co-operative difference. Journal of Socio-Economics 37: 2168-2177. Österberg, P. and J. Nilsson. 2009. Members’ perception of their participation in the governance of cooperatives: the key to trust and commitment in agricultural cooperatives. Agribusiness 25: 181-197. Pennings, J.M.E. and P. Garcia. 2001. Measuring producers’ risk preferences: a global risk attitude construct. American Journal of Agricultural Economics 83: 993-1009. Pennings, J.M.E. and R. Leuthold. 2000. The role of farmers’ behavioral attitudes and heterogeneity in futures contracts usage. American Journal of Agricultural Economics 82: 908-919. Pietola, K., M. Vare and A.O. Lansink. 2003. Timing and type of exit from farming: farmers’ early retirement programmes in Finland. European Review of Agricultural Economics 30: 99-116. Reynolds, B. 2003. Recruiting and selecting cooperative directors – a survey summary. USDA, Rural business, Cooperative Services. Richards, T.J., K.K. Klein and A. Walburger. 1998. Principal-agent relationships in agricultural cooperatives: an empirical analysis from rural Alberta. Journal of Cooperatives 13: 21-33. Siegel Levine D. and M.J. Strube. 2012. Environmental attitudes knowledge, intentions and behaviors among college students. Journal of Social Psychology 152: 308-326. Weber, E.U., A.R. Blais and N.E. Betz. 2002. A domain-specific risk-attitude scale: measuring risk perceptions and risk behaviors. Journal of Behavioral Decision Making 15: 263-290.

International Food and Agribusiness Management Review

OPEN ACCESS International Food and Agribusiness Management Review Volume 19 Issue 4, 2016; DOI: 10.22434/IFAMR2015.0110 Received: 14 July 2015 / Accepted: 22 August 2016

Do coffee cooperatives benefit farmers? An exploration of heterogeneous impact of coffee cooperative membership in Southwest Ethiopia RESEARCH ARTICLE Zekarias Shumeta a

and Marijke D’Haeseb

Assistant professor, Department of Agricultural Economics and Extension, Jimma University, P.O. Box 307, Jimma, Ethiopia b

Professor, Department of Agricultural Economics, Ghent University, Coupure Links 653, 9000 Gent, Belgium

Abstract Smallholder farmers’ participation in agricultural cooperatives is often promoted as a promising strategy for overcoming market imperfections and to increase farmers’ productivity and income. In recognition of this potential, in recent years, Ethiopia has shown renewed interest in promoting cooperatives. However, there is lack of empirical evidence of the impact that cooperatives have on farmers’ performance in Ethiopia. Using a matching technique, we evaluate the impact of coffee cooperatives on the performance of their member households in terms of income and coffee production. We use data from coffee farmers in south-west Ethiopia. The overall results suggest that members of cooperatives are not faring much better than non-members. The treatment effects we measured were not statistically significant from zero. Yet, the aggregate figures mask differences between cooperatives and amongst individual cooperative members. Average treatment effects on members differ between cooperatives, in general older members, those who have benefitted from more education and those with larger coffee plantations seem to benefit more from membership. Our analysis sheds light on the heterogeneity in the impact that membership of a cooperative can have: this differs by cooperative and by members within cooperatives, a finding that has important policy implications. Keywords: coffee, cooperatives, Ethiopia, performance, propensity score matching, heterogeneity JEL code: Q1, N5 Corresponding author: zekishum@yahoo.com

Shumeta and D’Haese

Volume 19, Issue 4, 2016

1. Introduction It is widely recognized that by participating more in markets smallholder farmers can increase their productivity levels and incomes, thereby improving their food security and experiencing less poverty. A thriving agricultural sector contributes to overall economic growth (World Bank, 2007). At the same time, changing economic, environmental and socio-political conditions around the world pose serious challenges to agricultural production and particularly to small-scale production. Today, as in the past, African smallholder producers face problems in accessing rewarding markets. Market liberalization, mainly in markets for traditional export products – such as coffee, and globalization have brought new opportunities and challenges to farmers (World Bank, 2007) who have had to start to deal with the quasi-monopsonistic powers of intermediaries with whom they have to make deals on spot markets or negotiate contracts with (Markelova and Mwangi, 2010). If unprotected or insufficiently supported, smallholder farmers are often disadvantaged and lack the bargaining power to secure fair trade conditions (Mujawamariya, 2013). Horizontal coordination among farms, in the form of cooperatives or producer groups is often promoted as a way to overcome market imperfections and constraints (World Bank, 2007; Verhofstadt and Maertens, 2015). These collective organizations can facilitate and leverage market linkages for small scale producers, help them to exploit the economies of scale that farmers are unable to achieve individually, improve their bargaining power and provide access to inputs, transport and market information that can enable farmers to engage with and benefit from existing value chains (Verhofstadt and Maertens, 2015; World Bank, 2007). Over recent decades, donors and governments have been supportive of producer cooperatives (Berdegué, 2001; Bernard et al., 2008; Collion and Rondot, 2001; World Bank, 2003, 2007) even though studies reveal that, in Ethiopia and elsewhere, they achieve different levels of success. Our study shows that cooperative membership brings mixed results and that this is partly due to differences between cooperatives (some cooperatives perform well while others fail to create impact for their members) and partly because some members benefit from membership more than others). Fisher and Qaim (2012), Ito et al. (2012) and Vandeplas et al. (2013) have all shown the positive and significant impact of cooperative membership on farm income and profits. By contrast, Bernard et al. (2008) and Francisconi and Heerink (2011) show that cooperatives have a limited influence on their members’ commercialization behavior. Barham and Chitemi (2009) examined the extent to which certain characteristics and asset endowments of smallholder farmer groups facilitate collective actions that can improve group marketing performance. Their findings suggest that more mature groups with stronger internal institutions, functioning group activities and a good base of natural capital are more likely to improve their members’ market situation. Markelova and Mwangi (2010) indicated the need to consider different types of markets and products, the characteristics of user groups, institutional arrangements, and external environment to determine the effectiveness and sustainability of collective marketing for smallholders. Cazuffi and Moradi (2012) found a net positive effect of group size on cooperative performance probably resulting from economies of scale. However, the average impact of cooperative membership on members’ performance seems to hide considerable heterogeneity between members. The World Bank (2007) expressed concerns about this trend, which was confirmed by Bernard et al. (2008) who found a positive and significant impact of cooperative membership on the degree of commercialization for large farms, yet a sometimes negative impact for some very small farms. Similarly, Verhofstadt and Maertens (2015) show that cooperative membership is more beneficial for larger farms and members in remote areas in Rwanda. Fisher and Qaim (2012) show that the effects of membership, in terms of commercialization, technology adoption and farm income, of banana cooperatives in Kenya are more noticeable for the smallest farms. Ito et al. (2012) conclude that the income effect of cooperative membership for watermelon farmers in China is twice as large for small farms than for larger farms. Abebaw and Haile (2013) assess the impact of cooperative membership on the likelihood of fertilizer adoption among farmers in Ethiopia and find that there is a significant positive effect for less educated farmers and an inverse U-shaped effect of distance to the market. Yet, none of these studies examined coffee cooperatives, despite the relevance of coffee for income generation and the mainly smallholder nature of its production. International Food and Agribusiness Management Review

Shumeta and D’Haese

Volume 19, Issue 4, 2016

Studying variance in treatment effects across cooperatives and members is useful in that it can help policy makers and researchers anticipate problems that could endanger the sustainability of cooperatives, and can also play an important role in improving program targeting. If only the top-performing farmers join a cooperative, the net benefit of membership could decrease if economies of scale are not increased significantly. The difference the cooperative makes for these top-performers would be small compared to when they would operate individually and the transaction costs involved in cooperating could be larger than benefits in economies of scale. On the other hand, attracting top-performers to join a cooperative which also has less successful or less qualified farmers as members, could increase the net impact, especially for those members who faced difficulties before joining the cooperative (Djebbari and Smith, 2008; Verhofstadt and Maertens, 2015; Xie et al., 2012). In short, supporting cooperatives may contribute to uplifting some members out of poverty but the average effect could be larger for less-well performing farmers. In this paper, we explore both the overall and the heterogeneous impact of membership of coffee cooperatives in Ethiopia. Our general hypothesis is that coffee farmers in south-west Ethiopia benefit from cooperative membership in terms of increasing income via improving the supply, the price they receive and the margins obtained (H1). We assume that this impact will differ among the different socioeconomic groups of member households (H2), and across cooperatives (H3). We collected data amongst coffee farmers in south-west Ethiopia and used propensity score matching to estimate the average treatment effect of cooperative membership. We analyzed how the estimated treatment effect differs with various household and farm characteristics. We find that in general cooperative membership does not have a significant impact on the selected performance indicators, including income, the volume of supply, the price and net margins. Nonetheless, significant differences on the estimated average treatment effects were observed between cooperatives and individual members. Older members, those who have better education and those who own large farms were found to benefit more. The result also indicated that different coffee cooperatives have different impacts on performance. Hence, although the average effect of cooperative membership is not significantly different from zero, this does not mean that cooperatives are ineffective: some are and some aren’t and members of some cooperatives fair better than non-members, while some members fair better than others.

2. Background and data collection Despite the tempestuous history of cooperatives under Ethiopia’s socialist regime (1974-1991), the present government of the Federal Democratic Republic of Ethiopia (FDRE) has expressed renewed interest in collective action to improve smallholders’ market involvement (Abate et al., 2014; FDRE, 1994, 1998). This renewed interest in cooperatives was also inscribed in the Sustainable development and poverty reduction program (FDRE, 2002 cited in Abate et al., 2014) as well as the Plan for accelerated and sustained development to end poverty (FDRE, 2005 cited in Abate et al., 2014), in which cooperatives are central to the country’s rural development strategy. Agricultural cooperatives are also recognized as privileged institutions by the recently established Agricultural transformation agency (Abate et al., 2014). As a result, cooperatives are widespread throughout the country and a substantial number of public improvement programs and private initiatives are directed through them in an attempt to reduce the exorbitant transaction and coordination costs that individual farmers face (Pingali et al., 2005). In this study, we focused on coffee cooperatives from the Jimma and Kaffa zones of south-west Ethiopia (Supplementary Figure S1). Jimma is one of the twelve zones in the Oromia region which has about 2.5 million inhabitants, most (88%) of who are Oromo, and manly speak Oromiffa and Amharic. Muslim is the dominant religion in the region, followed by orthodox Christianity and Protestantism. At an altitude of around 1,700 m.a.s.l., and with average temperatures that range from 8 to 28 °C, coffee (Arabica) thrives in the region and is the most important crop in terms of contributing to peoples’ livelihoods. Cereals, such as maize, and fruits, such as avocado and mango, are also widely produced in the region (personal communication, Jimma Zone Bureau of Agriculture 2012-2013). Kaffa is part of the Southern Region of Ethiopia and it contains a International Food and Agribusiness Management Review

Shumeta and Dâ&#x20AC;&#x2122;Haese

Volume 19, Issue 4, 2016

population of about one million people. The primary language in the area is Kaffa and the major religion is orthodox Christianity (80%). Temperatures are around 18-21Â Â°C and the altitude ranges from 500 m.a.s.l in the south to 3,000 m.a.s.l in the north and central highlands. Due to the favorable altitude and weather condition, much coffee (Arabica) is produced in the zone and provides the main source of income for farmers. Agricultural production, livestock rearing and collection of non-timber forest products are other important sources of livelihood in the zone (Personal communication, Kaffa Zone Bureau of Agriculture 2012-2013). Coffee cooperatives in the region are multi-purpose. Other than marketing coffee, they also process coffee and are involved in the sale of other crops. They also provide inputs (e.g. improved coffee seedlings, improved seed and fertilizer for other crops, etc.) and consumables, such as sugar and oil. Member farmers also receive at least one training session a year and a dividend (between 2 and 4 birr/kg of coffee cold1) from their cooperative. Cooperatives are organized by Kebeles or Peasant Associations, with each cooperative named after its location. Our preliminary study estimated that there are sixty-three (eight certified and fiftyfive uncertified) cooperatives in Jimma supplying coffee to the export market and twenty-seven (eighteen certified and nine uncertified) in Kaffa (personal communication, zonal cooperative agencies 2012-2013). Data collected from a survey of coffee farm households in 2012-2013 have been used for this study. In the first stage, three weredas (districts) from each zone were purposively selected on the basis of coffee production and the concentration of cooperatives. In the second stage, two Kebeles from each wereda were purposively selected using accessibility and existence of a relatively equal proportion of members and non-members as criteria. In our context, inaccessible cooperatives are cooperatives which require three to four hour walk on foot to reach due to the absence of any type of road for vehicle. But in order to avoid biases, we considered cooperatives which have similar performance with those of the inaccessible ones in our sample selection procedure. In the third stage, the households were stratified on the basis of their membership status. A random selection of 132 members and 124 non-member coffee producing households (the control group) were made across from twelve cooperative Kebeles with the help of experts from the Bureau of Agriculture and development agents. Despite the relevance of stratifying the sample on gender basis for a better understanding on determinants of cooperative membership, it was impossible to consider it due to the limited number of female headed households in the area. Most of the cooperative women are wives in the male member headed households. Respondents were interviewed by twelve trained enumerators using a structured questionnaire with different sections on household characteristics, farm characteristics, the volumes of coffee produced and supplied, prices received and costs incurred and cooperative membership. The household data were supplemented with information obtained from key informant interviews, focus group discussions with selected farmers and surveys among the twelve cooperatives (both certified and uncertified) that the sample of farmers belong to.

3. Analytical framework A particular challenge in assessing the effect of cooperative membership on performance is the need for a counterfactual; a control group of farmers who are not members of a cooperative (Heckman et al., 1997). We used the well-known propensity score matching (PSM) technique (Becker and Ichino, 2002; Dehejia and Wahba, 2002; Heckman et al., 1997; Rosenbaum and Rubin, 1985) to test our general hypothesis that coffee cooperatives have a substantial impact in improving the income of their members through increasing supply, price received and margins obtained (H1). In this technique the farmers in both treated and control samples are matched based on their observable characteristics. The impact is measured by the difference in performance between pairs of treated and control farmers. This allows us to partially control for non-random selection of cooperative members (Caliendo and Kopeinig, 2008; Imbens, 2004). Members are matched with non-members in order to search for differences in performance or the average treatment effect on the treated (ATT) in terms of supply volume, income levels, price received and margins obtained. 1 10

birr = 0.45 USD, calculated on the basis of the exchange rate on October 11, 2016.

International Food and Agribusiness Management Review

Shumeta and D’Haese

Volume 19, Issue 4, 2016

It should be noted that the PSM mimics the effects of a counterfactual and attempts to control for any bias caused by non-random selection. Yet it does not take into account any possible spillover effects of cooperative membership. In addition, members may side-sell some or all of their coffee outside the cooperatives (cf. Mujawamariya et al., 2014) and this also cannot be captured by PSM. As a first step the probability of being member of the cooperative was estimated as a function of observable pre-treatment covariates, using a logit model that included different sets of confounding variables that may explain the non-random distribution of cooperative membership among the population. Next, the predicted values of the logit model generated propensity scores for all treatment and control units. Mathematically, this is written as: PS = Prob (Z=1|X)

(1)

Where the PS is the propensity score obtained through a logit regression of observable covariates on cooperative membership, Z is the probability of sample farmers being members of cooperatives and the variables considered in vector X (age, years of schooling, number of family members in the productive age range, land planted with coffee, off-farm income, risk of price volatility on coffee income, location). These variables were inspired by previous research (see next section). The propensity scores were used to restrict the samples and ensure common support or overlap. The common support assumption requires balancing the covariate distribution between treated and untreated observations, so that treatment observations will have a comparable control observation close-by in the PS distribution. Once sufficient overlap is found, treated and control units with similar propensity scores are matched using the Kernel matching method and the ATT can be calculated. The estimation of ATT is given by: ATT = E[Y (1) – Y (0) |Z=1] = E[Y (1) |Z=1] – E[Y (0) |Z=1]

(2)

Where E[Y (1) |Z=1] is the mean value of the outcome variable in the treatment group, and E[Y (0) |Z=1] is the mean value of the outcome variable in the matched control group. A good matching estimator does not eliminate too many of the original observations from the analysis and should still, at the same time, yield statistically equal covariate means for households in the treatment and control groups (Caliendo and Kopeinig, 2008). In this regard, the use of kernel matching is helpful since it uses more information to construct the counterfactual outcome by using a weighted average of all the individuals in the control group with weights that are inversely proportional to the propensity score distance between the treated and control units, which reduces variance (Caliendo and Kopeinig, 2008). As PSM results are sensitive to matching methods (Caliendo and Kopeinig, 2008; Imbens, 2004), neighborhood matching was estimated as a check for robustness. Bootstrap standard errors were used to test the statistical significance of the estimated ATT in order to account for the variation caused by the matching process. Finally, the balancing of the covariates was checked by testing that the means of each covariate between the treated and control groups did not differ after matching. Next, the ATT was explored against the farms’ characteristics to test whether there is impact difference among the different groups of member households (H2). Inspired by similar work (Abebaw and Haile, 2013; Bernard et al., 2008; Mutuc et al., 2013; Verhofstadt and Maertens, 2015), the estimated ATT of each outcome variable was used as a dependent variable in a linear regression model to investigate how the cooperative effect may vary for different household and farm characteristics. Graphic assessment of the impact of heterogeneity was also made by plotting the ATT over the distribution of the propensity scores, household and farm characteristics are presented in Supplementary Figures S2 to S8. Finally, anova-post

International Food and Agribusiness Management Review

Shumeta and D’Haese

Volume 19, Issue 4, 2016

hoc test was applied to investigate the prevalence of significant variation among cooperatives in impacting their member households (H3). The definition of variables used in the analytical framework A number of variables are thought to influence membership of a coffee cooperative. These include household demographic characteristics, farm characteristics, income and some physical factors, such as distance to coffee collection points and geographic location. The matching of members and non-members was made on the basis of these observable characteristics in order to only point out the treatment effect on the outcome variables. Table 1 defines and quantifies the treatment, outcome and confounding variables. Table 1. Definitions of variables and their measurement.1 Variables Treatment variable cooperative membership Outcome variables income from agriculture (including coffee) income from agriculture (excluding coffee) income from coffee total volume of supply volume of berries supplied price received yield of berries yield of dry coffee net margin Confounding variables age years of schooling active household members

Type

Definitions and measurements

dummy

1 if member, 0 otherwise

continuous total income (in birr) obtained from the sale of all agricultural products including coffee in the 2012-2013 season continuous total income (in birr) obtained from the sale of agricultural products excluding coffee in the 2012-2013 season continuous total income (in birr) obtained from the sale of both berries and dry coffee in the 2012-2013 season continuous the amount of marketed berries and dry coffee (in kg) in the 2012/13 season continuous the amount of marketed coffee berries (in kg) in the 2012-2013 season continuous price (birr/kg) received from the sale of coffee berries in the 20122013 season continuous yield of berries (kg/hectare) produced in the 2012-2013 season continuous yield of dry coffee (in kg) obtained from a hectare of berries in the 2012-2013 season continuous net margin (in birr/kg) obtained from sale of coffee berries in the 2012-2013 season

continuous age of the household head in years continuous years of schooling of the household head continuous number of family members aged between 15 and 65 within a household area of coffee land continuous area of farm land planted with coffee (hectares) distance from the cooperative’s continuous time (in hours) needed by the farmers to travel to reach their cocoffee collection point operative’s coffee collection point (assuming travel on foot). availability of off-farm income dummy 1 if a household has an off-farm source of income, 0 otherwise. risk 1 of effect of price dummy 1 if ‘high’ and 0 otherwise volatility on coffee income risk 2 of effect of price dummy 1 if ‘medium’ and 0 otherwise volatility on coffee income zonal location dummy 1 if Jima, 0 if Kaffa living in certified cooperative dummy 1if yes, 0 otherwise village

1 10

birr = 0.45 USD, calculated on the basis of the exchange rate on October 11, 2016. International Food and Agribusiness Management Review

Shumeta and Dâ&#x20AC;&#x2122;Haese

Volume 19, Issue 4, 2016

The choice of the explanatory variables set out in Table 1 is made on the basis of available empirical studies on the determinants of cooperative membership. In relation to the household characteristics, Bernard et al. (2008), Bernard and Spielman (2009) and Abebaw and Haile (2013) have shown that the age of the household head is positively correlated with the likelihood of cooperative membership. Bernard and Spielman (2009) and Verhofstadt and Maertens (2015) illustrate a positive relationship between education level and the probability of cooperative membership. They also have depicted a direct and significant relationship between the number of economically active household members and the likelihood of cooperative membership. In terms of farm characteristics, Bernard et al. (2008), Bernard and Spielman (2009), Fischer and Qaim (2012) and Abebaw and Haile (2013) have found a positive relationship between the size of landholding and cooperative membership. Landholding size also may influence being a member in our two study areas since some of the cooperatives set a minimum coffee land size (0.25 or 0.5 hectare) as a requirement for membership. The literature reports mixed results on the relation between market or road distance and cooperative membership. Fischer and Qaim (2012) and Abebaw and Haile (2013) showed a direct and significant link between cooperative membership and the distance to the nearest road, although Verhofstadt and Maertens (2015) found a significant negative effect of market distance on cooperative membership. In this study, we assume an inverse relation between the distance to the cooperativeâ&#x20AC;&#x2122;s coffee collection point and the probability of cooperative membership; as farmers who live nearby may potentially benefit more from the marketing services that the cooperative provides. While Fischer and Qaim (2012) and Abebaw and Haile (2013) show a positive relation between off-farm income and cooperative membership, we assumed the opposite relation in this case study since having diverse sources of income makes farmers less vulnerable to poverty and potentially less likely to engage in collective action to safeguard their income from coffee. Jena et al. (2012), Mujawamariya et al. (2013) and Abate et al. (2014) mention that cooperatives are viewed as a safety net that protect their member farmers from low and fluctuating prices in the mainstream market. Hence, we assumed that feeling at risk of coffee price volatility would be an incentive for farmers to become members of a cooperative. Finally, zonal and certified village dummy variables were introduced to capture other institutional, market and socio-economic heterogeneities between the sample zones and villages that might otherwise remain unobserved.

4. Descriptive results Supplementary Table S1 gives an overview of the main characteristics of the cooperatives from which we drew our sample. Coffee is the main cash crop sold by the cooperatives. There are also other crops channeled via these cooperatives such as maize and fruits. The cooperatives were all established between 1976 and 1982. Most are certified to sell Fair Trade and organic coffee, a few not. The size of the membership varied greatly. In general the registration fees were low and the value paid in shares was reasonable. Table 2 compares the observable characteristics of households that were cooperative members and those that were not. Household heads who were members of cooperatives were on average older than non-members. Members, on average, had more land planted with coffee than non-members. Meier zu Selhausen (2016) also found out that female members owned more than the non-members in his study of determinants of women cooperative participation in Uganda. Certified villages had a higher proportion of cooperative members. In terms of the selected outcome performance variables, a substantially higher volume of supply and income from agriculture (with or without coffee) was noticed among member farmers than non-members, although there was no significant difference between members and non-members in terms of income from coffee, yield, the price received and margins obtained. However, these results cannot be used to draw inferences about the impact of coffee cooperatives on the performance of member farmers since other confounding factors would need to be controlled for.

International Food and Agribusiness Management Review

Shumeta and Dâ&#x20AC;&#x2122;Haese

Volume 19, Issue 4, 2016

Table 2. Comparison of the characteristics of cooperative members and non-members from survey data 2012-2013.1 Members age number of years of schooling family members in the productive age range (15<age<65) amount of land planted with coffee (ha) distance to coffee collection point of the cooperative (hours) availability of off-farm income (1=yes) risk of price volatility on coffee income (scale from 1=low to 3=high) zonal location (1=Jima) living in certified village (1=yes) income (birr) from agriculture including coffee from agriculture excluding coffee from coffee berries and dry coffee volume supplied (kg) berries and dry coffee berries only price received for berries (birr/kg) yield (kg/ha) of berries of dry coffee net margin on berries (birr/kg) 1

Non-members

t-values2 P-values

mean

std. dev.

mean

std. dev.

47.56 5.34 4.28

9.69 2.55 1.92

40.37 4.95 4.01

8.74 2.39 1.85

6.22*** 1.26 1.15

0.00 0.21 0.25

1.33 0.35

0.94 0.26

0.72 0.33

0.55 0.26

6.27*** 0.80

0.00 0.43

0.09 1.90

0.29 0.66

0.10 2.05

0.31 0.77

0.58 0.78

0.50 0.42

0.56 0.69

0.50 0.47

-0.37 -1.63

0.71 0.10

0.18 1.72*

0.86 0.09

34,994 17,765.98 11,307.77 94,35.69 23,686.29 14,556.20

29,626.57 15,837.82 8,504.44 10,133.87 21,293.85 13,821.73

2.546** 2.29** 1.34

0.011 0.023 0.18

961.03 775.96 9.36

801.91 711.11 1.02

639.32 494.38 9.19

660.09 548.24 1.28

3.09*** 3.53*** 1.16

0.00 0.00 0.25

1,330.96 443.65 8.25

1,028.90 342.97 1.04

1,420.42 470.86 8.05

827.34 278.64 1.08

-0.76 -0.69 1.52

0.45 0.49 0.13

10 birr = 0.45 USD, calculated on the basis of the exchange rate on October 11, 2016. and * denote significance at 0.01, 0.05 and 0.1 levels, respectively.

2 ***, **

5. Econometric results The econometric results are presented in the following three subsections. The first subsection provides the results of the estimation of the propensity scores and the probability of cooperative membership. The second sub-section presents the results of the PSM on the impact of cooperative membership on the selected performance indicators. Finally, a third sub-section discusses the heterogeneous treatment effect of cooperative membership on the performance indicators with cooperatives and among farm households. Estimation of propensity scores and the probability of cooperative membership We used a logit model to estimate the propensity scores and the probability of cooperative membership. A substantial number of covariates in the model showed the expected associations (Table 3). The mean value of the estimated propensity scores for the whole sample was 0.5156, with minimum and maximum values of 0.0138 and 0.9899 respectively. The propensity scores of the control group were between 0.0138 and 0.9618 with a mean score of 0.3425 while for the treated group these figures ranged from between 0.0713 and 0.9899 with a mean score of 0.6783. Hence, the region of common support for the distribution of the estimated propensity scores of the control (non-member) and treated (member) groups ranged between

International Food and Agribusiness Management Review

Shumeta and Dâ&#x20AC;&#x2122;Haese

Volume 19, Issue 4, 2016

Table 3. Results of estimates of the probability of cooperative membership (logit model) survey data 20122013. Variables

Marginal effect (dy/dx)1

Standard error

age (years) age squared schooling (years) schooling squared active hh members land planted with coffee (ha) land planted with coffee squared distance to coffee collection point (hours) off-farm income2 (1=yes) risk of price volatility2 (1=high) risk of price volatility2 (1=medium) zonal location2 (1=jima) living in certified village2(1=yes) pseudo r-square lr ch2 (13) prob>Ď&#x2021;2 % predicted correctly n

0.06* 0.00 0.03 0.00 0.01 0.49*** -0.05 -0.28* -0.18 0.30*** 0.01 0.19** 0.06 0.27 96.29*** 0.00 52.55 256

0.03 0.00 0.06 0.01 0.03 0.17 0.05 0.16 0.14 0.10 0.10 0.09 0.09

1 ***, ** 2

and * denote significance at 0.01, 0.05 and 0.1 levels, respectively. Marginal effects are calculated for a discrete change of dummy variable from 0 to 1.

0.0713 and 0.9618; this accounts for 127 members and 124 non-members with only 5 members outside this range. The propensity scores in the common support region were used to estimate the ATT. The estimation results (Table 3) revealed that cooperative membership was directly correlated with some household, farm and risk-related characteristics. Older household heads were more likely to be members of cooperatives. Households further away from the cooperativesâ&#x20AC;&#x2122; coffee collection points were less likely to be members. The estimated marginal effect indicated that, for each additional hour of travel to the coffee collection point, the likelihood of belonging to a cooperative decreased by 28%. In addition having more land planted to coffee was positively and significantly correlated with the probability of being a cooperative member. For each hectare of coffee cultivated the likelihood of being a cooperative member increased by 49%. This result is contrary to the findings of Verhofstadt and Maertens (2015) who found that limited access to land was one of the determining factors for land-poor households participating in cooperatives in Rwanda. It is however in line with the findings of Abebaw and Haile (2013). Respondents who said that they felt a high risk effect of price volatility on their income from coffee were also more likely to be cooperative members. This suggests that members see cooperatives as providing a safety net against price risks. Our results also suggested a positive and significant geographical influence on the probability of cooperative membership, with membership levels being higher in Jimma. Other variables, such as years of schooling, family size in the productive age group, availability of off-farm income and living in certified village did not have any significant impact on the likelihood of cooperative membership. The overall treatment effects of membership in coffee cooperatives The ATT was computed, using the kernel matching technique, in order to assess the impact of cooperative membership on the selected performance indicators: total income, income from coffee, agricultural income without coffee, total coffee supply, yields of berries and dry coffee, prices received and net margins obtained. International Food and Agribusiness Management Review

Shumeta and Dâ&#x20AC;&#x2122;Haese

Volume 19, Issue 4, 2016

We also employed neighborhood matching to check the robustness of the PSM estimates obtained from the kernel matching. Both matching methods showed that cooperative membership did not have a significant impact on any of the performance indicators (Table 4). Balancing the covariates In order to fulfill the balancing requirements of PSM, a balancing test was used to verify whether all the observed covariates were similar between members and non-members after matching (Supplementary Table S2). The results depict that the unmatched samples showed a systematic difference between members and non-members in terms of a number of observed characteristics: age, size of coffee land, number of family members in the productive age range and risk perception of price volatility. After the kernel based and nearest neighbor matching, there was no systematic difference in the observed characteristics of members and non-members, as depicted by the insignificant t-statistics for both sets of results. The percentage bias values of the covariates are all below 20% after matching, suggesting that the differences after both matching procedures were not significant. Only one variable has a percentage bias value that is slightly above twenty, which is tolerable in the PSM balancing. In relation to unobservable and hidden biases, we assumed that a positive and significant average treatment effect might partially result from member households having relatively better unobservable characteristics (for example talent, entrepreneurship or risk preference), as opposed to their solely being a result of the effect of cooperatives. In such cases, sensitivity analysis can be used to assess whether the ATT is overestimated as a result of those unobservable characteristics. Since our results indicated that cooperative membership had an insignificant impact on selected performance indicators it was not meaningful to do a sensitivity analysis since the insignificant impact of cooperative membership also reveals an absence of any hidden biases between members and non-members that would suggest that cooperative membership has a positive and significant impact (Faltermeier and Abdulai, 2009, cited in Abebaw and Haile, 2013; Hujer et al., 2004). Table 4. Estimates of the average treatment effect on the treated (ATT) from survey data 2012-2013.1,2,3 Outcome variables

Kernel matching

volume of total coffee berries and dry coffee supplied (kg) Ln (total income from coffee) Ln (total income from agriculture including coffee) Ln (total income from agriculture other than coffee) volume berries supplied (kg) price berries (birr/kg) yield of berries (kg/ha) yield of dry coffee (kg/ha) net margin for berries (birr/kg)

NN matching (5 neighbors)

ATT coef. BSE

z-value P-value

ATT coef. BSE

z-value P-value

-180.64

202.90

-0.89

0.37

-100.08

183.80

-0.54

0.59

-0.27 -0.03

0.16 0.14

-1.71 -0.18

0.09 0.86

-0.22 0.03

0.17 0.18

-1.27 0.14

0.20 0.88

0.30

0.23

1.31

0.19

0.35

0.24

1.44

0.15

-38.97 -0.22 -49.59 -16.53 -0.37

163.91 -0.24 0.20 -1.12 164.07 -0.30 55.19 -0.30 0.37 -0.98

0.81 0.26 0.76 0.76 0.33

32.38 -0.19 -28.88 -9.63 -0.37

159.32 0.20 0.20 -0.98 161.49 -0.18 48.84 -0.20 0.42 -0.90

0.84 0.33 0.86 0.84 0.37

1 Estimates of the matching were obtained using â&#x20AC;&#x2DC;psmatch 2â&#x20AC;&#x2122; command (Leuven and Sianesi, 2003) in Stata (StataCorp LP, College

Station, TX, USA). 2 BSE = bootstrap standard error; these values are calculated with number of replications of 100. 3 10 birr = 0.45 USD, calculated on the basis of the exchange rate on October 11, 2016.

International Food and Agribusiness Management Review

Shumeta and Dâ&#x20AC;&#x2122;Haese

Volume 19, Issue 4, 2016

The heterogeneous treatment effect between cooperatives The average ATT of performance indicators across cooperatives is given in Supplementary Table S3. It is rather tricky to interpret these figures, yet the main point is that the means of the ATT of members across cooperatives were different. The samples for each individual cooperative in this study were too small to have separate propensity score matches; hence it is not possible to verify if, for each of the cooperatives, the mean ATT reported is statistically significantly different from zero. Yet, the means do give an indication of the heterogeneity between cooperatives. The cooperatives in Supplementary Table S3 are arranged along a scale of the mean ATT on agricultural income. Cooperatives, such as Wodiyo, Dirri and Emicho, which performed best on this criterion also performed better on volumes and prices for berries, but not on net margins. It is difficult to explain why these cooperatives perform well on these criteria. The cooperatives are presented in the same order as in Supplementary Table S1, in which we list their basic characteristics. The top three cooperatives have several features in common: they are located in Kaffa, they were certified, and were amongst the smallest in terms of membership. However, location may not mean that all members in Kaffa will be better off (see also next section). Two other cooperatives in Kaffa were amongst the poorest performers. The heterogeneous treatment impact of cooperative membership The estimated ATT value of all the outcome variables assumes that the impact of cooperative membership is homogenous among all members. However, average treatment effects can also hide considerable heterogeneity of this impact between member farmers (Abebaw et al., 2010; Abebaw and Haile, 2013; Ali and Abdulai, 2010; Bernard et al., 2008; Cunguara and Darnhofer, 2011). We refined our analysis to try to assess the heterogeneity of the impact of cooperative membership across households. To this end, ordinary least square regressions were estimated to express the relation of some of the household and farm characteristics of member farmers with the estimated ATT values of all the outcome variables considered (see Supplementary Table S4). This approach has been used before (e.g. Abebaw and Haile, 2013; Verhofstadt and Maertens, 2015) for estimating heterogeneous impacts. A visual inspection of the heterogeneous impact was made by plotting the ATT values and the covariates (Supplementary Figures S2-S8). Our results corroborate those of other papers (Abebaw and Haile 2013; Bernard et al., 2008; Verhofstadt and Maertens, 2015) and demonstrate that not all members benefited equally from membership of a cooperative. Significant heterogeneities were observed that were related to the demographic, farm and the physical characteristics of member households (Supplementary Table S4). Heterogeneity in demographic characteristics The results show a positive and significant impact of cooperatives for relatively older member farmers in all of the performance criteria considered. For farmers who have enjoyed more education, membership seemed to have a larger impact on the total volume of coffee supplied and income from agriculture and coffee. These findings imply that cooperatives were less effective in improving performance for younger and less literate members. Yet, the effectiveness of cooperatives for old members can also be attributed to the lower likelihood of young farmers becoming members of cooperatives (Table 3). Despite the insignificant impact of education on the membership of coffee cooperatives (Table 3), the results suggest that incentives provided by cooperatives were largely utilized by the more knowledgeable member farmers. Heterogeneity in the size of coffee land owned A positive and significant impact of cooperative membership on income from agriculture and coffee, volume of supply, price received and yield was observed for members with large farms, implying that the cooperatives are less effective for smaller scale farmers. Cooperatives contribute to economies of scale in International Food and Agribusiness Management Review

Shumeta and Dâ&#x20AC;&#x2122;Haese

Volume 19, Issue 4, 2016

inputs, market access and a reduction in transaction costs and large farms that use more inputs and supply more coffee to the market can clearly benefit more. This result is similar to the findings of Verhofstadt and Maertens (2015) but contradicts the findings of Ito et al. (2012) and Fischer and Qaim (2012) who showed that cooperatives have a positive impact on small farms. Part of the result may be explained by that fact that land size is positively correlated with cooperative membership (Table 3). This may be the result of some cooperatives imposing physical capital constraints on membership, excluding the smaller-scale farmers from reaping the benefits of cooperative membership. Heterogeneity in market access and location Distance to the coffee collection point was positively associated with the ATT on the amount of coffee supplied. This could be related to the cooperativeâ&#x20AC;&#x2122;s marketing activities which can induce supply by reducing transaction costs, which are higher for more distant farms. Interestingly, farmers in distant places are less likely to join cooperatives (Table 3), although the potential benefits for them are large. Verhofstadt and Maertens (2015) found a similar positive relationship between market distance and the effectiveness of the cooperative, while Abebaw and Haile (2013) reported that market distance was negatively associated with the adoption of agricultural technologies (fertilizers) by member farmers. The impact of cooperatives on coffee income, volume of supply, and margins obtained was higher for member farmers living in Jimma than those in Kaffa. This can be traced to the better infrastructural facilities and services in Jimma which give easier access to markets, increases information sharing and also contribute to the higher probability of farmers in Jimma being cooperative members (Table 3).

5. Discussion Contrary to our hypothesis, cooperatives in the case study areas did not bring clear economic leverage to all smallholder members. Overall the impacts of cooperative membership on the performance of coffee farm households in the areas were insignificant although there were considerable differences between cooperatives and individual members. We identify three important institutional factors to explain why coffee cooperative membership overall has an insignificant impact. First, cooperatives in south-west Ethiopia are heavily financially constrained from purchasing coffee from their members. Most of the cooperatives (especially in Jimma) are in huge debt and have already lost trust from banks for borrowing money. Even when cooperative banks were established for the purpose of providing credit or loans to cooperatives, the service was not effective â&#x20AC;&#x201C; due to a range of different administrative and technical factors. Thus, the cooperatives are forced to get loans through the unions to which these cooperatives belong; the unions borrow money from the banks and transfer it to their member cooperatives. However, the money obtained through the unions is not always delivered on time and is insufficient to purchase all the coffee from member farmers at prices that are competitive with traders who pay cash immediately. Consequently, cooperatives are not able to offer a significantly better price for coffee to their members than that received by non-members working in the conventional spot market. In addition to the price issue, cooperatives are also heavily constrained by the ways they make payment. Traders in the mainstream independent market make full payment to producers immediately upon purchase, whereas with the cooperatives there are payment lags until all outstanding debts and costs are settled. Coupled with the price problem, such payment delays inevitably impairs the ability of cooperatives to make coffee a more lucrative business for more marginal producers or to transform the power asymmetry in the mainstream/conventional market. Financial constraints also mean that the cooperatives are not in a position to provide credit to their members. Government sponsored micro-finance schemes are the only financial institutions that provide credit services to producers in the study areas. However, they are not able to provide these services to all the producers who need them. As a result, a significant number of member farmers (more than 50%) are forced to have an interlocked contractual agreement with traders in which they take money in the form of loans with a promise to supply an equivalent amount of coffee to settle the debt at harvest time. In these contracts, the prices are International Food and Agribusiness Management Review

Shumeta and D’Haese

Volume 19, Issue 4, 2016

set by the traders at the time of delivering the loan with no possibility of improvement even if there is an increase in price at the time of harvest/supply. The study by Mujawamariya et al. (2013) also confirmed the need for credit as one main reason for cooperative farmers to side sell significant proportion of their coffee to traders in the mainstream market. Efforts should therefore be exerted to make the cooperative banks work in line with the credit requirements of the cooperatives. There should also be a mechanism for engaging individual cooperatives with other banks and introduce the uses of revolving funds. This can build the trust that the cooperatives have lost with the banks and eliminate the current inefficient practice of getting loan from banks through the union. The study by Pitt et al. (2006) also emphasized the relevance of establishing a smooth working relation between cooperatives and banks which could enable cooperatives to get bank loan with a relatively low interest rate and longer repayment period. Cooperatives should also be encouraged to establish a credit and saving unit in their internal structure which motivates member farmers to save and then deliver them the loan when the need arises. Second, provision of training and extension services by cooperatives help member farmers get the required knowledge and technical skill to improve their production/productivity and hence income. For example, the study by Meier zu Selhausen (2016) found out that previous trainings and extension services affect production and women’s choice to market their coffee through the cooperatives. Nonetheless coffee cooperatives in the study area are not in a position to provide sufficient training and technical advice to their members. Due to the limited number of experts and low commitment of the cooperative management, trainings are organized only once a year in collaboration with the district bureau of agriculture. As a result, the yield and the possible income benefit that cooperatives should have realized from the use of improved techniques of production could not be achieved at the required level. Third, the cooperative leaders lack managerial competencies. Almost all the cooperatives in the region are led by illiterate farmers who have no knowledge and skill in modern organizational management. Some 36% of respondents gave their cooperative leadership a low credibility, highlighting an absence of transparency and accountability on the management of some cooperatives in the region. Even if cooperatives are not in a position to deliver direct significant economic benefit to their members, it is imperative to remember the indirect economic impact that cooperatives bring to their members via improving the working of markets and competition. The fierce price competition between cooperatives and traders wishing to purchase coffee from farmers leads private traders to adjust their price to what the cooperatives offer and this in turn means producers get a better price for their product (this is known as the competitive yardstick effect). Coffee cooperatives are therefore used as a safety-net by member farmers against being exploited by traders in the region despite them having an insignificant direct impact in improving incomes. The study by Chagwiza et al. (2016) and Mujawamariya et al. (2013) mentioned the importance of cooperatives for inducing a general higher price at the local level and pointed out their safety-net role for their members.

6. Conclusions Despite the turbulent history of cooperatives mainly associated with the highly centralized governance of Ethiopia’s socialist regime of 1974-1991, the present government of FDRE has expressed renewed interest in collective action to promote greater market participation by smallholders. In this paper we explored if cooperative membership does really impact on farmers’ performance. Using a matching technique on household income, yield, volume of supply, price received and margins obtained as indicator variables, we evaluated the overall and heterogeneous impact of coffee cooperatives on performance of member farm households in south-west Ethiopia. Our results suggest that coffee cooperative membership does not have a significant overall impact on the performance of member farm households in any of the selected performance indicators. Yet, these average International Food and Agribusiness Management Review

Shumeta and Dâ&#x20AC;&#x2122;Haese

Volume 19, Issue 4, 2016

values hide considerable heterogeneity across member households. An analysis of the heterogeneity of these treatment effects showed that the impact levels differed across cooperatives and that cooperatives were more effective for member households whose household head is relatively older, educated and with a larger coffee farm. From a policy perspective, our findings stress the need to design strategies to improve the functioning of cooperatives through developing their financial power, and the competency of the members and management personnel, in order to promote the development of the coffee sector. The fact that cooperative membership and effectiveness are positively correlated with age and size of land under coffee suggests that cooperatives should avoid placing entry barriers based on human and physical capital and should be more welcoming to young and small-scale farmers, encouraging their membership and helping them become more effective. Our findings on the negative selection of the estimated income and supply effects of cooperative membership with distance to the cooperativesâ&#x20AC;&#x2122; coffee collection point implies the possibility of expanding membership and calls for continued promotion of cooperatives in more distant places. The higher probability of cooperative membership and effectiveness in the Jimma area also shows the need for a concerted effort to empower and promote cooperatives in the Kaffa area in order that they can attract more members and improve their efficacy. As cooperatives are relevant institutions to safeguard producers from the adverse effect of market liberalization, empowering them financially through the provision of credit and helping them organize themselves in a business/entrepreneurial principle is vital for leveraging their competitive power in the market and improve their contribution to the income of member households. Our results demonstrate the relevance of looking beyond overall treatment outcome, examining heterogeneous effects and assessing the impact of institutional innovation in the agricultural sector. We realize that the issue of gender is pertinent when dealing with coffee production and marketing though we couldnâ&#x20AC;&#x2122;t handle it due to the limited number of female headed households in the area. We therefore suggest future studies to give more focus on gender disaggregated impact assessment of cooperative membership. Finally, we would like to notify that our findings are not necessarily applicable to other coffee cooperatives in Ethiopia or elsewhere in Africa since the samples are relatively small and taken from specific localities.

Supplementary material Supplementary material can be found online at https://doi.org/10.22434/IFAMR2015.0110. Table S1. Overview of cooperative characteristics. Table S2. Results of balancing tests. Table S3. Mean (+ standard deviation) of the average treatment effect on the treated of selected variables by cooperative. Table S4. Heterogeneous impact of cooperative membership on income, volume of supply, yield, price received and margins obtained from survey data 2012-2013. Figure S1. IMap showing the location of the study sites (Jima and Kaffa). Figure S2. Heterogeneity of the average treatment effect on the treated (ATT) of income from agriculture including coffee over different covariates. Figure S3. Heterogeneity of ATT of income from coffee over different covariates. Figure S4. Heterogeneity of ATT of total volume of supply of cherries and dry coffee over the different covariates. Figure S5. Heterogeneity of ATT of volume of supply of cherries over the different covariates. Figure S6. Heterogeneity of ATT of yield over different covariates. Figure S7. Heterogeneity of ATT of price over different covariates. Figure S8. Heterogeneity of ATT of net margin over different covariates.

International Food and Agribusiness Management Review

Shumeta and D’Haese

Volume 19, Issue 4, 2016

Acknowledgements The authors are grateful for The Netherlands Organization for Cooperation in Higher Education (NUFFIC) for financing this research project.

References Abate, G.T., G.N. Francesconi and K. Getnet. 2014. Impact of agricultural cooperatives on smallholders’ technical efficiency; evidence from Ethiopia. Annals of Public and Cooperative Economics 85: 257-286. Abebaw, D., Y. Fentie and B. Kassa. 2010. The impact of the Food Security Program on household food consumption in north-western Ethiopia: a matching estimator approach. Food Policy 35: 286-293. Abebaw, D. and M.G. Haile. 2013. The impact of cooperatives on agricultural technology adoption: empirical evidence from Ethiopia. Food Policy 38: 82-91. Ali, A. and A. Abdulai. 2010. The adoption of genetically modified cotton and poverty reduction in Pakistan. Journal of Agricultural Economics 61: 175-192. Barham, J. and C. Chitemi. 2009. Collective action initiative to improve marketing performance: lesson from farmer groups in Tanzania. Food Policy 34: 53-59. Becker, S. and A. Ichino. 2002. Estimation of average treatment effects based on propensity scores. The Stata Journal 2: 358-377. Berdegue, J. 2001. Cooperating to compete, associative peasant business firms in Chile. Ph.D diss., Wageningen University, Wageningen, the Netherlands. Bernard, T. and B.J. Spielman. 2009. Reaching the rural poor through rural producer organizations? A study of agricultural marketing cooperatives in Ethiopia. Food Policy 34: 60-69. Bernard, T., A.S. Taffesse and E. Gabre-Madhin. 2008. Impact of cooperatives on smallholders’ commercialization behavior; evidence from Ethiopia. Agricultural Economics 39: 147-161. Caliendo, M. and S. Kopeinig. 2008. Some practical guidance for the implementation of propensity score matching. Journal of Economics Survey 22: 31-72. Cazzuffi, C. and A. Moradi. 2012. Membership size and cooperative performance: evidence from Ghanaian cocoa producers’ societies, 1930-1936. Economic History of Developing Regions 27: 67-92. Chagwiza, C., R. Muradian and R. Ruben. 2016. Cooperative membership and dairy performance among smallholders in Ethiopia. Food Policy 59: 165-173. Collin, M.H. and P. Rondot (eds. 2001). Agricultural producer organizations, their contribution to rural capacity building and poverty reduction. The World Bank, Washington DC, USA. Cungura, B. and L. Darnhofer. 2011. Assessing the impact of agricultural technologies on household income in rural Mozambique. Food Policy 36: 378-390. Dehejia, R. and S. Wahba. 2002. Propensity score matching methods for non-experimental causal studies. Review of Economics and Statistics 84: 151-161. Djebbari, H. and J. Smith. 2008. Heterogeneous impacts in PROGRESA. Journal of Econometrics 145: 64-80. Federal Democratic Republic of Ethiopia (FDRE), 1994. Agricultural Cooperative Societies Proclamation No. 85/1994. Negarit Gazeta 1 (45), 1 February 1994, 264-274. Available at: http://faolex.fao.org/ docs/pdf/eth8955.pdf. Federal Democratic Republic of Ethiopia (FDRE), 1998. Cooperative Societies Proclamation (No. 147/1998). Federal Negarit Gazeta 5(27), 29 December 1998, 942-956. Available at: http://faolex.fao.org/docs/ pdf/eth44002.pdf. Fischer, E. and M. Qaim. 2012. Linking smallholders to market; determinants and impacts of farmers’ collective action in Kenya. World Development 40: 1255-1268. Francesconi, G. N. and N. Heerink. 2011. Ethiopian agricultural cooperatives in an era of global commodity exchange: does organization form matter? Journal of African Economics 20: 153-177. Heckman, J., H. Ichimura, J. Smith,and P. Todd. 1997. Matching as econometric evaluation estimator; evidence from evaluating a job-training program. Review of Economic Studies 64: 605-654.

International Food and Agribusiness Management Review

Shumeta and D’Haese

Volume 19, Issue 4, 2016

Hujer, R., M. Caliendo and S. Thomson. 2004. New evidence on the effect of job creation schemes in Germany-a matching approach with threefold heterogeneity. Research in Economics .58: 257-302. Jena, R.P., T. Stellmacher and U. Grote. 2012. The impact of coffee certification on small-scale producers’ livelihoods; a case study from Jimma Zone, Ethiopia. Journal of Agricultural Economics 43: 429-440. Imbens, G. 2004. Nonparametric estimation of average treatment effects under exogeneity. Review of Economics and Statistics 86: 4-29. Ito, J., Z. Bao and Q. Su. 2012. Distributional effects of agricultural cooperatives in China; the exclusion of smallholders and potential gains on participation. Food Policy 37: 700-709. Markelova, H. and E. Mwangi. 2010. Collective action for smallholder market access: evidence and implications for Africa. Review of Policy Research 27: 621-640. Leuven, E. and B. Sianesi. 2003. PSMATCH2: Stata module to perform full mahalanobis and propensity score matching, common support graphing and covariate imbalace testing. Available at: http://tinyurl. com/jhanz7t. Meier zu Selhausen, F. 2016. What determines women’s participation in collective action? Evidence from a western Uganda coffee cooperative. Feminist Economics 22: 130-157. Mujawamariya, G., M. D’Haese and S. Speelman. 2013. Exploring double side-selling in cooperatives, a case study of four coffee cooperatives in Rwanda. Food Policy 39: 72-83. Mutuc, M., R.M. Rejesus and J.M. Yorobe. 2013. Which farmers benefit the most from Bt corn adoption? Estimating heterogeneity effects in the Philippines. Agricultural Economics 44: 231-239. Pingali, P., Y. Khwaja and M. Meijer. 2005. Commercializing Small farms; reducing transaction costs. ESA working paper: 05-08. The Food and Agriculture Organization (FAO, Rome, Italy. Pitt, M.M., R.S. Khandker and J. Cartwright. 2006. Empowering women with micro finance; evidence from Bangladesh. Ph.D Diss., University of Chicago, Chicago, IL, USA. Rosenbaum, P.R and D.B. Rubin. 1985. Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. American Statistician 39: 33-38. Vandeplas, A., B. Minten, and J. Swinnen. 2013. Multinationals vs cooperatives; the income and efficiency effects of supply chain governance in India. Agricultural Economics 64: 217-244. Verhofstadt, E. and M. Maertens. 2015. Can agricultural cooperatives reduce poverty? The heterogeneous impact of cooperative membership on farmers’ welfare in Rwanda. Applied Economic Perspectives and Policy 37: 86-106. World Bank. 2003. Reaching the rural poor, a renewed strategy for rural development. The World Bank, Washington DC, USA. World Bank. 2007. World development report 2008. The World Bank, Washington DC, USA. Xie, Y., Brand, J.E. and Jann, B. 2012. Estimating heterogeneous treatment effects with observational data. Sociological Methodology 42: 314-347.

International Food and Agribusiness Management Review

OPEN ACCESS International Food and Agribusiness Management Review Volume 19 Issue 4, 2016; DOI: 10.22434/IFAMR2016.0024 Received: 11 February 2016 / Accepted: 10 September 2016

Performance of small and medium-sized food and agribusiness enterprises: evidence from Indian firms RESEARCH ARTICLE Jabir Ali Centre for Food and Agribusiness Management, Indian Institute of Management Lucknow, Off Sitapur Road, Prabandh Nagar, Uttar Pradesh 226013, India

Abstract This paper analyses the size of food and agribusiness firms in India in relationship to business enterprise characteristics, performance, and obstacles through surveying 515 food and agribusiness firms operating in different regions of India using the World Bankâ&#x20AC;&#x2122;s Enterprise Survey 2014. Descriptive statistics, chi-square tests and analysis of variance were used to evaluate data using statistical software. Chi-square statistics identify significant differences in enterprise characteristics through examining firm size, location, gender ownership, type and age. An analysis of variance indicates significant differences in business performance across small, medium and large enterprises in term of input and output ratios. Obstacles facing firms are largely similar regardless of firm size in eleven of the sixteen business-obstacles surveyed. Results reveal that large enterprises perceive more challenges with telecommunication services, customs, trade regulations, and corruption, while small and medium firms face greater constraints gaining access to land and finance. This study is useful in helping to design policies that can efficiently support small and medium food and agribusiness enterprise development. Keywords: firm size, small and medium enterprises, business performance, business obstacles, enterprise survey JEL code: C83, D22, L25, L66 Corresponding author: jabirali@iiml.ac.in

ÂŠ 2016 Ali

Ali

Volume 19, Issue 4, 2016

1. Introduction Small and medium-sized food and agribusiness enterprises (SMEs) play a major role in global economic growth and development, particularly in the developing economies (Ayyagari et al., 2007; Berry et al., 2001; Cook, 2001; De and Nagaraj, 2014). The Indian food processing industry is widely recognized as having the potential to transform the Indian economy through large-scale food manufacturing that will not only benefit consumers in the long run but provide future employment and export earning opportunities. The output and employment contributions from these sectors are continuously increasing across regions (Mead and Liedholm, 1998) as demand for high processed foods rises due to urbanization, an expanding middle class, growing health awareness and evolving consumer preferences. The food and agribusiness sector is comprised of a large number of small and medium enterprises, a few large enterprises and multinationals, all competing with each other at every level of the supply chain. A competitive food and agribusiness sector requires effective innovation and entrepreneurship development in order to compete and strengthen growth (ACI and ETG, 2011). Indian food processing segment produces a broad spectrum of products including fruits, vegetables, legumes, spices, meats, poultry, and fisheries, milk and dairy products, alcoholic beverages, grain processing and specialty products such as confectionaries, cocoa products, soya-based and high protein foods, and mineral water, etc. It is estimated that the Indian agribusiness industry generated around $245 billion in 2015 and has grown 5.7% annually from 1991-2015 (BMI, 2016). The Ministry of Micro, Small and Medium Enterprises estimates about 51.06 million of micro-small-medium enterprises, employ about 117.2 million people. The food and agribusiness industry comprises about 6.9% of this sector employing 7.8%. The importance of business development, among small and medium enterprises, is widely recognized across developing and developed nations (Forsman and Temel, 2011; Tanabe and Watanabe, 2005). Dethier et al. (2011) found through an extensive survey of literature that good business practices tend to favor growth by encouraging productivity, further detecting that various infrastructures, finance, security, competition, and regulatory factors all significantly impact enterprise performance. De and Nagaraj (2014) argued that large-scale firms in the Indian manufacturing sector face different opportunities and challenges from small-scale firms. Although small firms managers are more flexible in their ability to respond quickly to market changes, larger firms have advantages of the economies of scale giving them more political clout and better access to government credits, contracts, and licenses. Considering the differences in managerial issues of firms by size, the government has provisioned separate regulatory and developmental arrangements. Moreover, development of micro, small and medium-sized enterprises (MSMEs) are important factors in employment, innovation, economic growth, and equity, so consequently are given policy thrusts in most developing countries. In India, the Micro, Small and Medium Enterprises Development Act 2006 subsequently merged into the Ministry of Small Scale Industries and the Ministry of Agro and Rural Industries in May 2007 to form the Ministry of Micro, Small and Medium Enterprises to better address policy issues affecting MSMEs. Understanding perceived business obstacles, enterprise characteristics and business performance from a scale perspective helps agribusiness managers align with the needs of the supply chain and policy makers in order to design better policy support.

2. Conceptual framework and research hypotheses Empirical evidence shows that firm size affects business performance and decision-making (Chang et al., 2013; Kalkan et al., 2011; Lee, 2009; Lun and Quaddus, 2011; Palmon and Wald, 2002; Vithessonthi and Tongurai, 2015; Youn et al., 2015). While some studies find business performance varies across firm size, others found mixed or no relationship between firm size and business performance (Bourlakis et al., 2014; Orlitzky, 2001). Beck et al. (2005), argue that firm size impacts a firmâ&#x20AC;&#x2122;s productivity, survival, and profitability. Bourlakis et al. (2014) analyzed sustainable performance differences within the Greek food supply chain by making statistical comparisons (of growers, manufacturers, wholesalers, and retailers) related to firm size. Kotey (2005) examined firm size and business performance in relation to profits, growth, efficiency and International Food and Agribusiness Management Review

Ali

Volume 19, Issue 4, 2016

liquidity differences between family and non-family, small-to-medium-sized enterprises (SMEs). Orlitzky (2001) analyzed relationships across firm size, corporate social performance and firm financial performance, concluding there is neither a significant positive correction between firm size and corporate social performance nor between firm size and firm financial performance. According to the literature, most large firms occupying dominant business positions have the advantage of economies of scale and efficiency. Laing and Weir (1999) noted that larger firms normally follow better governance structures and business compliance in achieving high-level corporate performance. Analyzing the legal-economic framework in Mexico, Laeven, and Woodruff (2007), showed that the legal system affects firm size by reducing the idiosyncratic risk faced by firm owners. Examining the linkage between firm size and technological changes, Antonelli and Scellato (2015) conclude that large firms are more likely to introduce science-based technological changes consisting of a shift effect in production functions, whereas smaller firms rely more on tacit, external knowledge involving technologies that use more locally abundant production factors. Researchers exploring the relationship between technical efficiency and firm size find that larger firms experience higher technical efficiency compared to smaller firms (Antonelli et al., 2015; Chow et al., 1997; De and Nagaraj, 2014). This paper analyses the size of food and agribusiness firms in India in relationship to business enterprise characteristics, performance, and obstacles. Nichter and Goldmark (2009) highlighted the following four types of factors affecting the growth and performance of business enterprises: (1) individual entrepreneur characteristics; (2) firm characteristics; (3) relational factors such as social networks or value chains; and (4) contextual factors such as the business obstacles. The following three hypotheses are formulated and tested in the study (Figure 1): H1: there are no differences in profiles among food and agribusiness enterprises across small, medium and large size firms. Several empirical studies have highlighted the importance of enterprise profile variables such as size, age, nature of ownership and location in understanding the performance of the firms (Coad and Tamvada, 2012; Shanmugam and Bhaduri, 2002). The location of firms is influenced by a variety of factors such as taxation, access to raw materials and availability of markets (Hebous et al., 2011; Sridhar and Wan, 2010). Sridhar and Wan (2010) found that large firms in India are normally concentrated in smaller cities, which is a surprise as larger cities have better provisions, infrastructure, and other public services. The evidence clearly indicates that the size of firms is an important implication in firm ownership and management by gender (Bardasi et al., 2011; Coleman, 2007). Ensuring gender diversity on corporate boards has become a mandatory provision under the New Companies Act of 2013 to improve corporate governance (Shrivastava and Chakraborty, 2015). Since 2013, the Securities and Exchange Board of India is implementing gender diversity in boardrooms. Shanmugam and Bhaduri (2002) analyzed that age positively influences growth across Indian manufacturing sector. Although, Park et al. (2010) found that firm size and age negatively impacted firm growth while positively impacting firm survival in the Korean manufacturing industry. H2: there are no differences in business performance among food and agribusiness enterprises across small, medium and large size firms. Empirical evidence finds some indicators were used to analyze business performance across firm size (Bourlakis et al., 2014; Campos-Climent and Sanchis-Palacio, 2015; Lee, 2009; Orlitzky, 2001). Chen (2009) argued that enterprise competitiveness is derived from the performance of its business units in the form of input-output relationships. Ponikvar et al. (2009) investigated the impact of firmsâ&#x20AC;&#x2122; growth rate based on various financial and non-financial performance ratios such as revenue per employee, average costs, labor costs, capital costs, capacity utilization, productivity, and efficiency.

International Food and Agribusiness Management Review

Ali

Volume 19, Issue 4, 2016 input ratios

(wages & salaries, raw material, fuel and rent)

business performance output ratios (sales, capacity utilization)

H2 size of the firms location ownership

enterprise characteristics

business obstacles

access to resources business regulations market externalities

age

Figure 1. Conceptual research framework and hypothesis testing. H3: there are no differences in business obstacles among food and agribusiness enterprises across small, medium and large size firms. Although De and Nagaraj (2014) assert that problems and challenges faced by small, medium and large firms vary greatly, several studies find that small firms face larger growth constraints and have less access to formal sources of finance (Beck et al., 2006; Hutchinson and Xavier, 2006; Kumar and Rao, 2016; Ramukumba, 2014). A variety of factors hinder business performance broadly categorized as access to resources, business regulations and market externalities (Das and Das, 2014).

3. Data and methods Data source This study uses primary data taken using a stratified random sampling of 515 Indian food and agribusiness firms operating in different regions of India, conducted under the World Bankâ&#x20AC;&#x2122;s Enterprise Survey in 2014. The strata for Enterprise Surveys are based on firm size, business sector, and geographic regions within a country. This study is a part of a comprehensive survey of 9,281 firms of different industry sub-sectors across the country. Firm size has been categorized based on the number of employees i.e. 5-19 employees as a small enterprise; 20-99 employees as a medium enterprise; and 100 and above employees as a largesized enterprise. The survey contained information on a variety of firmâ&#x20AC;&#x2122;s characteristics such as ownership, type of firms, size, location, and age; performance indicators regarding input and output ratios; and information on business obstacles as reported by the firms on a 5-points rating scale: 1 = no obstacle; 2 = minor obstacle; 3 = moderate obstacle; 4 = major obstacle; 5 = very severe obstacle. Obstacle indicators surveyed included: electricity, telecommunications, transport, customs and trade regulations, practices of competitors in the informal sector (such as unfair competition), access to land, crime, theft and disorder, access to finance, tax rates, business licensing and permits, corruption, labor regulations, and inadequately educated workforce. To measure the performance of business enterprises across sizes, a number of input and output ratios are used. Input ratios were generated for categories including wages and salary per employees, cost of raw material International Food and Agribusiness Management Review

Ali

Volume 19, Issue 4, 2016

by the total cost of production, cost of fuel by the total cost of production, cost of electricity by the total cost of production and rent for land and machinery by the total cost of production. Similarly, output ratios were generated to understand the business performance of firms in terms of total annual sales by the total cost of production and capacity utilization ratio of the firms. Further, business obstacles have been categorized into access to resources, regulatory system, and business externalities. Data analysis The World Bank’s Enterprise Survey data on food enterprises was analyzed using Statistical Package for the Social Sciences version 20.0 (SPSS, IBM, Chicago, IL, US). Simple statistical tools such as descriptive statistics, cross-tabulation, chi-square test and analysis of variance have been used to understand the business performance differences across size of food and agribusiness firms in India. The difference in enterprise characteristics across firm size was analyzed using the chi-square test statistics as follows: χ2 = ∑(O – E)2 / E With df = (r-1) (c-1), where r and c are the number of possible values for the two variables under consideration. Similarly, differences in business performance examining input-output ratios and business obstacles regarding access to resources, business regulations and market externalities across small, medium and large enterprises were analyzed using the analysis of variance technique.

4. Results and discussion Enterprise characteristics by size Several studies investigating differences in enterprise characteristics by firm size have shown various relationships with enterprise characteristics regarding location, ownership, type and age (Cerdan and Hernández, 2013; De and Nagaraj, 2014; Orser et al., 2000; Vithessonthi and Tongurai, 2015). Of the 515 food and agribusiness enterprises surveyed under the World Bank’s Enterprise Survey, about 52% were small, 30% were medium, and 18% were large firms. Table 1 provides details by firm size, location, type, and age. The results of chi-square statistics clearly indicate that firm characteristics vary significantly by size and substantial differences by size exist across various regions of the country (χ2=58,324, P<0.01). More food and agribusiness SMEs are located in the southern regions, whereas the northern and eastern regions have more large firms. Similarly, small and medium enterprises are primarily located in bigger cities (χ2=17.255, P<0.01). Most of the firms surveyed were small and medium enterprises, and female ownership was comparatively higher in larger firms (χ2=8.585, P<0.05). The result of the chi-square test for female participation in the management of firms reveals significant differences across the size of enterprises (χ2=9,907, P<0.01). Out of the 515 firms surveyed under the World Bank’s Enterprise Survey, about 7% were managed by a top female manager and about 16.5% firms reported having at least one female owner. It is important to note that among the female owners, about 8% reported having 100% ownership of the enterprises, while 26% reported ownership up to 50% and above of the firm’s resources. Distribution of firms by type of ownership also vary with the size of enterprises (χ2=42.002, P<0.01). Most of the small and medium enterprises are mainly sole proprietary/partnership firms whereas large enterprises utilize limited company/partnership. The distribution of enterprises by age across the size of firms also varies significantly (χ2=16.310, P<0.05). It is evident that the number of small and medium firms decline with the age of the firms whereas large firms are comparatively older in age.

International Food and Agribusiness Management Review

Ali

Volume 19, Issue 4, 2016

Table 1. Enterprise characteristics by size. Indicators

Regions Central Eastern Northeastern Northern Southern Western Size of locality City with population over 1 million Over 250,000 to 1 million 50,000 to 250,000 Less than 50,000 Female participation in ownership Yes No Firms with a female top manager Yes No Type of firm Limited company Sole proprietorship Partnership Others Age of the Firm <10 years 10-20 years 21-30 years >30 years 1 **significant

Small enterprise n=268

Medium enterprise n=154

Large enterprise n=93

Chisquare1

58.324**

0.000

17.255**

0.008

8.585*

0.014

9.907**

0.007

42.002**

0.000

16.310*

0.012

26 60 31 40 70 41

9.7 22.4 11.6 14.9 26.1 15.3

7 20 27 37 59 4

4.5 13.0 17.5 24.0 38.3 2.6

11 25 14 28 12 3

11.8 26.9 15.1 30.1 12.9 3.2

96 73 67 32

35.8 27.2 25.0 11.9

32 44 41 37

20.8 28.6 26.6 24.0

26 25 28 14

28.0 26.9 30.1 15.1

32 236

11.9 88.1

32 122

20.8 79.2

21 72

22.6 77.4

16 252

6.0 94.0

6 148

3.9 96.1

13 80

14.0 86.0

13 153 98 4

4.9 57.1 36.6 1.5

11 69 70 4

7.1 44.8 45.5 2.6

21 26 43 3

22.6 28.0 46.2 3.2

57 119 52 39

21.3 44.6 19.5 14.6

36 60 28 30

23.4 39.0 18.2 19.5

26 21 21 25

28.0 22.6 22.6 26.9

at 0.01 level, *significant at 0.05 level.

Therefore, hypothesis H1, which assumes that there is no difference in enterprise characteristics of firms by size, is not accepted, and small, medium and large firms significantly vary in enterprise characteristic parameters. In a nutshell, small and medium food and agribusiness firms comparatively belong to the southern region and are primarily concentrated in larger cities as compared to large firms, which operate in northern and eastern regions. It is interesting to note that SMEs preferably locate themselves in areas with proper infrastructural availability and within close proximity to markets. Female participation in ownership and management of food and agribusiness enterprises also increases with firm size. Similarly, most small and medium enterprises are sole proprietary/partnerships whereas as large enterprises are limited company/ partnerships. The age of the firm is positively correlated with the size. Difference in business performance by size of enterprises The cost of production and output varies with firm size. Business performance is analyzed using a variety of indicators such as inputs and outputs (Watson, 2002); profit margins and employment (Chirwa, 2008); closure rates, return on assets, risk (Robb and Watson, 2012) and the impact of return on assets on business International Food and Agribusiness Management Review

Ali

Volume 19, Issue 4, 2016

growth and survival (Basyith et al., 2014). Table 2 shows the differences in input and output performances across enterprise size. Analysis of Variance (ANOVA) was used to examine variances among the performance indicators. Large enterprises pay higher per unit costs (of producing goods and services) than small and medium enterprises in wages and salaries per employee (F=3.751, P<0.05); and rent by total cost of production (F=3.834, P<0.05). However, raw materials are comparatively less in bigger enterprises than small and medium agribusiness firms (F=3.751, P<0.05). Output is a key indicator to measure performance in terms of annual sales turnover and capacity utilization of firms. The analysis shows that annual sales by per unit cost of production in large firms are higher than small and medium agribusiness firms, which is statistically non-significant (F=10.017, P>0.10). However, capacity utilization across firm size is high for large firms compared to small and medium sized firms and is statistically significant at 0.10 (F=2.545, P<0.10). Therefore, Hypothesis H2, which assumes that there is no difference in business performance across small, medium and large enterprises, is not accepted, and large firms significantly vary on business performance parameters such as wages and salaries, raw material costs, rent and capacity utilization. It is clear from the analysis that large firms pay comparatively higher wages and salaries compared to SMEs as large firms are subjected to comply with more stringent labor regulations. However, SMEs spend more on raw materials due to low economies of scale in handling raw material for processing. Business obstacles by size of firms In a rapidly changing business environment, several factors hinder business performance (Kwong et al., 2012; Mbonyane and Ladzani, 2011; Roomi et al., 2009; Watson, 2006). Beck et al. (2005) explored the implications of financial, legal, and corruption obstacles affecting firms of different sizes. Formal government legislation, policies, and programs play a vital role in facilitating the growth and development of business enterprises (Aidis, 2005; Roxas et al., 2013). In this study, sixteen parameters were used to examine access to resources, regulatory, and business externalities of firms using a 5-points rating scale: 1 = no obstacle, 2 = minor obstacle, 3 = moderate obstacle, 4 = major obstacle, 5 = very severe obstacle. Table 3 provides the analysis of different business obstacles across firm size. Table 3 shows that a firmâ&#x20AC;&#x2122;s response to operational challenges varies greatly by firm size for five out of the sixteen indicators. However, a low mean value of responses concerning business obstacles suggests Table 2. Difference in business performance by size of firms. Business performance indicators

Input ratios Wages and salaries (Rs. in lakhs)/number of employees Cost of raw material/total cost Cost of fuel/total cost Cost of electricity/total cost Rent for land and machinery/total cost Output ratios Total annual sales/total cost Capacity utilization (%)2

Small Medium Large F1 enterprise enterprise enterprise n=268 n=154 n=93

Sig.

1.11

1.15

1.57

3.751*

0.024

0.71 0.03 0.06 0.02

0.75 0.03 0.06 0.02

0.66 0.03 0.05 0.04

3.152* 0.078 0.530 3.834*

2 2 2 2

0.044 0.952 0.589 0.022

1.72 78.8

1.83 77.3

2.23 82.3

1.051 2.545

2 2

0.350 0.079

significant at 0.05 level. The capacity utilization under the World Bankâ&#x20AC;&#x2122;s Enterprise Survey is defined in terms of output produced as a proportion of the maximum output possible if using all the resources available. 2

International Food and Agribusiness Management Review

Ali

Volume 19, Issue 4, 2016

Table 3. Responses on business obstacles by size of firms. Business obstacles

Access to resources Electricity Telecommunications Transport services Access to land Access to finance Business regulations Customs and trade regulations Tax rates Tax administrations Business licensing and permits Labor regulations Market externalities Crime, theft and disorder Courts Practices of competitors Political instability Corruption Inadequately educated workforce 1 **significant

Small enterprise n=268

Medium enterprise n=154

Large enterprise n=93

Sig.

2.7 1.3 2.1 1.9 2.2

2.9 1.2 2.1 1.5 1.9

2.9 1.5 2.2 1.8 2.0

1.475 7.617** 0.156 5.095** 6.149**

2 2 2 2 2

0.230 0.001 0.856 0.006 0.002

1.7 2.6 2.3 2.1 2.1

1.6 2.6 2.2 2.0 2.1

2.1 2.6 2.5 2.0 2.1

7.547** 0.080 2.139 0.133 0.090

2 2 2 2 2

0.001 0.923 0.119 0.876 0.914

1.5 1.7 1.9 2.0 2.8 2.1

1.4 1.7 1.9 2.0 2.7 2.0

1.6 1.8 1.8 1.9 3.1 2.0

1.802 0.274 0.847 0.493 4.044* 0.342

2 2 2 2 2 2

0.166 0.760 0.429 0.611 0.018 0.711

at 0.01 level, *significant at 0.05 level.

that firms are mainly facing minor or moderate levels of challenges in their business operations. Among the most significant differences in mean values are large enterprises who perceive comparatively more obstacles than small and medium enterprises for three indicators: telecommunication (F=7.617, P<0.01), customs and trade regulations (F=7.547, P<0.01) and corruption (F=4.044, P<0.01). This implies that large firms expect improvements in information provisions and communication technologies in order to meet business obligations and compliance with higher tax regulations. Lee et al. (2010) analyzed the incidences of bribery and the size of bribes using the residual control theory and argues that firms pay bribes based on their exposure and vulnerability to residual rights of control by government officials. Similarly, small enterprises perceive comparatively more obstacles than large enterprises for two business obstacles â&#x20AC;&#x201C; access to land (F=5.095, P<0.05) and finance (F=6.149, P<0.01). This clearly indicates that SMEs face challenges in accessing the land and credit for business expansion. Therefore, Hypothesis H3, which assumes that there is no difference in business obstacles across firm size, is largely accepted, as small, medium and large firms perceive similarly on the majority of business obstacles parameters.

5. Conclusions and managerial implications Small and medium enterprises play a crucial role in the growth and development of the economy through generating employment opportunities, reducing regional imbalances, industrialization of rural and backward areas and assuring equitable distribution of resources. The problems and opportunities are different for small, medium versus large enterprises. While small firms have more flexible management and lower response time to market changes, larger firms have the advantages of economies of scale, political clout and better access to government credits, contracts and licenses (De and Nagaraj, 2014). Therefore, there is a need to analyze the nature and magnitude of business performance and obstacles faced by firms across the size of enterprises. The Indian government has adopted a focused approach in developing and promoting MSMEs and large enterprises separately by dedicated central ministries, policies, and plans. International Food and Agribusiness Management Review

Ali

Volume 19, Issue 4, 2016

By acknowledging the differences in managerial needs by firm size, governments are making separate regulatory and developmental provisions to address the issues. As the majority of business enterprises are micro, small and medium, this sector is given a policy thrust in most developing countries. In India, the Ministry of Micro, Small, and Medium Enterprises addresses policy issues affecting MSMEs. This paper analyzed the differences in business performance and obstacles faced by food and agribusiness firms in India. This analysis reveals that business needs vary depending on size. Small and medium food and agribusiness firms were located primarily in the southern region of the country and concentrated in larger cities, while larger firms normally operate in the northern and eastern regions. Female participation in ownership and management of food and agribusiness enterprises increases with firm size. Similarly, most small and medium enterprises are sole proprietary/partnership firms whereas large enterprises are limited company/partnership firms. The age of the firm was positively correlated with the size. This study provides some insight into the differences in firm performance and business obstacles across the small, medium and large enterprise and can serve as a resource in helping researchers, bankers, entrepreneurs, and policymakers develop effective business models which address the greatest challenges faced by small and medium food and agribusiness enterprises. The study is based on a larger survey of data from the World Bank. The secondary data has provided limited choices in selecting the performance indicators of business enterprises as well as business obstacles encountered by these enterprises. Future research can be conceptualized based on theoretical models with suitable indicators by incorporating in-depth interviews of respondents and their characteristic variables.

References Agrifood Consulting International and Economic Transformation Group (ACI and ETG). 2011. Growing food, products, and businesses: applying business incubation to agribusiness SMEs. Agribusiness Incubators Assessment Report, Agrifood Consulting International and Economic Transformation Group. Bethesda, Maryland, MD, USA. Available at: http://tinyurl.com/h47hpto. Aidis, R. 2005. Institutional barriers to small- and medium-sized enterprise operations in transition countries. Small Business Economics 25: 305-317. Antonelli, C., F. Crespi and G. Scellato. 2015. Productivity growth persistence: firm strategies, size and system properties. Small Business Economics 45: 129-147. Antonelli, G. and G. Scellato. 2015. Firms size and directed technological change. Small Business Economics 44: 207-218. Ayyagari, M., T. Beck and A. Demirguc-Kunt. 2007. Small and medium enterprises across the globe. Small Business Economics 29: 415-434. Bardasi, E., S. Sabarwal, and K. Terrell. 2011. How do female entrepreneurs perform? Evidence from three developing regions. Small Business Economics 37: 417-441. Basyith, A., M. Idris and Fitriya. 2014. The gender effect on small business enterprisesâ&#x20AC;&#x2122; firm performance: evidence from Indonesia. Indian Journal of Economics and Business 13: 21-39. Beck, T., A. DemirgĂź Ă&#x2021;-Kunt and V. Maksimovic. 2005. Financial and legal constraints to growth: does firm size matter? The Journal of Finance 60: 137-177. Berry, A., E. Rodriguez and H. Sandee. 2001. Small and medium enterprise dynamics in Indonesia. Bulletin of Indonesian Economic Studies 37: 363-384. BMI 2016. India agribusiness report Q4 2016. BMI Research. Business Monitor International Ltd., London, UK. Bourlakis, M., G. Maglaras, E. Aktas, D. Gallear and C. Fotopoulos. 2014. Firm size and sustainable performance in food supply chains: insights from Greek SMEs. International Journal of Production Economics 152: 112-130. Campos-Climent, V. and J.R. Sanchis-Palacio. 2015. How much does size matter in agri-food firms? Journal of Business Research 68:1589-1591.

International Food and Agribusiness Management Review

Ali

Volume 19, Issue 4, 2016

Cerdan, A.L.M. and A.J.C. Hernández. 2013. Size and performance in family managed firms: surviving first generation. Journal of the Iberoamerican Academy of Management 11: 13-34. Chang, C.L., H.K. Hsu and M. McAleer. 2013. Is small beautiful? Size effects of volatility spillovers for firm performance and exchange rates in tourism. North American Journal of Economics and Finance 26: 519-534. Chen, T-H. 2009. Performance measurement of an enterprise and business units with an application to a Taiwanese hotel chain. International Journal of Hospitality Management 28: 415-422. Chirwa, E.W. 2008. Effects of gender on the performance of micro and small enterprises in Malawi. Development Southern Africa 25: 347-362. Chow, C., K. Wing, M. Fung, and K. Yiu. 1997. Firm size and performance of manufacturing enterprises in P. R. China: the case of Shanghai’s manufacturing industries. Small Business Economics 9: 287-298. Coad, A. and J.P. Tamvada. 2012. Firm growth and barriers to growth among small firms in India. Small Business Economics 39: 383-400. Coleman, S. 2007. The role of human and financial capital in the profitability and growth of women-owned small firms. Journal of Small Business Management 45: 303-319. Cook, P. 2001. Finance and small and medium-sized enterprise in developing countries. Journal of Developmental Entrepreneurship 6: 17-40. Das, S. S. and A. Das. 2014. India shining? A two-wave study of business constraints upon micro and small manufacturing firms in India. International Small Business Journal 32: 180-203. De, P.K. and P. Nagaraj. 2014. Productivity and firm size in India. Small Business Economics 42: 891-907. Dethier, J-J., M. Hirn and S. Straub. 2011. Explaining enterprise performance in developing countries with business climate survey data. The World Bank Research Observer 26: 258-309. Forsman, H. and S. Temel, S. 2011. Innovation and business performance in small enterprises: an enterpriselevel analysis. International Journal of Innovation Management 15: 641-665. Hebous, S., M. Ruf and A.J. Weichenrieder. 2011. Decision of multinational firms : M & A versus Greenfield investments. National Tax Journal 64: 817-838. Hutchinson, J. and A. Xavier. 2006. Comparing the impact of credit constraints on the growth of SMEs in a transition country with an established market economy. Small Business Economics 27: 169-179. Kalkan, A., O. Erdil and Ö. Çetinkaya. 2011. The relationships between firm size, prospector strategy, architecture of information technology and firm performance. Procedia – Social and Behavioral Sciences 24: 854-869. Kotey, B. 2005. Are performance differences between family and non-family SMEs uniform across all firm sizes? International Journal of Entrepreneurial Behavior and Research 11: 394-421. Kumar, S. and P. Rao. 2016. Financing patterns of SMEs in India during 2006 to 2013 – an empirical analysis. Journal of Small Business and Entrepreneurship 28: 97-131. Kwong, C., D. Jones-Evans and P. Thompson. 2012. Differences in perceptions of access to finance between potential male and female entrepreneurs: evidence from the UK. International Journal of Entrepreneurial Behaviour and Research 18: 75-97. Laeven, L. and C. Woodruff. 2007. The quality of the legal system, firm ownership, and firm size. Review of Economics and Statistics 89: 601-614. Laing, D. and C.M. Weir. 1999. Governance structures, size and corporate performance in UK firms. Management Decision 37: 457-464. Lee, J. 2009. Does size matter in firm performance? evidence from US Public firms. International Journal of the Economics of Business 16: 189-203. Lun, Y.H.V. and M.A. Quaddus. 2011. Firm size and performance: a study on the use of electronic commerce by container transport operators in Hong Kong. Expert Systems with Applications 38: 7227-7234. Mbonyane, B. and W. Ladzani. 2011. Factors that hinder the growth of small businesses in South African townships. European Business Review 23: 550-560. Mead, D.C. and C. Liedholm. 1998. The dynamics of micro and small enterprises in developing countries. World Development 26: 61-74. Nichter, S. and L. Goldmark. 2009. Small firm growth in developing countries. World Development 37: 1453-1464. International Food and Agribusiness Management Review

Ali

Volume 19, Issue 4, 2016

Orlitzky, M. 2001. Does firm size confound the relationship between corporate social performance and firm financial performance? Journal of Business Ethics 33: 167-180. Orser, B. J., S. Hogarth-scott and A.L. Riding. 2000. Performance, firm size, and management problem solving. Journal of Small Business Management 38: 42-59. Palmon, O. and J.K. Wald. 2002. Are two heads better than one? The impact of changes in management structure on performance by firm size. Journal of Corporate Finance 8: 213-226. Park, Y., J. Shin and T. Kim. 2010. Firm size, age, industrial networking, and growth: a case of the Korean manufacturing industry. Small Business Economics 35: 153-168. Ponikvar, N., M. Tajnikar and K. Pusnik. 2009. Performance ratios for managerial decision-making in a growing firm. Journal of Business Economics and Management 10: 109-120. Ramukumba, T. 2014. Overcoming SMEs challenges through critical success factorsâ&#x20AC;Ż: a case of SMEs in the western Cape Province, South Africa. Economic and Business Review 16: 19-38. Robb, A.M. and J. Watson. 2012. Gender differences in firm performance: evidence from new ventures in the United States. Journal of Business Venturing 27: 544-558. Roomi, M.A., P. Harrison and J. Beaumont-Kerridge. 2009. Women-owned small and medium enterprises in England. Journal of Small Business and Enterprise Development 16: 270-288. Roxas, B., D. Chadee and R. Erwee. 2012. Effects of rule of law on firm performance in South Africa. European Business Review 24: 478-492. Shanmugam, K.R. and S.N. Bhaduri. 2002. Size, age and firm growth in the Indian manufacturing sector. Applied Economics Letters 9: 607-613. Shrivastava, B. and S. Chakraborty. 2015. Scramble for women directors as Indian companies miss deadline. March 31, 2015. Bloomberg. Available at: http://tinyurl.com/heygsye. Sridhar, K.S. and G. Wan. 2010. Firm location choice in cities: evidence from China, India, and Brazil. China Economic Review 21: 113-122. Tanabe, K. and C. Watanabe. 2005. Sources of small and medium enterprises excellent business performance in a service oriented economy. Journal of Services Research 5: 5-20. Vithessonthi, C. and Tongurai, J. 2015. The effect of firm size on the leverage-performance relationship during the financial crisis of 2007-2009. Journal of Multinational Financial Management 29: 1-29. Watson, J. 2002. Comparing the performance of male-and female-controlled businesses: relating outputs to inputs. Entrepreneurship Theory and Practice 26: 91-100. Watson, J. 2006. External funding and firm growth: Comparing female- and male-controlled SMEs. An International Journal of Entrepreneurial Finance 8: 33-49. Youn, H., N. Hua and S. Lee. 2015. Does size matter? Corporate social responsibility and firm performance in the restaurant industry. International Journal of Hospitality Management 51: 127-134.

International Food and Agribusiness Management Review

OPEN ACCESS International Food and Agribusiness Management Review Volume 19 Issue 4, 2016; DOI: 10.22434/IFAMR2015.0204 Received: 8 November 2015 / Accepted: 26 August 2016

The relevance of business practices in linking smallholders and large agro-businesses in Sub-Sahara Africa RESEARCH ARTICLE Linda Kleemann Researcher, Kiel Institute for the World Economy, Kiellinie 66, 24105 Kiel, Germany

Abstract Smallholders often have to certify according to international standards and produce under contract for large agro-businesses to access export markets. While mostly positive effects for the farmers have been found for contracts and certifications, often these effects do not persist because contracts fail and certifications are not renewed. We suggest that individual firm behavior is crucial for the long-term success of farmeragro-business relationships. In this article, we use data of 386 smallholders in the pineapple export sector in Ghana, analyze them quantitatively and enrich it by a detailed case study of a large-scale agro-business in Ghana. The results show that, in an environment with weak contract enforcement, certification is an agent of change in farmer-agro-business relations and that building trust and aligning expectations of farmers and firms largely determine success. We conclude that individual firm behavior matters more than taken into account in previous research. Our case study shows that three ‘R’ – reliability, reputation and respect – constitute the basis for contract relationships that benefit all. Keywords: contract farming, certification, smallholders, Ghana, firm management practices JEL code: O13, Q13, Q17, Q56 Corresponding author: linda.kleemann@ifw-kiel.de

Kleemann

Volume 19, Issue 4, 2016

1. Introduction and literature review Sub-Saharan African smallholders that target global food markets usually produce under contract for medium or large agro-businesses and certify according to international food standards. While certification with GlobalGAP is a market entry condition for conventional fresh products food, especially for horticultural products, organic certification is required for the high-value organic food market. Even though private, and hence voluntary, standards, retailers normally require that their suppliers adhere to one or more such standards for suppliers from developing countries (Henson et al., 2011; Swinnen et al., 2015). While only about 1% of land in Africa is organic certified as compared to 5% in Europe, organic certified production in Africa is almost exclusively destined for the export market (Willer et al., 2016). In Ghana 0.1% of total land is organic certified (15,563 hectares of agricultural land plus 35,695 hectares used for wild collection based on 2014 data; Willer et al., 2016). GlobaGAP certification operates in over 100 countries and is, in many developing countries, required by agribusiness exporters for fresh fruit and vegetables exports (ITC, 2013; Masood and BrĂźmmer, 2014).1 This certification requirement creates an entry barrier for suppliers and in particular for smallholders (Schuster and Maertens, 2013). However, those that master the barrier may gain financially (Bellmare, 2012; ITC, 2011; Maertens and Swinnen, 2009; Miyata et al., 2009; Subervie and Vagneron, 2013; Warning and Key, 2002), albeit this may not necessarily be enough to lift poor farmers out of poverty (Beuchelt and Zeller, 2011). In addition, failure rates are relatively high (Bellemare, 2012). Since certification is usually too costly for smallholders themselves, it is commonly supported by NGOs, bilateral aid organizations or the exporter to which the smallholder is contracted. Generally, the literature finds positive short-run income effects for this kind of contract farming (Barrett et al., 2012; Bellemare, 2012; Bolwig et al., 2009). Most papers study effects one or two years after certification (ITC, 2011), a period after which, when taking into account the full certification costs, these have not been recovered. High contract failure rates and frequent non-renewal of certifications in developing countries (Kersting and Wollni, 2012) may render net benefits negative. One reason for failures may be rooted in firm specific effects, which have hardly been studied so far (Lemeilleur, 2013). The literature has so far neglected the role of certification as a structural element driving contract outcomes, as well as the role of individual firm behavior in shaping long-term effects (Steen and Maijers, 2014). Considering the large initial investment required, the existence of positive net effects crucially depend on survival rates, i.e. the length of a specific contract farming relationship or certification period. In this article we tackle the research gap by showing that firm management practices are crucial to contract survival. Data of 386 either GlobalGAP or organic certified smallholders in the pineapple export sector in Ghana is analyzed quantitatively and enriched by a qualitative analysis of a case study of a large-scale agro-business called Blue Skies. Certification imposes sunk costs on both contract partners, which they can only recover after several years of working together. Hence, a contractual relationship that enables certification only leads to material benefits if it lasts. This is the main difference to the situation without certification, making it a driver of change in farmer-agro-business relationships. Self-reported changes include a more intense relation and an improved overall relationship following certification, an important basis for long-term relationships. Longterm relationships allow for renewal of the certification. Hence, with certification a longer-term contractual relationship is a necessary precondition for mutual benefit, due to the high initial certification cost. Nevertheless, many contracts still fail after a short period of time. These failures can only partly be explained by market changes and part-externalization of the certification costs. Firm management practices play a key role. In an environment with weak contract enforcement, trust is particularly important, and this is shaped by contract management.

1 For Ghana, the GlobalGAP website lists 67 certified pineapple suppliers. However, small-scale producers are certified under a group option. Hence

this figure does not represent the number of small-scale farmers actually certified (GlobalGAP, 2016).

International Food and Agribusiness Management Review

Kleemann

Volume 19, Issue 4, 2016

In the next step we identify firm management practices as important success factors for the latter. We use subjective self-statements to gain deeper information about the farmers’ perceptions and motivations as these shape expectations. The match or mismatch between farmer motivation to join a certification or contract arrangement and the perception of the outcome of this process defines whether the farmer will be satisfied or disappointed with the outcome (Hidayat et al., 2015). Non-alignment may explain many failures of contract schemes. Contracts last when both sides stick to their promises and respect each other. Some agro-businesses manage this much better than others. Our case study shows that three ‘R’ – reliability, reputation and respect – constitute the basis for contract relationships that benefit all. These successful firms accomplish to establish their corporate culture among their contract farmers and buffer them against international market volatility, contrarily to what Suzuki et al. (2011) suggests in common practice. Standards linked with contracts are short-run agents of change; individual firms determine whether they translate into long-run benefits. The rest of the article is organized as follows. Section 2 describes the data that is used in this paper. Section 3 presents the analysis, Section 4 concludes.

2. Data The data used in this paper are a farmer survey in Ghana and a detailed case study of Blue Skies, a largescale agro-business in Ghana. The data sources are linked through farmers identified in the survey and in the case study. We are hence able to compare farmers producing for Blue Skies with farmers in the same sector but producing for another firm. The farmer survey was conducted from January to March 2010 in six different districts (Ajumako Enyan Esiam, Akuapem South, Ewutu-Efutu-Senya, Ga, Kwahu South and Mfantseman) of the Central, Eastern and Greater Accra regions in a radius of about 100 km north and west of Accra. Stratified random sampling in three stages was used. First, districts with significant amounts of commercial smallholder pineapple production were selected, using information from Sea Freight Pineapple Exporters of Ghana. Next, lists of all pineapple farmer groups in the selected districts that were GlobalGAP or organic certified were obtained.2 Finally, a percentage of farmers in each group were selected randomly from the lists. The sample is representative of the selected districts. Identified farmers answered a detailed questionnaire that bordered on the management of the pineapple farm, inputs for the production, harvesting and marketing of the pineapples, the certification process, and relations with exporters and processors that were exclusively medium- and large-scale agrobusinesses. Respondents were also made to provide information on household characteristics, social capital and land disposition, as well as non-income wealth indicators and perceptions of different statements about environmental values, organic farming techniques and the use of fertilizers and pesticides. The dataset includes 386 farmers from 75 villages with either GlobalGAP or organic certification for their pineapple farms. In total, 185 organic farmers and 201 conventional (GlobalGAP) certified farmers were interviewed. Organic farmers sold part of their produce as organic certified to exporters or processors and part of it on the local market, without any reference to the certification. Conventional farmers sold their produce as GlobalGAP certified to exporters or processors and on the local market, without reference to GlobalGAP certification. In principle, organic certified farmers could sell their produce as organic certified (which has the highest price) as first preference, as conventional export produce as second preference, or on the local market. It is not possible for conventional farmers to sell on the export organic market. Organic certification refers to the European standards according to EU regulation (EC) 834/2007 and (EC) 889/2008, which prohibits in particular the use of mineral fertilizers and inorganic pesticides, herbicides and fungicides. All conventional farmers are GlobalGAP certified in our sample. GlobalGAP refers to a catalogue of criteria based on so called good agricultural practices. A particular focus is on food and worker safety, as well as chemical residues.3 2 Smallholders 3 The

are certified in groups under the so-called option 2 certification. full list of criteria for fruit can be found here: http://tinyurl.com/zj9zasl.

International Food and Agribusiness Management Review

Kleemann

Volume 19, Issue 4, 2016

Table 1 presents an overview of descriptive statistics for Blue Skies farmers against all other farmers in the sample. Almost all farmers are male, on average 45 years old and live in a five to six person household. Significant differences between Blue Skies farmers and other farmers exist for farming experience, pineapple land and varieties planted. Farms are relatively large within the group of what is considered the group of small farmers in Ghana. This is a result of two particularities of the sector. First, many agro-businesses require minimum farm sizes as a precondition for contract farming. Second, pineapple is usually planted in a rotation system with about half of the land at one point in time fallow. Three pineapple varieties are planted, MD2, Smooth Cayenne and Sugarloaf. Most farmers plant only one of these, with some exceptions having two different varieties on their farm. Economic and agro-business specific variables are presented further in this paper. The case study of Blue Skies, a large-scale agro-business in Ghana, was based on interviews with its suppliers, i.e. farmers, employees, management and communities in which Blue Skies was active in 2013.4 For the purpose of this paper, only the farmer and management interviews will be used. Blue Skies produces fresh cut fruit and fruit salads for export mainly to Europe and freshly squeezed juice for the local market. It buys both organic and GlobalGAP certified fruit from local farmers, mostly in a contract farming arrangement, but occasionally also on the spot market. Set up in Ghana in 1998, it was in 2013 the second biggest private sector employer in Ghana with around 2,000 employees, depending on the season. It has grown into a group of factories processing fresh fruit locally with additional smaller sites in Egypt, South Africa, Senegal, Brazil 4 The data gathered for the case study was collected by the author for a report commissioned by Waitrose, one of the buyers of Blue Skiesâ&#x20AC;&#x2122; products.

The information is used with permission from Blue Skies and Waitrose.

Table 1. Descriptive statistics of selected variables. Variable

Definition

Mean Blue Mean of all Skies Farmers other farmers n=71 n=282

t-stat.1

GENDER AGE HHSIZE ADULT EDUC FSIZE OWNLAND PINLAND CREDIT

gender of farmer: 1 if male, 0 if female age of farmer (years) household size (persons living in household) fraction of adults (older than 15) in household maximal educational level in household (years) farm size (acre) share of land owned pineapple land (acre) access to credit during the last five years:1 if yes, 0 otherwise bank account with more than 200 GHS2: 1 if yes, 0 otherwise number of durable goods owned years of experience in pineapple farming importance of preserving the environment 1= very important; 4= not important distance to the closest local market (hours) variety MD2 (1 if yes, 0 otherwise) variety Smooth Cayenne (1 if yes, 0 otherwise) number of certified years organic certified (1 if yes, 0 otherwise)

0.930 45.37 5.213 0.601 9.572 14.72 0.293 4.852 0.304

0.935 44.71 5.643 0.674 9.679 14.01 0.287 3.112 0.278

0.200 1.071 1.193 1.137 0.974 0.651

0.412

0.422

0.123

6.183 14.79 1.269

6.152 9.049 1.513

0.933 3.149*** 1.564

0.867 0.016 0.420 3.987 0.541

0.714 0.398 0.416 3.456 0.483

1.857* 5.872*** 0.051 1.011 1.312

BANK WEALTH EXPER ENV DIST MD2 SC CERTYEAR ORGANIC 1

2.091** 1.265

Significance levels: * = 10%; ** = 5%; *** = 1%.

2 A conversion factor of 1 Ghana Cedi (GHS) = 0.46 Euros (calculated on the basis of the exchange rate on January 12, 2010) was used.

International Food and Agribusiness Management Review

Kleemann

Volume 19, Issue 4, 2016

and UK. 55% of the total production value is created in Ghana, followed by UK with 16% and Egypt with 15%. It employs over 2,500 people at all sites together and had a turnover of â&#x201A;Ź47.6m and profit of â&#x201A;Ź1.1m in 2012. At the time of the case study, Blue Skies Ghana had 70 supplying farmers, of which 59 were contracted suppliers. The rest are larger independent farms, including outside Ghana (Figure 1). Contracts with farmers are renewable yearly. They specify the certifications, crop variety, quality, brix levels and terms of payment. Prices are fixed in â&#x201A;¤GBP and renegotiated yearly. The overall acreage grown for Blue Skies is 1,928. A combination of a standardized questionnaire and open qualitative interviews using the most significant change technique were used. Ten farmers were interviewed, representing the major crops grown for Blue Skies by smallholders: pineapple, mango, papaya and coconut. They were randomly drawn from the contracted supplier list, after two selection criteria were fulfilled: to cover all main crops grown for Blue Skies and to include both farmers that have been supplying to Blue Skies for a long time, and farmers that started recently. 4 pineapple farmers, 3 papaya farmers, 3 mango farmers and 1 coconut farmer were interviewed at their farms, where one grows both mango and papaya and the coconut farm is a sharecropping system with many families working and living on the farm. One person from the Blue Skies agronomy team always introduced us. Hence farmer interviews were not entirely conducted confidentially. This was the only possible way to be well received. Farmers were nevertheless very opinionated and sometimes even asked the agronomy team to listen and witness their complaints or requests. All interviews were made on the farms and with the farmer himself, all of whom were male. Each interview took between 30 minutes to one hour. In addition, interviews and informal discussions were led with the management of Blue Skies throughout the study period. The management of Blue Skies provided us with all the information requested on management practices, farmer statistics and policies, extension, certification, etc. and was available for clarifications and feedback. The two samples are linked in the following way: farmers who produce for Blue Skies were identified in the farmer survey.

3. Analysis We now combine and analyze the two datasets in two steps. We first study the role of certification in shaping expectations and business relationships and then look at the role of individual business practices in shaping longer-term results. Certification as an agent of change in farmer-agro-business relations The farmer survey has been quantitatively analyzed in particular with respect to return on investment in certification and agricultural practices in Kleemann et al. (2014) and Kleemann and Abdulai (2013), showing that different types of certification, namely organic and GlobalGAP can have very different returns on

Smooth Cayenne MD2 Sugarloaf Mango Coconut Papaya Independent suppliers 0

Supplying farmers

Figure 1. Blues Skies approved suppliers in 2013. International Food and Agribusiness Management Review

Kleemann

Volume 19, Issue 4, 2016

investment and foster different types of agricultural practices. Here, we focus on the subjective statements that farmers were asked to give during the survey. Since the share of organic and GlobalGAP certified farmers among the sample firm and in the rest of the dataset are both around 50%, we abstract from differences between different certifications in this paper. In our sample of farmers, self-reported changes of the certification process in general include a more intense relationship (farmer and agro-business meet or talk more often to each other), an improved overall relationship following certification and longer contract durations with pre-specified volumes. Specifically, 63% of farmers report an improved overall relationship, against 36% reporting no change and 2% a worse relationship and the results for the intensity of the relationship are 76, 24 and 1%, respectively. Hence, certification alters not only prices and costs; it is also a driver of change in farmer-agro-business relationships and contract specification. One likely reason is the upfront commitment and investment from both sides that is required. The certification process can take several years (e.g. three years for organic) and hence there is a considerable time lag between the decision and the first market transaction as certified product. Longer-term contracts with pre-specified volumes are more frequent among certified farmers. Longer relationships (a larger number of years selling to the same buyer) allow for renewal of the certification. This is in the interest of both sides, due to the high initial certification cost. Certification, if managed as part of a contract relationship, could hence be an element of structure that shapes strategy. Going more into detail, the information flow and roles of different actors during the certification process are strongly correlated with farmer satisfaction after the process is completed, which in turn is likely to be correlated with contract survival. First, we asked all farmers directly what their motivations for one or the other certification where5 and how they first got to know about the possibility of organic and GlobalGAP certification respectively. Organic farmers got information usually from buyers (i.e. large-scale agro-businesses) or other farmers whereas half of all GlobalGAP farmers were informed through NGOs or donors (Table 2). To our knowledge, GlobalGAP certification was at that time intensively supported by US and German development aid and this picture may be the result. For both groups government extension services were hardly ever (in 9% of all cases) relevant information providers. Buyers and other farmers stress ‘hard’ information (on prices, markets and yield), on certification, whereas NGOs and donors put the focus on ‘soft’ information such as environmental hazards and safe handling. This is the outcome from feedback discussions with buyers and donors and is also reflected in the certification training material provided by those groups. Even more, it is also mirrored in the personal motivation that farmers stated for aiming at a particular certification. When asked in an open question for their motivation for certification, those informed by NGOs or donors stated far more often than those informed by agro-businesses that environmental concerns, health or food safety as determining factors whereas those informed by agro5 By stressing differences between statements and actions, many researchers affirm that believing in what people say can be misleading (Manski, 2004). Consequently, one rarely sees subjective data in empirical papers. We want to break with this tradition and compare our quantitative results with qualitative information about farmers’ own statements on the subject. We are aware of the measurement errors that come with individual differences in interpretation of questions and expressions, and can thus not make accurate statements using this methodology.

Table 2. From whom farmers first learned about the possibility to get organic/GlobalGAP certification in percentages. Information provider

Organic (%)

GlobalGAP (%)

Export agro-businesses Government/extension NGO or donor Other farmers Relatives Other

42 9 12 30 7 0

14 9 50 24 1 1

International Food and Agribusiness Management Review

Kleemann

Volume 19, Issue 4, 2016

businesses considered yields, prices and contracts most important (Table 3). Blue Skies mostly responded that they were informed by agro-business (94%) and their responses mirror those informed by agro-businesses. How might this be relevant for the rest of the certification process and even more for the success of the farmer-agro-business contract? According to our data, the quality of the relationship between farmer and agro-business, as subjectively perceived by the farmer, is significantly better in those cases where the agro-business provided the initial information (test statistic of two sided t-test is 2.77). Going further in the procedure, when the agro-business provided the initial information, it likely also organized the certification process and even paid for it (correlation of 0.6). In addition, in those cases where the exporter paid for the certification, farmers received on average double the amount of training than in all other cases. And the quality of the relationship is also perceived as significantly better when the agro-business organized the certification process (test statistic of two sided t-test is 6.11). For our case study specifically, the data shows that usually when Blue Skies is the buyer, they also organized and paid for the certification. This result points towards the importance of integrating and engaging both partners early in the certification process in order to align expectations and build trust. This was confirmed in the discussions with both sides, with the key factors being building mutual trust. Disappointments in terms of wrong expectations or unreliability of the other party were mentioned as the main reasons for failures of contract schemes. This means that the ability of both certification and contract to deliver on the expectations it created will determine its success. As proclaimed in management theory, aligning expectations and building trust is important for longer-term success. Our case study of a successful agro-business below shows that Blue Skies puts particular emphasis in its farmer-relations management, on frequent and transparent communication and reliability on both sides. Analysis of the firm factor The above analyzed positive drive in contractual relationships only transforms into material benefits, if the sunk cost of certification can be recovered, which usually takes several years. Certification demands an upfront investment from both parties, a sunk cost in case the contract fails. This cost is usually higher than the additional benefit generated within the first contract year. Hence, any business should only invest in this contract, if the contract duration is long enough to recover the initial cost. This is the additional incentive to strengthen the relationship. While the literature finds modest short run income and welfare effects which disappear quickly after the end of the contract, stronger beneficial effects of contract farming should manifest themselves primarily over the medium and long run in higher regular incomes and farm or asset growth. If long-term contract relationships Table 3. Stated motivation to become certified in percentages.1 Motivation for certification

Blue Skies farmers (%) n=71

Others informed by agro-business (%) n=162

Others informed by NGO or donor (%) n=153

Better yields Better prices Health or food safety reasons Environmental concerns Better contracts with exporters Easier to sell Cultural reasons/tradition Customer demands Other reasons

1 20 30 20 15 17 0 9 0

11 23 21 5 15 7 1 16 1

10 14 26 34 2 5 0 11 1

Multiple selection of responses was possible (up to three). International Food and Agribusiness Management Review

Kleemann

Volume 19, Issue 4, 2016

allow for positive partner-specific investments on both sides such as on farm pack houses, planting of specific varieties, etc. Contracts last when both sides stick to each other. Some agro-businesses seem to manage these farmer relations much better than others, among them our case study firm Blue Skies. Table 4 shows that Blue Skies farmers are better off in several respects. They are compared with all other farmers in the sample. Bilateral t-tests show that Blue Skies farmers have a higher return on investment and a higher profit that is determined by higher revenues, not by lower costs. While we do not have a direct measure of the extent of use of state of the art farming technologies and the overall the level of use of fertilizer, mulch, and other productivity enhancing agricultural practices, we see that Blue Skies farmers use more good agricultural practices and organic farming techniques and this even though they did not receive more training. Training and support institutions named by the farmers are mostly government (mainly ministry of agriculture) and international donors. Training and support by firms mentioned was done by Blue Skies. Other exporters/ processors are mentioned in less than 1% of the cases by those farmers selling to other agro-businesses. However, it appears that Blue Skies contract farmers receive slightly less training by other organisations (the difference is significant at 10%). This could be supply or demand driven. Bellemare (2012) and others argue that contract farming is a driver of farm modernization. However, we see here that there are huge differences between contract firms (all farmers in the sample produce under contract). Table 4. Blue Skies farmers in comparison with other contract farmers.1 Variable

Definition

Blue Skies All other farmers farmers n=71 n=282

ROI PRODCOS_KG REV_KG PROFIT_KG TRAIN

return on investment in pineapple farming (one year) 3.13 production cost GHS3 per kg fruit 0.11 revenue GHS per kg fruit sold 0.26 profit GHS per kg fruit sold 0.15 training received in last 5 years from exporters, NGOs, 17.96 donors or ministry of agriculture GAPPRACT number of good agricultural practices and organic farming 4.13 practices used YEARS_BUYER number of years already selling to the same buyer 2.42 REL_BUYER quality of relationship to buyer on a scale from 1 (very 1.31 good) to 4 (very bad) Details of the quality of the relationship between buyer and seller: PICKUP_BUYER 1 if satisfied with delivery/pickup arrangements, 0 0.94 otherwise VOL_BUYER 1 if satisfied with volumes bought by buyer, 0 otherwise 0.76 BUY_GUARANT 1 if guaranteed volumes bought, 0 otherwise 0.83 TIME_PAY time lag from pickup to payment (1 = same day; â&#x20AC;Ś; 5 = 3 2.91 months or more) Intensity of the relationship between buyer and seller: MEET_BUYER frequency of meetings between buyer and seller (times per 10.65 year) PHONE_BUYER 1 if phone number of buyer known, 0 otherwise 0.68 1

t-stat.2

1.90 0.12 0.17 0.06 14.82

3.56*** 0.63 9.17*** 5.21*** 1.03

1.89

11.91***

0.97 2.34

10.62*** 10.66

0.36

8.24***

0.15 0.26 3.14

7.00*** 10.08*** 2.25**

5.96

5.38***

0.20

5.09***

Differences between organic and GlobalGAP certified farmers were the focus of the paper Kleemann et al. (2014). Since the share of organic farmers amongst Blue Skies farmers and among the rest of the sample is similar (54% and 48% respectively), this difference between the two certifications in particular in terms of production costs and prices can be neglected here. 2 Significance levels: ** = 5%; *** = 1%. 3 1 Ghana Cedi (GHS) = 0.46 Euros (calculated on the basis of the exchange rate on January 12, 2010).

International Food and Agribusiness Management Review

Kleemann

Volume 19, Issue 4, 2016

But Blue Skies farmers are also better off in another respect, which was identified as a crucial success factor in the previous section. Farmers producing for Blue Skies state to have a significantly better and more intensive relationship with their buyer, than all other farmers in the sample. This relationship also lasts for longer already, on average more than two years instead of less than one year.6 Intensity is measured by whether or not they have the phone number of the buyer, the frequency of meetings, and how much they know about the further use of their pineapple (destination/processing). Whether the better relationship results in better economic outcomes or vice versa remains unclear. These correlations might not be causal. Therefore we have tried to verify these results through qualitative interviews. In addition, selection for the ‘better’ farmers may be an issue that affects some of these results. We tested some standard measures such as farm size, production costs, and experience and while there are no significant differences between Blue Skies farmers in these respects, there may be other factors that we did not capture such as fruit quality and farmer reliability. The selection process is one of the main aspects we focused on the in the qualitative interviews. All agro-business firms in our sample have similar selection mechanisms for farmers in particular concerning minimum farm size and/or level of organization in groups. The common target of this selection mechanism is to find those farmers that produce good quality in a reliable way at an acceptable distance to the firm. Because firms cluster in a small area, they target the same regions for supplier farmers. Nevertheless, because Blue Skies has a good reputation they might have the first choice in terms of supplier farmers. We asked the farmers directly in order to find out whether the selection process differs among firms and what they considered as the most important factors leading to their satisfaction with the contract relationship as well as economic benefits from it. We call the latter results channels below. As most farmers had experiences with other agro-business firms, they were able to compare. While they stated that the selection from the firm side does not differ significantly between firms7, they had very clear answers concerning results channels (Table 5). Farmers considered the secure and reliable long-term market and payment stream that Blue Skies is providing as the most important results channel, in particular in comparison with other buyers who were criticized for their unreliability. Blue Skies is respected for its corporate culture of respect, social equality and openness up to the extent that farmers imitate it themselves. Several farmers mentioned their admiration for Blue Skies’ management, especially related to mastering past market challenges, such as failures in export due to the ash cloud in island in 2010. Blue Skies is also respected for the quality of its advice and training to the farmers, which is considered, compared to others, much more targeted to their needs and takes up their suggestions and ideas. In addition, Blue Skies, in partnership with two of its buyers Waitrose and Albert Heijn, supports community projects through a foundation. Projects are proposed by the farmers and owned by the communities. The Foundation manager at Blue Skies supports and overviews the implementation and visits each project regularly. The interviews showed, that the foundation is an important add-on because it gives Blue Skies and its farmers a good standing in the communities. They are judged as important by the farmers, but nevertheless second to a stable market (Table 5). The transparent management of problems and difficulties by Blue Skies was particularly mentioned by several farmers, including for unsuccessful examples. In addition, the quality rather than the quantity of the training was highlighted. We provide a characteristic unsuccessful example from the interviews. The typical organic Sugarloaf pineapple farmer has been growing Sugarloaf for many years and has been with Blue Skies from the beginning of their operations. He emerged from a poor family background. His farm is comparably small, but has grown considerably over time together with Blue Skies. Pineapple production is his family’s main income source. He sells about 50% of his fruit to Blue Skies. The rest of his harvest is sold at a lower price on the local market. This is the only alternative market for him. The additional income from selling to Blue Skies not only helped him to increase his farm size, but also to send his children to 6 The

sample was random and representative at the time of survey, which implies that there should be no differences in average contract duration if there is no ‘firm factor’. 7 However, Blue Skies, due to their long presence, may be better informed about the characteristics of different farmers, and therefore able to select more effectively. Without panel data, it is not possible to assert or reject this supposition.

International Food and Agribusiness Management Review

Kleemann

Volume 19, Issue 4, 2016

Table 5. Most important results channels from the perspective of the farmers in descending order. Reliability and consistency

Volume Corporate culture Training

Credit

Community projects

• ‘Blue Skies is the most reliable buyer and always pay everything and on time. Prices are fair and we are told about quantities in advance. And there are additional incentives that other buyers do not provide.’ • ‘Other exporters were not reliable.’ • ‘Sometimes we expect to sell more but we understand it is because of the orders that Blue Skies receives from their customers. We stay with Blue Skies because it is reliable and we can constantly supply them. There is no other consistent buyer in Ghana.’ • ‘I would prefer to sell to Blue Skies even if I get a higher price elsewhere.’ • ‘We sell almost all our fruit to Blue Skies. And we would sell more. We want to expand the farm and improve housing for workers if we are able to sell more.’ • ‘We feel that we are all part of the Blue Skies family. We can openly discuss our problems and complaints with the agronomy team. Aspiring farmers are built up to succeed by Blue Skies. We admire how they manage, especially in difficult times.’ • ‘The constant training from Blue Skies is very beneficial. They visit us every 2-6 weeks for audits and trainings that cover amongst others certification, cropping, farm management. We also ask for advice with current prevalent problems. They take our concerns seriously.’ • ‘We would like to receive a loan for the expansion of the farm. We cannot get it from Blue Skies and the banks are not helping either. They have very high interest rates and demand huge collateral. But on an individual basis, needs are considered. We can get soft loans (without interest) as advance payment. We know that we can count on Blue Skies that they will do their best.’ • ‘I was very involved in getting the Foundation project in my community and I am very proud of it. I am now also in the management committee.’ • ‘We are applying to the Foundation to get a Junior High School to our community. But more important, is more demand for fruit.’

better schools and to invest in a taxi as additional off-farm business. However, in the past few years, the demand for Sugarloaf from Blue Skies has decreased and become unstable. Many farmers, especially the smaller ones, had to leave Blue Skies and are now selling exclusively on the local market or switched to staple crops. To try to counter this trend, Blue Skies is actively promoting the Sugarloaf variety among its customers as well as trying to find new customers for Sugarloaf. They are supporting farmers in testing new farming techniques, e.g. using plastic mulch, while being careful not to induce high expectations that they cannot meet. While the farmer is not happy with the low demand, he understands the demand situation and respects Blue Skies for its efforts. We conclude that the satisfaction of Blue Skies farmers and their economic success is at least in a considerable part due to the way that Blue Skies treats its farmers and not due to selection effects. But does this also benefit Blue Skies, i.e. is it a win-win situation? Without being able to establish causality, we observe that Blue Skies has had its operations in Ghana since 1998, over time considerably increasing in size. During this period, many others have failed (e.g. Coastal Groves, Kingdom Fruit Juice, Nsawam Cannery and Athena) or remained much smaller (e.g. Peelco and WAD). In the next step, we try to understand the corporate causes behind the big difference between Blue Skies farmers and other contracted farmers that we found in the farmer survey and in the qualitative interviews. We benchmarked Blue Skies with other similar firms. A list of firms used for the benchmarking can be found in Supplementary Table S1. In particular, we looked at three points: smallholder orientation, prices and corporate social responsibility. While Blue Skies does not differ significantly from other agro-businesses in

International Food and Agribusiness Management Review

Kleemann

Volume 19, Issue 4, 2016

terms of buying practices and corporate social responsibility, it differs in terms of soft factors.8 Reliability and consistency, corporate culture and training and were identified as most important results channels by the farmers (Table 5). We hence contoured the main factors within Blue Skies by reviewing their policies and observing their actual behavior in day-to-day business. The Joint Effort Enterprise is the Blue Skies model for a sustainable business. It is built upon three strands, as can be seen in Figure 2. First, Blue Skies invests strongly into building up a long-term supplier base through its, in comparison, large and well equipped agronomy department. 15 people take care of the permanent suppliers, dealing with training, certifications, audits, quality assurance, crop planning, etc. Farmers also receive individual assistance and access to subsidized inputs such as compost. Extension workers know ‘their’ farmers personally and treat them on an equal basis. They encourage farmers to think in an entrepreneurial way taking their thoughts and ideas seriously. At the same time, Blue Skies invests heavily in high quality training of its farmers and staff. This way they gain confidence, skills and experience, while Blue Skies gains a good reputation as buyer. Second is the active practice of the strong corporate culture and undisputable values implied by the business model summarized in Figure 2. This is much harder to benchmark against other similar firms. However, it was observed that corporate culture is very strong and lived in day-to-day business practice with the management acting as role models. In addition, farmers, which had experiences with other firms, frequently explained the difference between Blue Skies and other firms in terms of corporate culture. The Blue Skies culture is based on mixing people from diverse backgrounds minimizing hierarchies and visible distinctions between people. On the social side everyone is treated equally and with respect. Management is based on trust and peer pressure, which is unusual in Ghana, where it is usually based on supervision. This culture creates a strong identification with Blue Skies among farmers. It also implies that those who do not fit in leave voluntarily. The third success factor is reliability. Blue Skies behaves in a protective way towards its farmers and surrounding communities, trying to buffer them against market volatilities, while transparently communicating own challenges. This is in contrast to the findings by Suzuki et al. (2011), which explain the opposite as common practice. This combination of protection and open communication creates a trustworthy and resilient relationship between suppliers and Blue Skies. In summary, our case study shows that three ‘R’ – reliability, reputation and respect – constitute the basis for contract relationships that benefit 8 In addition, not the focus of this paper, but highly relevant for the overall local impact of Blue Skies is its principle of value adding at source. This principle translates into local employment opportunities in Ghana and up to 70% of the production value stays in the country, compared to about 15% when processing takes place outside Ghana.

Employing people from different backgrounds and cultures because we believe that we will generate better ideas if we have a diverse range of skills, experience and perspectives.

Respecting people equally

Operating profitably and efficiently

because we believe that if we respect each other for who we are, then we will feel happier about our work and proud to do a good job. This is our culture.

because we know that we cannot continue to produce the best fruit products in the world unless we generate the funds that will enable us to survive and grow.

Figure 2. The Blue Skies business model at a glance (adapted from Blue Skies: http://blueskies.com/about). International Food and Agribusiness Management Review

Kleemann

Volume 19, Issue 4, 2016

both sides. Certification creates a mutual dependency that did not exist before. Successful certification for smallholders in turn depends in many cases on a successful contractual relationship with an agribusiness. This research shows that this strengthens the relationship only if the R exist.

4. Conclusions With increasing relevance of certification standards such as GlobalGAP, organic and Fairtrade and associated contract relationships between exporters and smallholders, many researchers have analyzed the income and welfare effects of such arrangements. But they have so far neglected the role of certification as a structural element driving an increase in contracts, which result in a win-win situation only if they can be maintained over several years. The length of the contractual relationship is in turn to a large extent determined by individual firm behavior. This paper shows that certification is an agent of change in farmer-agro-business relations. Because it requires a large upfront investment in terms of certification cost, training and changes in farm management and involves a considerable time lag between decision to invest and first benefits, aligning expectations of farmers and buyers (i.e. agro-businesses) and building trust between the partners is crucial for the success of the whole process. Some agro-businesses are more successful than others in managing the required kind of trustful and strong relationship with their contracted smallholders. Individual firm behavior matters more than taken into account in previous research both before certification (expectations) and after (income effects/personal satisfaction). Our case study of Blue Skies shows that three ‘R’ – reliability, reputation and respect – constitute the basis for win-win contract relationships. Successful firms manage to establish a joint corporate culture among their staff and contract farmers and buffer risks of international market volatility while demanding high quality and reliability. Given that beneficial effects of smallholder agro-business relationships primarily show up in the longer run in the form of recovered investment and higher regular incomes, individual firm management is crucial. Standards linked with contracts are shortrun agents of change, the individual firms determine whether they translate into long-run success. Future research would benefit from calculating survival rates of agro-business-smallholder contracts and link them to economic benefits. For supporters of certification processes, be it NGOs, donors, or agro-business firms, this means that more importance should be placed on longevity of contracts and to ‘soft’ factors such as trust building and forming a joint culture in addition to ‘hard’ facts such as market opportunities and entry barriers.

Supplementary material Supplementary material can be found online at https://doi.org/10.22434/IFAMR2015.0204. Table S1. Other companies analyzed for benchmarking.

References Barrett, C., M. Bachke, M. Bellemare, H. Michelson, S. Narayanan and T. Walker. 2012. Smallholder participation in contract farming: comparative evidence from five countries. World Development 40: 715-730. Bellemare, M.F. 2012. As you sow, so shall you reap: the welfare impacts of contract farming. World Development 40: 1418-1434. Beuchelt, T. and M. Zeller 2011. Profits and poverty: certification’s troubled link for Nicaragua’s organic and Fairtrade coffee producers. Ecological Economics 70: 1316-1324. Bolwig, S., P. Gibbon and S. Jones. 2009. The economics of smallholder organic contract farming in tropical Africa. World Development 37: 1094-1104. GlobalGAP. 2016. GlobalGAP Database. Available at: http://tinyurl.com/kpada2c. Henson, S., O. Masakure and J. Crandfield. 2011. Do fresh produce exporters in Sub-Saharan Africa benefit from GlobalGAP certification? World Development 39: 375-386.

International Food and Agribusiness Management Review

Kleemann

Volume 19, Issue 4, 2016

Hidayat, N.K., P. Glasbergen and A. Offermans. 2015. Sustainability certification and palm oil smallholders’ livelihood: a comparison between scheme smallholders and independent smallholders in Indonesia. International Food and Agribusiness Management Review 18: 25-48. International Trade Centre (ITC). 2011. The impacts of private standards on producers in developing countries. International Trade Centre literature review series on the impacts of private standards, part II, Geneva, Switzerland. International Trade Centre (ITC). 2013. Key features of the sustainability standard. Available at: http:// tinyurl.com/j7fah5z. Kersting, S. and M. Wollni. 2012. New institutional arrangements and standard adoption: evidence from small-scale fruit and vegetable farmers in Thailand. Food Policy 37: 452-462. Kleemann, L. and A. Abdulai. 2013. Organic certification, agro-ecological practices and return on investment: evidence from pineapple producers in Ghana. Ecological Economics 93: 330-341. Kleemann, L., A. Abdulai and M. Buss. 2014. Is organic farming worth its investment? The adoption and impact of certified pineapple farming in Ghana. World Development 64: 79-92. Lemeilleur, S. 2013. Smallholder Compliance with private standard certification: the case of GlobalGAP adoption by mango producers in Peru. International Food and Agribusiness Management Review 16: 159-180. Maertens, M. and J.F.M. Swinnen. 2009. Trade, standards, and poverty: evidence from Senegal. World Development 37: 161-178. Manski, C.F. 2004. Measuring Expectations. Econometrica 72: 1329-1376. Masood, A. and B. Brümmer. 2014. Determinants of worldwide diffusion of GlobalGAP certification. GlobalFood Discussion Papers No. 48, Georg-August-University, Göttingen, Germany. Miyata, S., M. Nicholas and H. Dinghuan. 2009. Impact of contract farming on income: linking small farmers, packers, and supermarkets in China. World Development 37: 1781-1790. Schuster, M. and M. Maertens. 2013. Do private standards create exclusive supply chains? New evidence from the Peruvian asparagus export sector. Bioeconomics working paper series working paper 2013/1. Department of Earth and Environmental Sciences, Division of Bioeconmics, University of Leuven, Leuven, Belgium. Steen, M. and W. Maijers. 2014. Key success factors for Ethiopian agribusiness development. International Food and Agribusiness Management Review 17: 83-88. Subervie, J. and I. Vagneron. 2013. A drop of water in the Indian ocean? The impact of GlobalGap certification on lychee farmers in Madagascar. World Development 50: 57-73. Suzuki, A., S.J. Lovell and R.J. Sexton. 2011. Partial Vertical integration, risk shifting, and product rejection in the high value export supply chain: the Ghana pineapple sector. World Development 39: 1611-1623. Swinnen, J., K. Deconinck, T. Vandemoortele and A. Vandeplas. 2015. Quality standards, value chains, and international development. Cambridge University Press, New York, USA. Warning, M. and N. Key. 2002. The social performance and distributional consequences of contract farming: an equilibrium analysis of the arachide de bouche program in Senegal. World Development 30: 255-263. Willer, H. and J. Lernoud (eds.). 2016. The world of organic agriculture – statistics and emerging trends. 2016. Research Institute of Organic Agriculture, Frick, Switzerland.

International Food and Agribusiness Management Review

OPEN ACCESS International Food and Agribusiness Management Review Volume 19 Issue 4, 2016; DOI: 10.22434/IFAMR2015.0201 Received: 4 November 2015 / Accepted: 22 August 2016

The influence of value chain integration on performance: an empirical study of the malt barley value chain in Ethiopia RESEARCH ARTICLE Mulugeta D. Watabaji a, Adrienn Molnarb, Manoj K. Dorac, and Xavier Gellynckd aStaff

member, College of Business and Economics, Haramaya University, P.O. Box 138, Dire Dawa, Ethiopia; PhD student in the Division of Agri-food Marketing and Chain Management, Department of Agricultural Economics, Faculty Bioscience Engineering of Ghent University, Coupure Links 653, 9000 Ghent, Belgium bPost-doctoral

researcher, Department of Agricultural Economics, Faculty of Bioscience Engineering of Ghent University, Coupure Links 653, 9000 Ghent, Belgium; staff member Hungarian Academy of Sciences, Institute of Economics cPost-doctoral

researcher, Division of Agri-food Marketing and Chain Management, Department of Agricultural Economics, Faculty Bioscience Engineering, Ghent University, Coupure Links 653, 9000 Ghent, Belgium dProfessor

and head of the Division of Agri-food Marketing and Chain Management, Department of Agricultural Economics, Faculty of Bioscience Engineering, Ghent University, Coupure Links 653, 9000 Ghent, Belgium

Abstract The purpose of this study is to examine the interplay between value chain integration dimensions and value chain performance along the malt barley value chain in Ethiopia. The analyses were based on survey data sets obtained from 320 farmers and 100 traders and qualitative interview responses captured from sixty-two key informants selected from members of the chain. The structural equation modelling technique was employed to seek answer for the question of how value chain integration dimensions are related to performance. The results of the analyses showed the existence of positive relationships between coordination of activities and performance; between joint decision-making and performance at farmers-cooperatives interface; and between commitment towards long-term relationships and performance at farmers-traders interface. The study has made important empirical contributions in areas of value chain integration and performance and their interplays within the context of the studied malt barley value chain. The key findings of the study make important policy implications for agribusiness value chains in the developing countries. The study would open a venue for robust investigation based on a wider database from various agribusiness chains in Ethiopia or even beyond, for better validation of the findings. Keywords: value chain integration, value chain performance, malt barley value chain, Ethiopia JEL code: Q13 Corresponding author: mulugetadamie.watabaji@ugent.be; mulied75@yahoo.com ÂŠ 2016 Watabaji et al.

Watabaji et al.

Volume 19, Issue 4, 2016

1. Introduction Value chain is a set of three or more members, either organizations or individuals or both. They take part in the forward and reverse flows of materials, services, finances and information from their sources to destinations to create values in the form of products and/or services for customers (Bagchi et al., 2005). In the view of same authors, value chain integration (VCI) deals with the management of these flows to provide superior values to end users (Bagchi et al., 2005). In simple terms, VCI is defined as a set of relationships among suppliers, processors, distributors, retailers and consumers that facilitate the conversion of raw materials into products or services of more value (Darroch and Mushayanyama, 2006; Wever et al., 2009). VCI is a means to create a match between demand and supply of products and/or services at every stage along the value chain (Barratt, 2004). In this study, VCI is defined with the help of four latent concepts termed as ‘VCI dimensions’ throughout the paper. These are: (1) collaboration among value chain members in terms of resources, capabilities and risks sharing; (2) commitment towards long-term relationships; (3) coordination of activities along the value chain; and (4) joint decision-making on key issues like product specification and prices and process improvements. Since past studies focused on VCI as a single variable (Lotfi et al., 2013b), this study is relevant for its completeness. Many past studies generally claimed that VCI improves value chain performance (VCP) outcomes (Arshinder and Deshmukh, 2008; Kim, 2009;Vickery et al., 2003; Wever et al., 2009; Zhao et al., 2008) commonly measured in terms product quality, responsiveness, flexibility and efficiency (Wu et al., 2014). However, the results of these studies are inconsistent (Wiengarten et al., 2010). Moreover, there is a dearth of literature to empirically verify the association between VCI dimensions and VCP (Vereecke and Muylle, 2005; Sezen, 2008; Vanpoucke, 2009; Vickery et al., 2003), especially empirical data from developing countries are scanty (Chin et al., 2014). In the view of Lotfi et al. (2013b) past studies dealt with dyadic interactions between a single value chain member and its chain partners; while chain-level studies were not only few but also descriptive. On the other hand, Bagchi et al. (2005) noted variations in the types of associations between VCI dimensions and VCP whereby commitment showed negative association with VCP while collaboration is positively associated. Moreover, the types of relationships exhibited between VCI dimensions and VCP under one context may not be equally valid under another (Hausman, 2001) and VCI may not always guarantee higher VCP (Vanpoucke, 2009). Therefore, the purpose of this study is to shade light on this research gaps with the help of empirical data obtained from the malt barley value chain (MBVC) in Ethiopia. More specifically, the study aims to: (1) conceptualize the multidimensional constructs of VCI and VCP; (2) measure the current levels of MBVC integration and performance; (3) investigate the relationship between VCI dimensions and VCP at chain-level; and (4) provide some policy implications to address VCI and VCP related challenges in the MBVC in particular and in the agribusiness value chains of developing countries in general. The MBVC is a suitable source of empirical data for this study given the big paradox of chain’s failure to meet more than 40% of the demands for malt from local breweries, though the country produces the largest volume of barley in the African continent. The chain is characterized by limited participation of cooperatives, marginalization of upstream members, involvement of highly opportunistic traders, and dominance of single malt factory both as a buyer of malt barley and seller of malt. The malt factory expresses bitter complaints about the supply of inferior quality malt barley from local sources. The country spends huge amount of foreign currency on imported malt. This study, therefore, seeks an answer as to how VCI dimensions influence VCP outcomes within the context of the MBVC in Ethiopia. The remaining parts of the paper are structured as follows. In the next section, we provide theoretical underpinning of the conceptual framework to set the bases for our research hypotheses. Subsequently, the research methodology is explained, followed by results and discussions. Finally, conclusions are drawn and practical implications are indicated.

International Food and Agribusiness Management Review

Watabaji et al.

Volume 19, Issue 4, 2016

2. Conceptual framework and research hypotheses A conceptual framework for this study was adapted from past study to postulate possible associations between VCI dimensions and VCP which were tested using empirical data obtained from the MBVC in Ethiopia. The framework is primarily based on the resource based view (RBV) which creates a conducive environment to pool resources and capabilities through VCI for superior VCP outcomes (Chin et al., 2014). In the view of Barratt (2004), VCI can only be materialized when members collaborate through resources, capabilities and risks sharing. Similarly, Kim (2009) stressed on the concepts of RBV as key enablers of VCI. According to RBV, resources refer to both tangible and intangible assets, whereas capabilities refer to members’ ability to utilize these resources to achieve higher performance outcomes. No matter how diverse and huge the resources owned by a single member are, it is still not feasible for this member to own every kinds of resources and capabilities in-house. Therefore, VCI is a strategic tool with which members may acquire inimitable complementarities of resources, capabilities and risks that lead to superior VCP. As indicated earlier, VCI is conceptualized in terms of four key dimensions. These are: collaboration (Lotfi et al., 2013b; Wu et al., 2014), commitment (Cechin et al., 2013), coordination (Van Donk et al., 2008), and joint decisions making (Malhotra et al., 2005) to capture its broader and important aspects. As indicated earlier, the other core construct in this study is VCP. In the view of Chan et al. (2003), VCP can be measured using both qualitative and quantitative indicators. In the view of Lotfi et al. (2013a), measurement indicators like added values, efficiency, and customers’ satisfaction can be used to measure VCP. The study by Simatupang and Sridharan (2001) suggests the use of process efficiency, customer satisfaction and financial indicators. In their study on the relationship between VCP and members’ linkages, Won Lee et al. (2007) measured performance using efficiency and effectiveness as indicators. Though various performance measurement indicators were proposed, they are all highly interrelated (Vickery et al., 2003). In most cases, financial indicators are used to measure VCP, though they are not inclusive of all aspects of performance and they are also exposed for misinterpretations (Wu et al., 2014). In immature value chains like the MBVC, data on financial indicators are either unavailable or inaccessible even if available. In line with past studies and data availability, four key indicators were identified to measure MBVC performance. These are: quality, responsiveness, flexibility and efficiency (Gellynck et al., 2008; Vickery et al., 2003; Wu et al., 2014; Zhao et al., 2008). These indicators are broadly acceptable as complete and inclusive (Vereecke and Muylle, 2005). In line with the study by Schloetzer (2012), MBVC members’ perceptions on these indicators were used in this study: ■■ Quality refers to a fitness of products and services to the needs of customers (Lotfi et al., 2013b). In the view of Cao and Zhang (2010), quality refers to the extent to which value chain members offer reliable products that can create greater value for customers. In this paper, quality refers to the moisture content, mix level with other barley varieties, and neatness of the malt barley grains. According to the quality standard set by the malt factory, malt barley grains with low moisture level, admixture free, neat and white are ranked high on the quality scale. These measures of quality are equivalent to ‘attractiveness’ in the view of Molnar (2010) which explains how appealing the appearance of product is to the eyes of customers. ■■ Responsiveness is the measure of capability of value chain members to provide the right product or appropriate service or both within the shortest possible time after receiving orders from the customers (Molnar, 2010). According to her study, lead-time and customers complaints are key indicators of responsiveness. ■■ Flexibility refers to value chain members’ capacity and capability to support changes in products and service specifications to meet the changing needs of customers (Cao and Zhang, 2010). In the view of Sezen (2008), product flexibility, delivery flexibility, mix flexibility and volume flexibility are important aspects of flexibility. ■■ Efficiency refers to the wise use of available resources to generate the maximum possible return while achieving cost competitiveness (Cao and Zhang, 2010). It is a comparison between costs incurred

International Food and Agribusiness Management Review

Watabaji et al.

Volume 19, Issue 4, 2016

and benefits gained in connection with value adding undertakings. It deals with process optimization to produce outputs of higher value using inputs of less value. Based on the literature, the conceptual framework presented under Figure 1 was developed to guide hypotheses formulation, research design, and data analysis and discussion. In the framework, the main constructs are presented in bold and the conceptual indicators are placed in smaller boxes. Collaboration Collaboration among value chain members is identified as VCI dimension and is understood as a win-win philosophy whereby resources, capabilities, and risks are shared among value chain members to achieve higher VCP (Vereecke and Muylle, 2005). In the views of Vieira et al. (2009) and Arshinder and Deshmukh (2008), collaboration is a trustful, loyal and mutual interactions between value chain members and joint efforts towards improved VCP. Collaboration materializes only when value chain members cooperate (Cao and Zhang, 2010). Collaboration is conceptualized to express the extent to which resources (Cao and Zhang, 2010; Wiengarten et al., 2010) and capabilities (Vieira et al., 2009) are shared along the value chain for the purpose of complementarity. In the view of Stank et al. (2001), collaboration is a low-cost strategy that reduces operational wastes and redundancies to improve product and service quality. Whereas, Wiengarten et al. (2010) reported inconsistencies among findings of past studies that relate collaboration and VCP. In their study, Vereecke and Muylle (2005) call for additional empirical underpinning to substantiate the positive interplay between collaboration and performance. Based on the above premises, the following hypothesis was proposed. H1: collaboration between value chain members positively relates to VCP. Commitment Commitment is defined as an enduring desire to maintain long-term relationship between value chain members (Hausman, 2001). Value chain members are committed to long-term relationship when they believe in its importance to enable them achieve higher performance (Darroch and Mushayanyama, 2006; Morgan and Hunt, 1994; Zhao et al., 2008). In the view of Brown et al. (1996), commitment can be classified as normative and instrumental. Normative commitment is a mutual and ongoing relationship over an extended time period based on high trust level between value chain members. Whereas, instrumental commitment refers to value chain membersâ&#x20AC;&#x2122; readiness to bear influences imposed by other value chain members, its ultimate goal being either receipt of rewards or avoidance of punishments. In the view of Wu et al. (2004), commitment is a multifaceted construct of three key aspects: affective, continuance and normative commitments. The affective aspect refers to value chain membersâ&#x20AC;&#x2122; sense of belongingness and attachment to the value chain; the continuance aspect refers to the perceived high costs if value chain members exit from the value chain; and the normative aspect explains both implicit and explicit obligations on value chain members to stay in the value chain. Past studies asserted that commitment towards long-term relationships positively relates to VCP (Brown et al., 1996). In the view of Hausman (2001), less committed value chain members make less effort and resource contributions to ensure higher performance. Similarly, Clarke (2006) suggests that commitment to long-term relationships is a chief strategic tool to improve VCP. Based on these premises, the following relationship was proposed. H2: commitment towards long-term relationships positively relates to VCP.

International Food and Agribusiness Management Review

Collaboration

H1 (+)

Commitment

H2 (+)

Coordination

H3 (+)

Joint decisionmaking

Value chain performance

Volume 19, Issue 4, 2016

Value chain integration

Watabaji et al.

)

H 4 (+

Quality Responsiveness Flexibility Efficiency

Figure 1. Hypothetical conceptual framework. H1 to H4 = hypotheses 1 to 4 (adapted from Vickery et al., 2003). Coordination As noted by Arshinder and Deshmukh (2008), coordination of activities along the value chain requires clear definition of all activities and their proper alignment with value chain goals. It is the act of managing interdependences of the procurement, production and distribution activities along the value chain to improve VCP (Arshinder and Deshmukh, 2008; Vickery et al., 2003). In the view of Darroch and Mushayanyama (2006), coordination of activities along the value chain lowers transaction costs and raises VCP. Furthermore, coordination of activities along the value chain improves membersâ&#x20AC;&#x2122; responsiveness by shortening lead times and increasing membersâ&#x20AC;&#x2122; flexibility through capacity building. Based on these premises, the following hypothesis was forwarded. H3: coordination of activities along the value chain positively relates to VCP. Joint decision-making Joint decision-making refers to the level of participation of value chain members in the decision-making processes of chain partners or the level of sharing decision support information or both (Malhotra et al., 2005; Wiengarten et al., 2010). In the view of Wiengarten et al. (2010), joint decision-making positively relates to operational performance in chain settings, but only if substantiated with free flow of sufficient and quality information along the value chain. Though some authors conceptualize joint decision-making as part of collaboration, members of the malt MBVC consider it as an essential dimension of VCI that should be separately treated. Based on the above premises, the following hypothesis was forwarded. H4: joint decision-making on critical issues like product specifications and prices positively relates to CVP.

3. Research methodology The study contexts and data sources In order to test the validity of proposed associations between conceptual constructs, survey data and interview responses were collected from sample respondents and key informants drawn from MBVC members in Ethiopia. The MBVC is one of the most comprehensive agribusiness value chains in Ethiopia in which several members participate at various stages. The key members of the chain are small-scale farmers, traders, cooperatives, the malt factory, and breweries performing various value adding activities to produce malt barley and ultimately convert it to beer. According to the malt factory, half a million small-scale farmers produce an aggregate of 2.1 million metric tons of barley, which makes Ethiopia the first in the African continent in terms of production volume of which 20% (i.e. 420 thousand metric tons) is suitable for malting. Hence, malt barley makes significant contributions to the national economy (Legesse et al., 2007). Both survey International Food and Agribusiness Management Review

Watabaji et al.

Volume 19, Issue 4, 2016

data and interview responses needed for this study were obtained from selected small-scale farmers, traders, cooperatives staff, and malt factory managers. Small-scale farmers, one of our data sources, are price takers. Due to subsistence nature and risk aversive behavior, these farmers produce malt barley along with other crops for diversification purpose. Since malt barley is also suitable for food and feed, farmers consume nearly 60% of malt barley in-house and sell only about 20% to meet cash needs after reserving some portion for seeds (Legesse et al., 2005). These farmers sell malt barley mostly to traders and rarely to cooperatives at very low prices. Few farmers make direct sales to the malt factory either individually or in groups, because the minimum procurement lot of 5 tons per transaction set by the malt factory discourages the farmers to use this option. Even though hundreds of traders participate in malt-barley collection, only about thirty large ones supply nearly 90% of malt factory’s needs. The large traders collect malt barley from farmers, small traders, and commission agents. Most traders, both large and small, have very good experience to easily identify good quality malt barley from bad ones. If the malt factory pays premium prices, traders can supply best quality malt barley to the factory. Unfortunately, traders opt to mix high quality malt barley with malt barley of low quality to claim good prices since premium prices paid by the factory for best quality is not as such attractive. Cooperatives, another data source of this study, rarely participate in malt barley collections though the malt factory always encourages them to engage on this business. Except one cooperative union in Lemu-bilbilo and another one in Kofele districts, cooperatives in the study area are not engaged in malt barley collection due to structural rigidity, capital limitation, unfair competition from traders, farmers’ reluctance to sell to them, and over-stretching situations regarding the supply of agricultural inputs. The other data source for this study is the malt factory. It is the single dominant buyer of malt barley from farmers, traders and cooperatives (a monopsony) and the single dominant local seller of malt to local breweries (monopoly). The factory can produce 36,000 metric tons of malt per annum out of 50,000 tons of malt barley if operates at full capacity. Presently, the factory’s capacity utilization rate hovers around 80% mainly due to shortage of supply of malt barley with the required quality standards. Its dominance both in the malt barley market as a buyer and malt market as a seller makes it a single price maker in the chain. Sampling and data collection In line with past studies, both qualitative and quantitative data were collected through field surveys and qualitative interviews with selected farmers, traders, cooperatives staff, and malt factory managers. Farmers, traders and cooperatives were selected from Lemu-bilbilo and Tiyyo districts of Arsi zone and from Kofele and Shashemene districts of West Arsi zone. These districts were purposively selected for their wider coverage of malt barley production and market surplus based on the information obtained from the malt factory. From each selected district, random samples of 80 farmers were systematically drawn whereby the kth farmers in the intervals were selected for inclusion in the samples, the starting point being randomly selected from the first interval. The lists of farmers, which are our sampling frames, were obtained from district offices of agriculture. A total of 100 traders, 25 from each selected districts, were included in the survey. Farmers’ and traders’ surveys were conducted during June to August 2013. Prior to data collection, structured questionnaires and interview guides were prepared. The English version of farmers questionnaire was translated into Afan Oromo, the language spoken in the study area, and then re-translated to English to verify the correctness of the translation and to improve clarity. Since traders speak different languages, experienced and multilingual enumerators were hired to translate the English version questionnaire to languages of traders while conducting the surveys (Vanpoucke, 2009). The survey questionnaires and interview guides were pilot tested with few farmers and traders in months of April and May 2013 to ensure content validity. The structure, readability, clarity and completeness of the questionnaires and guides were also reviewed by senior researchers in Agro-food marketing and chain management division International Food and Agribusiness Management Review

Watabaji et al.

Volume 19, Issue 4, 2016

of the Department of Agricultural Economics at Ghent University, Belgium to further improve the validity and clarity of these instruments based on feedbacks from the pilot tests and comments from the researchers. Intensive literature review was done to identify suitable indicators for VCI dimensions and VCP and formulated into various statements to develop the survey questionnaires and interview guides. Survey respondents (i.e. farmers, traders, cooperatives staff, and malt factory managers) were asked to rate the extent of their agreements or disagreement on the statements under VCI dimensions and VCP on five-point scales, 1 = ‘strongly disagree’ and 5 = ‘strongly agree’. In addition to the field surveys, 62 qualitative interviews were conducted of which 27 were with farmers, 13 with traders, 17 with cooperatives staff, and 5 with malt factory managers. Farmers and traders were interviewed to triangulate the survey data sets. Surveys were not conducted with cooperatives staff and the malt factory managers due to small sample size. For all qualitative interviews, MBVC members with good know-how on the operation of the value chain were purposively selected (Vanpoucke, 2009). In total, 320 farmers and 100 traders completed the survey questionnaires. Whenever sampled farmers had refused to fill the survey questionnaire for whatsoever reasons, the next farmers in the list were asked to fill the questionnaire. The detailed profiles of respondent farmers and traders were presented in Table 1. Table 1. Respondents’ profile. Characteristics

Gender distribution male female Age distribution ≤20 years 21-40 years 41-50 years ≥51 years Marital status single married divorced widow/er Educational status not educated read and write primary school secondary school college/university Work experience ≤5 years 6-10 years 11-15 years 16-20 years ≥20 years

Malt barley farmers

Malt barley traders

301 19

94.1 5.9

98 2

2 202 72 44

0.6 63.1 22.5 13.8

2 68 23 7

16 288 8 8

5 90 2.5 2.5

6 92 0 2

43 60 141 65 11

13.4 18.8 44.1 20.3 3.4

0 2 31 58 9

41 120 43 54 62

12.8 43 13.4 16.9 19.4

36 34 25 3 2

International Food and Agribusiness Management Review

Watabaji et al.

Volume 19, Issue 4, 2016

In the study area, farmers produce malt barley along with other competing agricultural crops on an average landholding of 1.86 hectares. On top of that, the average productivity of malt barley is 2 tons per hectare which is lower compared to food barley (2.7 tons) and wheat (2.5 tons) in the study area. The malt barley productivity in the study area is far lower than it is for Europe (7 to 8 tons per hectare) due to poor supply of inputs, limited access to mechanized services, poor linkages along the chain and lack of incentives for farmers. Data analysis After data sorting, within-scale factory analyses (Lin et al., 2005; Sezen, 2008) and Cronbach’s alpha reliability estimate test (Lin et al., 2005; Yu et al., 2013; Zhao et al., 2008) were performed. The factory loadings within-scale were computed to check the validity of all observable indicators to measure the intended multivariate latent variables. Cronbach’s alpha reliability estimates, also called scales of reliability, were used to measure the internal consistency of indicators under a given construct. This is the measure of relatedness of the indicators to manifest a single construct they intend to measure. The summary of factor loadings and alpha reliability estimates for each construct are presented in Table 2. The within-scale factor loadings for all measurement indicators are greater than 0.70 except for PRF1 at farmers-traders interface and for PRF3 at farmers-cooperatives interface that loaded 0.645 and 0.690 respectively (Table 2). In past studies, factor loadings higher than 0.50 are assumed to demonstrate sufficient validity (Lin et al., 2005; Yu et al., 2013). Therefore, few observable indicators loading lower than 0.50 were dropped from further analyses (Table 2). Except for coordination of activities at the traders-malt factory interface, Cronbach’s alpha reliability estimates are higher than 0.70 to reveal strong consistencies among observable items under each multivariate latent variable (Lin et al., 2005; Zhao et al., 2008). In this study, structural equation modelling (SEM) technique was used for data analyses. This technique was chosen for its strength and suitability for the conceptual model developed for this study. As indicated by Tomarken and Waller (2005), SEM technique has the ability to specify latent variable models by providing separate estimates for the associations between latent variables and their manifest indicators (measurement models) and show the relationship among exogenous and endogenous latent variables (structural model); it always provides higher R2 values compared to other techniques; and it provides more information on the relative strength of observed indicators to explain the latent variables as factor analysis is nested in it. As noted by Nachtigall et al. (2003), model suitability can easily be checked by model-fit-statistics under SEM technique. Acceptable fit statistics somehow indicate whether or not (1) observable measurement items fairly manifest the intended latent constructs – measurement models; and (2) the data sets support the proposed associations between exogenous and endogenous variables – structural model (Figure 2). Though the SEM technique provides outputs for both measurement and structural models, outputs of the former were not reported since these outputs are quite similar to factor loadings reported in Table 2. Therefore, we presented only the model-fit-statistics and the path-coefficients of the structural models of the SEM technique. Similar to the works of Wang et al. (2015), Won Lee et al. (2007), and Lin et al. (2005), four SEM diagrams were formulated at four interfaces (Table 3) along the MBVC based on farmers’ and traders’ data sets. In all cases, the models treat collaboration, commitment, coordination and joint-decision as latent-independent (exogenous) variables and VCP as latent-dependent (endogenous) variable. All measurement items with factor loadings of 0.50 or more were used to construct SEM diagrams and to run further analysis while other variables that loaded lower than the threshold were dropped (Table 3). The SEM diagram at farmers-cooperatives interface was presented as a sample (Figure 2) though four SEM diagrams were formulated for the entire analyses. The summated median values for the set of observable indicators were used to explain multivariate exogenous and endogenous latent variables to run the models since summated mean values can only show the locations of estimates that do not exist among the five-point measurement scale (Molnar, 2010). Four separate SEM models were run, two for each data set to assess the relationship between four exogenous latent variables and an endogenous latent variable. International Food and Agribusiness Management Review

Watabaji et al.

Volume 19, Issue 4, 2016

Table 2. Factor loading and the Cronbach’s α estimates (farmers' and traders' survey).1,2 Code

CLB1 CLB2 CLB3 CLB4 CLB5 CMT1 CMT2 CMT3 CMT4 CMT5 CRD1 CRD2 CRD3 CRD4 CRD5 JDM1 JDM2 JDM3 PRF1 PRF2 PRF3 PRF4

Construct and item

Collaboration (α scores) We and our partners form joint teams to work on common projects. We and our partners combine resources on common projects. We unreservedly share our knowledge with our partners. Our partners unreservedly share their knowledge with us. We and our partners expend joint efforts to improve our relations. Commitment (α scores) Our relations with our partners are based on mutual benefits. Our relations with our partners continue for a long future. We like to maintain our association with our partners. We are ready to invest in the relationship with our partners. We have stable relations with our partners. Coordination (α scores) We and our partners jointly manage our activities. We work closely with our partners for effective executions of activities. We and our partners always share activity schedule. We have clear guidelines for interactions with our partners. Our partners strictly follow our interaction guidelines. Joint decision-making (α scores) We and our partners jointly decide on product type. We and our partners jointly decide on process improvements. We and our partners jointly set product prices. Value chain performance (α scores) We improved product quality by working closely with our partners. We improved our responsiveness to customers by working closely with our partners. We enhanced our flexibility by working closely with our partners. We improved our efficiency by working closely with our partners.

F-interfaces

T-interfaces

F-C

F-T

T-F

T-AMF

0.792 – – 0.810 0.868 0.844 0.817 – 0.843 0.843 0.732 0.792 0.778 0.772 0.771

0.791 0.737 – 0.792 0.812 0.833 0.810 – 0.819 0.831 0.774 0.769 0.791 0.827 0.777

0.733 – – 0.751 0.867 0.815 0.882 0.873 0.907 0.753 0.898 – 0.716 – 0.885

0.828 0.804 – 0.814 0.747 0.866 0.701 – 0.765 0.855 0.750 – 0.620 0.825 –

0.800 – 0.759 0.812 0.837 0.880 0.841 0.743 0.821 0.727

0.793 – 0.726 0.807 0.831 0.897 0.826 0.834 0.821 0.727

0.885 – – 0.849 0.901 0.877 0.854 0.711 0.654 0.843

– 0.825 – 0.816 0.800 0.902 0.869 0.707 – 0.821

0.691 0.785

0.901 –

0.842 0.761

1 F-C = farmers-cooperatives interface; F-T = farmers-traders interface; T-F = traders-farmers interface; and T-AMF = traders-Assela

malt factory interface. 2 The empty cells had values lower than 0.50 and were dropped from further analyses.

The models were run on SPSS-AMOS version 22 statistical software (IBM, Armonk, NY, USA). The works of Yu et al. (2013) and Wang et al. (2015) were followed in which case the goodness-of-fit statistics of the models were assessed by (1) chi-square (χ2); (2) normalized chi-square (χ2/df); (3) comparative fit index (CFI); (4) root mean squared errors of approximation (RMSEA); and (5) incremental fit index (IFI). An acceptable χ2 value relative to a given degrees of freedom measures how well the observed distribution of the data set fits the distribution that is expected if the variables are independent. This implies that the theoretical model significantly replicates the samples variance-covariance relationships in the matrix (Schumacker and Lomax, 2004). The CFI measures the improvements of non-centrality obtained by switching from one model to another. The RMSEA, also called discrepancy per degree of freedom, provides an indication of a discrepancy between observed and implied variance-covariance matrices (Hailu et al., 2005). These goodness-of-fit statistics were computed at two interfaces each and presented in Table 4 for farmers and Table 5 for traders along with applicable threshold values.

International Food and Agribusiness Management Review

Watabaji et al.

Volume 19, Issue 4, 2016 Measurement models for exogenous variables

Measurement models for endogenous variables

Structural modal

Figure 2. Structural equation modelling diagram at farmers-cooperatives interface. e1-e19: are codes for error variables; CLB3S, CLB4S and CLB5S are codes for observed indicators under collaboration (CLB) while CLB1S, CLB2S were dropped for loading low; CMT2S-CMT5S are codes for observed indicators under commitment (CMT); CRD1S-CRD5S are codes for observed indicators under coordination (CRD) while CRD4S was dropped for loading low; JDM1S-JDM3S are codes for observed indicators under joint decision-making (JDM); and PFR1S-PFR4S are codes for observed indicators under performance (PRF) (see Table 2 for explanation of the specific codes). Table 3. Malt barley value chain integration interfaces. Interface F-C F-T T-F T-AMF

Farmers’ perceptions about cooperatives’ contributions towards MBVC performance Farmers’ perceptions about traders’ contributions towards MBVC performance Traders’ perception about farmers contributions towards MBVC performance Traders’ perceptions about Assela malt factory’s (AMF’s) contributions towards MBVC performance

Table 4. Model fit statistics from farmers’ survey (n=320).1,2 Statistics

F-C interface

F-T interface

Threshold values

χ2 df χ2/df CFI RMSEA IFI

359.24 124 2.897 0.915 0.077 0.916

333.86 124 2.692 0.926 0.073 0.927

≤2,793.8 ≤300 ≤5.00 ≥0.90 ≤0.08 ≥0.90

1 P<0.001. 2 F-C

= farmers-cooperatives interface; F-T = farmers-traders interface; threshold values adopted from Yu et al. (2013).

International Food and Agribusiness Management Review

Watabaji et al.

Volume 19, Issue 4, 2016

Table 5. Model fit statistics from traders’ survey (n=100).1,2 Statistic

T-F interface

T-AMF interface

Threshold values

χ2 df χ2/df CFI RMSEA IFI

141.67 79 1.793 0.929 0.090* 0.931

134.19 78 1.720 0.914 0.085* 0.917

≤2,793.8 ≤300 ≤5.00 ≥0.90 ≤0.08 ≥0.90

P<0.001. T-F = traders-farmers interface; T-AMF = traders-Assela Malt Factory interface; * = values are slightly higher than the threshold values by Yu et al. (2013). 2

4. Results and discussions According to SEM steps, the research hypotheses in this study can be tested once our survey data sets’ goodness-of-fit to the SEM models are assured (Tables 4 and 5). The study findings were discussed in line with the proposed research hypotheses. Along with our conceptual framework presented in Figure 1, positive relationships between VCI dimensions and VCP were proposed at four interfaces (Table 3). The goodness-of-fit statistics generated from SEM models based on farmers’ and traders’ data sets are within acceptable ranges, except RMSEA values computed at traders’ interfaces. The RMSEA values at tradersfarmers and traders-malt factory interfaces were 0.090 and 0.085 respectively (Table 5) which are slightly higher than the threshold value of 0.08 (Yu et al., 2013). In order to improve models’ goodness-of-fit, a double headed covariance arrow was drawn between two error variables, e16 and e17, in the SEM diagram (Figure 2) as hinted by the modification indices generated by SPSS-AMOS statistical software package (Janssens et al., 2008; Wang et al., 2015). The modification has reduced the χ2 from 378.01 to 359.24 and RMSEA value from 0.080 to 0.077. Even though RMSEA values of ≤0.05 demonstrate the best model fit, still values between 0.05 and 0.10 are acceptable (Han, 2009). Therefore, the generated model-fit-statistics show that our survey data sets fit the models quite well, except the higher RMSEA value for traders’ data set is slightly high probably due to the small sample size. According to results of the structural models from farmers’ data set, coordination (H3) and joint decision-making (H4) are the only exogenous variables that demonstrate significant positive correlation with performance at farmers-cooperatives with standardized path weights of 0.56 and 0.36 respectively. Similarly, commitment (H2) has a significant positive relationship with performance at farmers-traders interface with standardized path weights of 0.62 (Table 6). The t-values for coordination (H3) and joint decision-making (H4) at farmerscooperatives interface are significant at P<0.05, and t-value for commitment (H2) at farmers-cooperatives interface is significant at P<0.01. The t-values for other proposed associations between variables at farmers’ interfaces are less than the minimum threshold of 1.96 which implies insufficient empirical supports (Janssens et al., 2008). According to the standardized path weights for farmers’ data set, coordination of activities (H3), and joint decision-making (H4) at farmers-cooperatives interface significantly correlate with VCP. Interviewed cooperative staff also noted the existence of positive relationship between coordination of various malt barley farming related activities and performance at farmers-cooperatives interface. Moreover, they expressed that joint decision-making on the type, quantity, quality, terms of shipment of agricultural inputs improves performance at farmers-cooperatives interface. Therefore, active participation of farmers in the decision-making processes of cooperatives positively relates to performances. Consistent with the International Food and Agribusiness Management Review

Watabaji et al.

Volume 19, Issue 4, 2016

Table 6. Results of structural model at cooperatives-farmers-traders interfaces (farmers’ survey; n=320).1 Hypothesis: path

F-C interface

H1: collaboration → performance H2: commitment → performance H3: coordination → performance H4: joint decision-making → performance

F-T interface

path coefficient

t-value

path coefficient

t-value

-0.22 0.18 0.56 0.36

0.948 1.039 1.994* 2.427*

0.20 0.62 0.18 -0.22

1.077 3.124** 0.685 1.524

*P<0.05; **P<0.01. 1 F-C

= farmers-cooperatives; F-T = farmers-traders.

finding of this study, Van Donk et al. (2008) noted a positive relationship between joint decision-making on inventory types and batch sizes and performance as it allows an extra flexibility to value chain members. The fact that farmers’ data set provided significant backing to the proposed positive relationships between coordination and performance statistically (H3), joint decision-making and performance (H4) at farmerscooperatives interface and between commitment and performance (H2) at farmers-traders interface goes handin-hand with the findings of past studies. For instance, Simatupang et al. (2002) noted a positive relationship between coordination and performance as coordination improves both flexibility and responsiveness. Similarly Stank et al. (2001) noted a positive correlation between coordination and performance as coordination reduces costs associated with duplication of activities and hence improves efficiency. At farmers-traders interface, commitment towards long-term relationships has significant positive correlation with performance. In the view of interviewed farmers, most malt barley traders are egocentric who always try to maximize own interests at the expense of other value chain members with no commitment towards long-term relationships. Small-scale farmers and other interviewed chain members categorize egotism of traders as critical performance menace. In our opinion, the positive correlation between commitment and performance at farmers-traders interface is resulted from farmers’ desire to work with committed traders. In line with this finding, Clarke (2006) noted a positive relationship between value chain members’ commitment towards long-term relationships and performance as commitment reduces the time and costs associated with recurrent disputes, posturing and renegotiations. In the view of Morgan and Hunt (1994), commitment towards long-term relationships improves performance particularly when complemented with high level of trust and free information flow along the value chain. On the other hand, many researchers noted the existence of positive relationship between collaboration between value chain members and performance ( Cao and Zhang, 2010; Vereecke and Muylle, 2005), farmers’ data set failed to support this hypothesis. Such a contradiction may be due the fact that MBVC members are unaware of the strategic importance of VCI to improve VCP. In the view of interviewed farmers, it was learnt that traders are egotist towards collaboration with farmers which has lowered performance. The malt factory considers traders as opportunists and always reluctant to engage them in any of its MBVC improvement programs. On the other hand, interviewed traders expressed their resentment about an exclusive strategy of the malt factory. Contrary to our expectation, the path coefficients based on traders’ data set are not statistically significant to support the proposed hypotheses at traders’ interfaces (Table 7). Therefore, it is opined that traders’ localized-thinking, non-inclusiveness, and egotism must have contributed to the lack of empirical support. In the view of interviewed malt factory managers, traders are self-seeking and mischievous who always try to serve their greedy profit motives. They, for instance, soak the malt barley in water to deceive the factory on weight and mix superior qualities/varieties malt barley with inferior one to cheat on price. In the view of Cao and Zhang (2010), egotistic actions of value chain members always diminishes VCP. It is harmony, International Food and Agribusiness Management Review

Watabaji et al.

Volume 19, Issue 4, 2016

Table 7. Results of the structural model (traders’ survey; n=100).1 Hypothesis: path

T-F interface

H1: collaboration → performance H2: commitment → performance H3: coordination → performance H4: joint decision-making → performance

1 T-F

T-AMF interface

path coefficient

t-value

path coefficient

t-value

-0.78 0.45 0.47 -0.59

1.724 0.808 0.530 0.660

-0.28 -0.49 0.25 0.09

0.701 1.037 1.344 0.213

= traders-farmers; T-AMF = traders-Assela malt factory.

not isolation, of value chain members that would lead to superior VCP (Gellynck et al., 2008; Vanpoucke, 2009). Moreover, the small sample size of traders could have influenced the statistical significance of the coefficients. The malt factory managers express worries about the poor quality of malt barley supplied through traders which constitutes over 90% of the factory’s malt barley purchases. Similarly, Yu et al. (2013) noted no significant correlation between VCI dimensions and VCP when value chain members are dissatisfied by low service level of chain partners. The study by Wiengarten et al. (2010) on collaborative value chain practices also reported no significant relationship between joint decision-making and VCP with poor information flow along the value chain. The traders’ data set offered no support for the proposed relationships between variables, partly because of lack of awareness of members regarding these relationships. Likewise, interviewed farmers strengthened managers’ views by saying that traders adjust the measurement scale in order to read as low as 85% of the actual weight of supplied malt barley which is even difficult to control since the act is done mischievously. On the other hand, traders regard farmers’ and the factory’s accusations as character assassination which always threatens their long-term participation in the chain. It is, however, interesting to point out that farmers’ data set has moderately supported our hypotheses than traders’ data set which failed to support even a single hypothesis. The varying recognition levels given to farmers and traders by the malt factory are suspected to cause perception differences. The malt factory has been providing several direct and indirect supports to farmers to improve their productivity and establish direct linkages or bridge through cooperatives, though this effort remained unsuccessful. Moreover, MBVC members have not yet started to consider VCI dimensions as part of their strategic means to revive the performance of the chain. Generally speaking, the findings of this study highlight the assertion that VCI dimensions do not always perceived to higher VCP, rather, it depends on the context of the value chain.

5. Conclusions and practical implications This study provides better insights on the relationship between VCI dimensions and VCP based on the data sets from the MBVC in Ethiopia. The fact that very few of the proposed relationships received significant empirical support at the studied interfaces must be due to the particularity of the contexts in a country where the MBVC operates which makes the findings more interesting. The study hinted that the MBVC members, particularly farmers and traders, have not yet started to use VCI dimensions as part of their strategic means to revive VCP. In our views, the low level of maturity of the MBVC and lack of awareness of its members about the strategic importance of VCI dimensions to improve performance are the major contribution to the unique findings. Among the hypothesized relationships, only coordination and joint decision-making at farmers-cooperatives interface and commitment at farmers-traders interface received significant empirical support to be positively related to VCP which show the entry points for interventions. The lack of empirical supports for the proposed International Food and Agribusiness Management Review

Watabaji et al.

Volume 19, Issue 4, 2016

relationships, mostly at traders’ interface, is mainly due to traders’ feelings of exclusion from any VCI programs in addition to the effect of small sample size. The strategy that excludes traders cannot be successful as about 95% of malt barley is collected and supplied to the malt factory by them. The other MBVC members and relevant policymakers should look for policies and strategies that lead to better inclusiveness of traders so as to make them understand the importance of VCI for better performance. Otherwise, cooperatives organizations should be supported to replace traders for the collection and supply malt barley to the malt factory. Though enforcing VCI dimensions can be too expensive, MBVC members had better include them in their strategic plans to revive performance. The huge agro-processors in the chain should create awareness among the upstream small-scale farmers and traders concerning the importance of VCI dimensions in this regard. Moreover, MBVC members and policymakers should establish salient ‘rules of the game’ at every stage of the chain to promote value chain-thinking and VCI practices to enhance performance. Though the use of data sets collected from a single agribusiness value chain in a developing country is an important empirical contribution by itself, more research should be done for better generalizability of the key findings to other agribusiness value chains in Ethiopia and even beyond.

Acknowledgements The study was funded by the Netherlands Organization for International Cooperation in Higher Education (Nuffic) through the NICHE-ETH-019 project, a consortium project of four Ethiopian universities in collaboration with partner universities in the Netherlands and Belgium, to support the commercialization of Ethiopian agriculture. Moreover, the second author acknowledges the financial support of the Hungarian Scientific Research Fund (OTKA, PD 116226), Supply chain and network performance and relationships in the agribusiness sector.

References Arshinder, A. and G. Deshmukh. 2008. ‘Supply chain coordination:perspectives, empirical studies and research directions’. International Journal of Production Economics 115: 316-335. Bagchi, P., B. Ha, T. Skjoett-Larsen and L. Soerensen. 2005. ‘Supply chain integration: a European survey’. International Journal of Logistics Management 16: 275-294. Barratt, M. 2004. ‘Understanding the meaning of collaboration in supply chain’. Supply Chain Management 9: 30-42. Brown, J.R., R.F. Lusch and C.Y. Nicholson. 1996. ‘Power and relationship commitment: their impact on marketing channel member performance’. Journal of Retailing 71: 363-392. Cao, M. and Q. Zhang. 2010. ‘Supply chain collaborative advantage: a firm’s perspective’. International Journal of Production Economics 128: 358-367. Cechin, A., J. Bijman, S. Pascucci and O. Omta. 2013. ‘Decomposing the member relationship in agricultural cooperatives: implications for commitment’. Agribusiness 29: 39-61. Chan, F.T.S., H.J. Qi, H.K. Chan, H.C.W. Lau and R.W.L. Ip. 2003. ‘A conceptual model of performance measurement for supply chains’. Management Decision 41: 635-642. Chin, T.A., A.B.A. Hamid, A. Raslic and L.H. Heng. 2014. ‘The impact of supply chain integration on operational capability in Malaysian manufacturers’. Procedia-Social and Behavioral Sciences 130: 257-265. Clarke, N. 2006. ‘The relationships between network commitment, its antecedents and network performance’. Management Decision 44: 1183-1205. Darroch, M.A. and T. Mushayanyama. 2006. ‘Improving working relationships for smallholder farmers in formal organic crop supply chains: evidence from KwaZulu-Natal, South Africa’. Agrekon 45: 339-360. Gellynck, X., A. Molnár and L. Aramyan. 2008. ‘Supply chain performance measurement: the case of the traditional food sector in the EU.’ Journal on Chain and Network Science 8: 47-58. Hailu, G., P.C. Boxall and B.L. McFarlane. 2005. ‘The influence of place attachment on recreation demand.’ Journal of Economic Psychology 26: 581-598. International Food and Agribusiness Management Review

Watabaji et al.

Volume 19, Issue 4, 2016

Han, J. 2009. Supply chain integration, quality management and firm performance in the pork processing industry in China. Wageningen Academic Publishers, Wageningen, the Netherlands. Hausman, A. 2001. ‘Variations in relationship strength and its impact on performance and satisfaction in business relationships’. Journal of Business and Industrial Marketing 16:600-616. Janssens, W., K. Wijnen, P. De Pelsmacker and P. Van Kenhove. 2008. Marketing research with SPSS. Pearson Publishing, Cambridge, UK. Kim, S. 2009. ‘An investigation on direct and indirect effect of supply chain integration on performance’. International Journal of Production Economics 119:328-346. Legesse, G., S. Debebe and T. Alemu. 2007. Assessing the uncomparative advantage of malt barley production in Ethiopia: application of policy analysis matrix. Paper presented at the African Crop Science Conference, El-Minia, October 2007. Legesse, G., M. Hassena and B. Bedassa. 2005. Malt barley production, marketing and utilization in Arsi. Assela, Ethiopia: Ethiopian Agricultural Research Organization, Kulumsa Agricultural Research Center. Lin, C., W.S. Chow, C.N. Madu, C.-H. Kuei and P.P. Yu. 2005. ‘A structural equation model of supply chain quality management and organizational performance’. International Journal of Production Economics 96: 355-365. Lotfi, Z., S. Sahran and M. Mukhtar. 2013a. ‘A product quality-supply chain integration framework’. Journal of Applied Sciences 13: 36-48. Lotfi, Z., S. Sahran, M. Mukhtar and A.T. Zadeh. 2013b. ‘The relationships between supply chain integration and product quality’. Procedia Technology 11: 471-478. Malhotra, A., S. Gosain and O.A.E. Sawy. 2005. ‘Absorptive capacity configurations in supply chains: gearing for partner-enabled market knowledge creation’. MIS Quarterly: 29: 145-187. Molnar, A. 2010. Supply chain performance and relationships: the European traditional food sector. Ph.D. diss., Ghent University, Ghent, Belgium. Morgan, R.M. and S.D. Hunt. 1994. ‘The commitment-trust theory of relationship marketing’. Journal of Marketing 58: 20-38. Nachtigall, C., U. Kroehne, F. Funke and R. Steyer. 2003. ‘Pros and cons of structural equation modeling.’ Methods of Psychological Research Online 8(2): 1-22. Schloetzer, J. 2012. ‘Process integration and information sharing in supply chains.’ The Accounting Review 87: 1005-1032. Schumacker, R.E. and R.G. Lomax. 2004. A beginner’s guide to structural equation modeling. Psychology Press, Abingdon, UK. Sezen, B. 2008. ‘Relative effects of design, integration and information sharing on supply chain performance’. Supply Chain Management 13: 233-240. Simatupang, T. and R. Sridharan. 2001. ‘A characterization of information sharing in supply chains’. In Proceedings of the 36th Annual ORSNZ Conference, pp. 16-25. Simatupang, T., A. Wright and R. Sridharan. 2002. ‘The knowledge of coordination for supply chain integration’. Business Process Management Journal 8: 289-308. Stank, T.P., S.B. Keller and D.J. Closs. 2001. ‘Performance benefits of supply chain logistical integration’. Transportation Journal 41: 32-46. Tomarken, A.J. and N.G. Waller. 2005. ‘Structural equation modeling: strengths, limitations, and misconceptions’. Annual Review of Clinical Psychology 1: 31-65. Van Donk, D.P., R. Akkerman and T. Van der Vaart. 2008. ‘Opportunities and realities of supply chain integration: the case of food manufacturers’. British Food Journal 110: 218-235. Vanpoucke, E. 2009. Supply chain integration and performance: empirical essays in a manufacturing context. Ph.D. diss., Ghent University, Ghent, Belgium. Vereecke, A. and S. Muylle. 2005. ‘Performance improvement through supply chain collaboration: conventional wisdom versus empirical findings’. International Journal of Operations and Production Management 26: 1176-1198. Vickery, S.K., J. Jayaram, C. Droge and R. Calantone. 2003. ‘The effects of an integrative supply chain strategy on customer service and financial performance: an analysis of direct versus indirect relationships’. Journal of Operations Management 21: 523-539. International Food and Agribusiness Management Review

Watabaji et al.

Volume 19, Issue 4, 2016

Vieira, J., H. Yoshizaki and L. Ho. 2009. ‘Collaboration intensity in the Brazilian supermarket retail chain’. Supply Chain Management 14: 11-21. Wang, O., H. De Steur, X. Gellynck and W. Verbeke. 2015. ‘Motives for consumer choice of traditional food and European food in mainland China’. Appetite 87: 143-151. Wever, M., N. Wognum and J. Trienekens. 2009. ‘Supply chain integration and coordination in the agri-food sector’. In 15th International Conference on Concurrent Enterprising, pp. 22-24. Wiengarten, F., P. Humphreys, G. Cao, B. Fynes and A. McKittrick. 2010. ‘Collaborative supply chain practices and performance: exploring the key role of information quality’. Supply Chain Management 15: 463-473. Won Lee, C., I.-W.G. Kwon and D. Severance. 2007. ‘Relationship between supply chain performance and degree of linkage among supplier, internal integration, and customer’. Supply Chain Management 12: 444-452. Wu, I.-L., C.-H. Chuang and C.-H. Hsu. 2014. ‘Information sharing and collaborative behaviours in enabling supply chain performance: a social exchange perspective’. International Journal of Production Economics 148: 122-132. Wu, W.-Y., C.-Y. Chiag, Y.-J. Wu and H.-J. Tu. 2004. ‘The influencing factors of commitment and business integration on supply chain management’. Industrial Management and Data Systems 104: 322-333. Yu, W., M.A. Jacobs, W.D. Salisbury and H. Enns. 2013. ‘The effects of supply chain integration on customer satisfaction and financial performance: an organizational learning perspective’. International Journal of Production Economics 146:346-358. Zhao, X., B. Huo, B.B. Flynn and J.H.Y. Yeung. 2008. ‘The impact of power and relationship commitment on the integration between manufacturers and customers in a supply chain’. Journal of Operations Management 26: 368-388.

International Food and Agribusiness Management Review

OPEN ACCESS International Food and Agribusiness Management Review Volume 19 Issue 4, 2016; DOI: 10.22434/IFAMR2014.0177 Received: 2 December 2014 / Accepted: 3 October 2016

Dairy farm households, processor linkages and household income: the case of dairy hub linkages in East Africa RESEARCH ARTICLE Elizaphan J.O. Rao a, Immaculate Omondib, Aziz A. Karimovc, and Isabelle Baltenweckd aAgricultural

economist, bMLE scientist, and dProject leader, International Livestock research Institute (ILRI), P.O. Box 30700, 00100, Nairobi, Kenya

cAgricultural

economist, International Maize and Wheat Improvement Center (CIMMYT), Sehit Cem Ersever Caddesi 9-11, 06511 Ankara, Turkey

Abstract In this study we have analysed the effects of household linkages to milk market via dairy hubs currently implemented under the East African Dairy Development project. Our analyses show that participation in dairy hubs increases dairy revenues by USD 1,022 on average. Impacts are higher for households participating in hubs supplying exclusively to processors (USD 1,673) relative to ones supplying hubs that pursue mixedlinkage approach. Moreover, participation in dairy hubs also yields significant effect on household income. Appropriate measures should be undertaken to widen the reach of such processor linkages while also safeguarding existing gains, more so as the processing sector becomes more concentrated. Keywords: smallholder dairy farmers, large processors, dairy hubs, East Africa, dairy revenues, propensity score matching JEL code: D4, L11, O13, Q12, Q13 Corresponding author: J.Rao@cgiar.org

ÂŠ 2016 Rao et al.

Rao et al.

Volume 19, Issue 4, 2016

1. Introduction Dairy production remains an important livelihood option for many poor rural households in the developing world; providing an important source of nutrients and contributing to household income (Duncan et al., 2013; Thorpe et al., 2000). In most developing countries, however, dairy value chains are characterized by multiple market failures that impede participation by dairy producers (Amorim et al., 2013; Barrett, 2008; Wiggins et al., 2010). First, majority of dairy producers in the developing world are smallholders producing low volumes (Holloway et al., 2000; Thornton and Herrero, 2001). This coupled with the scattered nature of their location makes them unattractive suppliers to more structured and reliable market outlets such as processors. Moreover, these poor households are also remotely located with limited access to reliable infrastructure, which leads to higher transaction costs, further compromising their ability to access structured markets (Jayne et al., 2010). Limited access to input markets also heighten cost of production further restricting households to low-input-low-output vicious cycle. In response to these limitations, development agencies continue to promote approaches aimed at enhancing market participation by smallholder dairy producers. These initiatives include provision of market information and promotion of collective action as a means of enhancing access to both input and output markets (Fischer and Qaim, 2012; Njuki et al., 2011). It is expected that with better linkages to markets producers would realize higher and sometime less volatile output prices and lower transaction costs in accessing inputs leading to higher net returns from dairy production. This would lead to improved income for participating households and possibly enhance access to inputs and improved technologies. One such initiative is the dairy hub model implemented under the East Africa Dairy Development (EADD) project that was initiated in 2008 and is currently in its second phase (2013 to 2018). Dairy hubs are geared towards upgrading dairy value chains via linkages to input and output markets mainly through collective action. Through aggregation and bulk selling of milk farmers accrue bargaining advantage when negotiating with milk buyers. This is likely to have positive effect on milk prices for participating farmers. Similarly, households benefit from bulk sourcing of inputs and collectively negotiated rates with service providers thus lowering the cost of production and subsequently leading to improved dairy and household income. The EADD project also provides capacity support to hubs with an aim of making the supported hubs selfsustaining by the end of the projectâ&#x20AC;&#x2122;s implementation period â&#x20AC;&#x201C; the year 2018. Regarding milk sales, dairy hubs pursue diverse market strategies including: exclusive sales of milk to processors and a mixed strategy involving sales to processors and local consumer outlets. In an attempt to understand the impacts of this development initiative, an evaluation was carried out at the end of phase one of EADD and findings showed improved welfare for participating households in terms of household income and dietary diversity score. This was true across Kenya, Uganda and Rwanda where the project was implemented. However, the study largely adopted descriptive approaches with less statistical rigor. Moreover, the analyses did not expressly look at impacts of different marketing strategies on returns to dairy production. In this study we analyse the impacts of different marketing strategies adopted by dairy business hubs on dairy and household income. In particular, we evaluate whether linkages to processors have greater impacts on household welfare. Using data collected from smallholder dairy farm households living within the catchment areas of dairy hubs in Kenya and Uganda (dairy hubs participants and non-participants) and a mix of descriptive analyses and propensity score matching approaches, we provide evidence on the market linkage mechanism that yields the greatest impact on dairy and household income. In the next section we present the analytical framework guiding this study. We then present data and descriptive statistics. This is followed by a presentation and discussion of the results before providing concluding remarks on possible policy implications.

International Food and Agribusiness Management Review

Rao et al.

Volume 19, Issue 4, 2016

2. Analytical framework Program evaluation often follows approaches suggested by Maddala (1983): y = Xβ + γI + u

(1)

Where y can be considered as household or dairy income or any other household welfare indicator; X is a vector of farm, household and contextual characteristics that could influence dairy/household income; and I is a dummy indicating whether or not a household participates in a hub. Holding other factors constant then, the coefficient (γ) captures partial effects of household participation in a hub (in general; that are processorlinked or; hubs linked to other market outlets) on dairy revenue/household income. However, because dairy producers may self-select into participation in hubs, this estimate may be biased. In other words, it is possible that some determinants of hub participation may also affect dairy revenue/household income. If such factors are not included explicitly in Equation 1, as is the case when such variables are unobserved, then the indicator for hub participation in Equation 1 will be correlated with the error term (u) leading to a biased estimation of γ. If participation were randomized, the counterfactual would be observable, making it possible to derive causal inference. Unfortunately, this is not the case in our example. The cross-sectional nature of our data also rules out the possibility of addressing selection bias through panel data approaches. To address the potential selection bias, we propose a matching technique, which assume that conditioning on observable variables eliminates sample selection bias (Heckman and Navarro-Lozano, 2004). Similar approach has been used in the context of agricultural technology adoption (Ali and Abdulai, 2010; Faltermeier and Abdulai, 2009). Matching models essentially create an experimental condition in which participation in dairy hubs is randomly assigned, thus allowing for identification of causal link between hub participation and dairy/household incomes. We use a class of matching models known as propensity score matching (PSM) to measure the effects of participation in processor-linked hubs of various types on household dairy revenue/household income. Instead of directly comparing dairy revenues or household income between households participating in dairy hubs (processor-linked or otherwise) and their counterparts not participating in these hubs, PSM compares between only households participating in dairy hubs (‘treated’) and those households not participating in dairy hubs (‘control’). Moreover, PSM only compares ‘treated’ and ‘control’ households that are similar in terms of observable characteristics, thus reducing the bias that would otherwise occur if the two groups are systematically different (Dehejia and Wahba, 2002). PSM involves two stages. In the first stage, we generate propensity scores P(z) from a probit model that estimates the probability that a household participates in a market linkage program (processor-linked hub for instance). The vector z is of observed conditioning variables that may overlap with variables included in X in Equation 1. We then construct a control group by matching participants in the hub arrangement with participants in other market linkage programs based on similarity of their propensity scores. Households in ‘control’ group for whom appropriate matches cannot be found as well as those not used as matches are dropped. In the second stage, we calculate the average treatment effect on the treated (ATT) households for the outcome variable (household and dairy income), using matched observations of households in the ‘treated’ and ‘control’ groups. The PSM estimator of the ATT is the difference in outcomes between treatment and control groups, appropriately matched by the propensity scores: PSM = E τ ATT P(z│I=1) [E{R1│I = 1, P(z)} – E{R0│I = 0, P(z)}] (2)

where R1 and R0 are outcomes for the treated and control farms respectively; I = 1 indicates treated households and I = 0 control households.

International Food and Agribusiness Management Review

Rao et al.

Volume 19, Issue 4, 2016

There are various matching techniques, but the most common ones include nearest neighbour matching (NNM), kernel-based matching (KBM), stratified radius matching, and Mahalanobis matching (Caliendo and Kopeinig, 2008). In this study, we apply the KBM and the NNM methods. NNM involves pairing farmers in processor-linked hubs and non-processor linked hubs that are closest in terms of P(z) as matching partners. KBM, on the other hand, uses a weighted average of the outcome variable for all individuals in the control group (households in non-processor linked hubs) to construct a counterfactual outcome. Observations that provide better matches are given more weight. The weighted average is compared to the outcome for households in processor-linked hubs, and the difference provides an estimate of the treatment effect for each household supplying the processor-linked hub. A sample average over all processor linked households then provides an estimate of ATT. It is worthwhile noting that PSM can control only for selection bias that is due to observed factors z. In other words, systematic differences between processor linked farmers and non-processor linked farmers may still exist even after conditioning, especially if part of the selection process is based on unobserved variables (Smith and Todd, 2005). Our estimation of ATT is based on the assumption that the distribution of such unobservables is the same for treatment and control groups. However, this is ultimately an empirical question that should be tested (Imbens, 2004). Therefore, we apply the standard bounding test proposed by Rosenbaum (2002), which evaluates how strongly the unobserved variables would have to influence selection to invalidate the implications of the matching process.

3. Data and descriptive statistics Data Dairy farming households in EADD-supported hubs catchment areas were surveyed as part of a baseline for the second phase of EADD project (EADD phase II). Two performance indicators for EADD, increase in milk production and income from milk production, were the main response variables used in estimating the required sample size. Due to the large number of hubs in Uganda, stratification by cattle production system (intensive versus intensive/semi-intensive) was done, for Uganda, in order to estimate the required sample size that is sufficient to elicit the desired response. Consequently, all the dairy hubs supported by phase II of EADD in Kenya (8) and a sample of 24 hubs (out of 33) in Uganda, were included in a household survey conducted between October and December 2014. The required sample sizes for the two countries were estimated to be 322 and 671 cattle keeping farm households from 8 EADD-hubs in Kenya and 24 EADDhubs Uganda respectively, with equal sample sizes per hub in each country. Geo-spatial random sampling technique was used to randomly select smallholder dairy farm households living within catchment area of each dairy hub in Kenya and Uganda. Using a structured questionnaire, data was collected from these households through personal interviews. In Kenya, the 8 hubsâ&#x20AC;&#x2122; catchment area covered Nandi East, Nandi North, Nandi south, Sotik, Narok South, Trans Nzoia and Wareng districts in North and South Rift (Western Kenya). In Uganda, the hub catchment areas span across 13 districts, i.e. Isingiro, Ibanda, Kiruhura, Ssembabule, Masaka, Mukono, Jinja, Kayunga, Kamuli, Kyankwanzi, Wakiso, Kiboga and Mityana districts. Socio-demographic data, data on livestock assets, milk production and utilization, and input use were collected from sampled households in the two countries. In addition to the household survey data, information regarding market outlets used by the dairy hubs, their linkages with processors and other market outlets, were collected between September 2014 and January 2015, from monthly business reports submitted to EADD by the hubs. A sustainability assessment study, conducted between March and April 2015 was used to obtain information on levels of hub sustainability.

International Food and Agribusiness Management Review

Rao et al.

Volume 19, Issue 4, 2016

Dairy hubs and marketing strategies Dairy hubs are farmer-owned collective action-based mechanisms for enhancing market linkages for smallholder dairy farmers. In a situation where smallholder producers are scattered and produce low volumes, it is uneconomical for milk traders and/or processors, as well as input and business service providers to deliver services to farmers. Through bulking and/or chilling, the hubs enable farmers to supply milk to large dairy processors who are the main players in the dairy output market. Courtesy of collective action dairy hubs also command large demand for inputs and services, which could be attractive to inputs and service providers. Besides, most hubs implement a check-off system that enables farmers to access inputs and services on the account of milk delivery, which allows households to access inputs and services even when they do not have cash. Among other services, hubs also provide farmers easy access to loans and training on animal husbandry, which in turn improves smallholder farm management. Hubs therefore reduce transaction costs both for suppliers and buyers of milk thus improving margins from dairy production. Processors on the other hand benefit from reliable supply of high-quality local milk and achieve better control over the supply chain. Finally, local/regional consumers realize gains from affordable and safe milk delivered via efficient milk marketing mechanism and better regulated milk supply. While development partners have been supporting this initiative, hubs remain purely farmer-owned with development partners only playing a facilitative role. This is achieved through capacity building at both farmer and hub management level. Development partners also conduct studies to evaluate performance of the hubs both at farm and hub levels. Development partners also facilitate sharing of lesson that can enhance farm and hub level performance and thus move hubs towards sustainability. With regards to milk marketing strategies, some hubs sell milk exclusively to processors (pure processor hubs) what we refer to here as ‘strategy 1’. On the other hand some hubs sell milk to diverse outlets with large processors being just one of the clients (mixed-linkage hubs) – ‘strategy 2’. Selling to more than one outlet is a risk managing strategy through which hubs may take advantage of potentially higher prices in non-processor outlets. However, non-processor prices are subject to wide fluctuations and this may erode gains from period of higher milk prices. Therefore depending on the share of hub’s milk going to nonprocessor outlet, farmers attached to such hubs may experience lower prices on average and hence lower annual revenues. On the other hand, while processor prices may be low, the option offer stable prices and hence more stable revenue flows for associated farmers throughout the year. Differences in market outlets therefore imply differences in profits that a hub can generate. We hypothesize that these differences would trickle down to farmers in terms of prices that farmer receive for milk sales, thus impacting on dairy and possibly household income. In order to understand the effect of linkage to large processors, this study compares smallholder dairy farmers participating in these different types of dairy hubs: ■■ First we compare households participating in dairy hubs versus those not participating in the hubs (Figure 1). By participation we mean those households that either deliver milk or access inputs and/ or services via hub arrangement. Both elements of participation are expected to impact dairy income – milk sales via revenues and input/service access via cost of production. Non participants in this

all farmers

participants

non participants

Figure 1. The first step compared participating households to those not participating in the hubs. International Food and Agribusiness Management Review

Rao et al.

■■ ■■

Volume 19, Issue 4, 2016

case are livestock keepers who are residents in the catchment areas of respective hubs but neither sell milk nor access inputs and/or services from the hub. We then compare participants and non-participants who reside within the catchment of dairy hubs supplying milk exclusively to large processors, i.e. hubs pursuing ‘strategy 1’ (Figure 2). In order to understand the differential impacts of different market linkages, we also compare participants and non-participants in the catchment areas of hubs pursuing ‘strategy 2’ (Figure 2).

Descriptive statistics Table 1 shows a profile of dairy hubs categorized by country. Approximately 31,700 smallholder dairy farmers were registered as suppliers in the dairy hubs in the two countries by April 2015. These dairy hubs supplied processors and other market outlets with a total of 82,700 liters of milk per day. The dairy hubs in Kenya were all chilling plant hubs with 2 of them adopting a ‘pure processor linkage’ approach while 6 hubs had a ‘mixed linkage’ approach. On the other hand, 10 out of 24 dairy hubs in Uganda adopted a ‘pure processor linkage’ approach. In Table 2, we show some selected socio-economic characteristics of smallholder dairy farmers categorized by participation in dairy farmer groups. We see that farmers actively participating in dairy hubs are significantly more educated. They also have significantly more experience with dairy farming. We also see that majority of farmers in our sample that participate in dairy hubs are from Uganda (58% of participants are from Uganda compared to Kenya’s 42%). Finally we note that significantly more active participants are found in EADD-supported dairy hubs. Similarly significantly more active participants are in hubs linked to processors compared to non-active participants.

all farmers

farmers around pureprocessor hubs

participants

farmers around mixedlinkage hubs

participants

non-participants

Figure 2. Participants and non-participants using mixed linkage hubs are compared. Table 1. Description of the dairy hubs by country, marketing strategy and hub types. Form of linkage

Pure processor linkage

Mixed linkage

number of dairy hubs number of registered suppliers volume of milk (liters) supplied to market outlets per day

Kenya 2 6,507 11,435

Kenya 6 21,534 21,805

Uganda 10 1,483 41,007

International Food and Agribusiness Management Review

100

Uganda 14 2,230 8,539

Rao et al.

Volume 19, Issue 4, 2016

Table 2. Selected socio-economic characteristics of small holder dairy farmers by country and participation in dairy farmer groups.1

age of operator annual dairy revenues (USD) total household income (USD) education (years of schooling) dairy farming experience (years) proportion of male operators farming as a primary occupation (%) sample households from Kenya) (%) sample households in EADD supported groups (%) sample households in non-EADD supported groups (%) sample households in processor-linked hubs (%) 1 Asterisks

Participants n=193

Non-active participants n=800

Characteristic St. error

52 1,640* 12,954* 8* 20* 83.4 77.5 42* 70* 4 36*

50 365 5,379 7 16 83.1 77.3 30 4 5 23

15 2,446 26,288 5 14 37.3 41.8 49 46 19 43

14 983 14,731 5 12 37.5 41.9 46 20 21 42

indicate significance at 1%.

Income distribution We also present a breakdown of sources of household income in addition to average income comparison. The illustration in Figure 3 shows that households in the EADD regions engage in other income generating activities besides dairy. As expected, majority of households engage in sale of milk and other dairy products that are own produced. This is followed by the proportion of households generating income from sale of cattle and other livestock as well as livestock products. Crop revenues and trade in agricultural products (not own produced) follow next in that order.

sale/trade in livestock & livestock products 30% sale of milk & other dairy products 35%

trade in agricultural products 7% formal salaried employment 4%

crop revenues 8%

micro-enterprises 2%

sale of fodder products 3% land & property rent 7%

remittance 0%

pension 1%

farm wages 1% sale of natural resource products 2%

Figure 3. Proportion of households engaging in respective income generating activities.

International Food and Agribusiness Management Review

101

Rao et al.

Volume 19, Issue 4, 2016

We further illustrate in Figure 4 the average amount of revenue generated from respective sources. First, we see that households in the catchment of ‘pure processor’ hubs have higher household income than households in the catchment of ‘mixed-linkage’ hubs. For the whole sample, Figure 4 also shows that household income is largely generated from sale and trade in livestock and livestock products. This is also true for households residing in the catchment of dairy hubs pursuing ‘mixed linkage’ approach. As for households residing in the neighbourhood of ‘pure processor’ dairy hubs, micro-businesses provide the highest contribution to household income. Finally, while dairy revenue ranks low relative to income from other sources, households residing in the catchment of ‘pure-processor’ dairy hubs tend to generate higher revenues from this source than their counterparts in the ‘mixed-linkage’ hubs.

4. Empirical results and discussions Descriptive results discussed in the previous section reveal some differences in dairy and household income between active participants and non-active/non-participants in dairy hubs. However, it is impossible to determine if these differences are due to household participation in respective market linkage programs. In order to attribute these differences to hub participation we conduct statistical matching as described in the analytical framework. Results of these analyses are shown in Table 3. In Supplementary Table S1 we also show results of the probit model used to predict propensity scores that form the basis for our matching. The statistical matching results shown in Table 3 reveal that participation in EADD dairy hubs leads to significantly higher revenues from dairy production. For the whole sample, we find that holding all factors constant, participation in dairy hubs increases annual dairy revenue by USD 1,022 on average. The analyses also show that participation in dairy hubs has a positive and significant effect on total household income – increasing total annual household income by USD 4,628 on average. The larger effect on total household income implies a multiplier effect of dairy revenue; dairy revenues may actually be used as capital in other household income generating activities thus enlarging the total effect of these revenues on household income. For instance, dairy revenues could be supporting micro-businesses which as Figure 4 revealed, is a major contributor to household income especially for participants in the ‘pure processor’ hubs. These results are irrespective of the marketing strategy adopted by the dairy hubs.

ck household income/revenue (USD) tra & li v de es in to ck a gr fo pr i cu rm ltu odu al ct ra sa s lp lar ro ie du d em ct s pl m oy ic sa m ro le en -e of nt t na r ep tu ris ra fa lr es r es eo m w ur ag ce es pr od uc ts pe ns io la n re nd & mit ta sa pr le o p n ce o e f sa rt fo le dd y re of nt e rp m ilk ro cr du & op ct ot s re he ve r to d n ai ue tal s ho ry p r us od eh uc ol ts d in co m e

18,000 15,000 12,000 9,000 6,000 3,000

sa le /

tra de

liv e

sto

whole sample

pure processor

mixed linkage

Figure 4. Average household earnings from different activities. International Food and Agribusiness Management Review

102

Rao et al.

Volume 19, Issue 4, 2016

Table 3. Results of the propensity score matching. Matching algorithm

ATT1

Outcome

z-statistic2 Critical level of hidden bias (Γ)

Effect of household participation in dairy hubs nearest neighbour annual dairy revenue 1,022.81 4.26*** total household income 4,628.69 2.13** kernel annual dairy revenue 990.31 4.62*** total household income 4,690.60 2.12** Effect of household participation in pure processor-linked dairy hubs nearest neighbour annual dairy revenue 1,673.22 2.64*** total household income 7,463.28 1.34 kernel annual dairy revenue 1,676.41 3.02*** total household income 7,682.69 1.59 Effect of household participation dairy hubs with mixed-linkage nearest neighbour annual dairy revenue 747.41 2.81*** total household income 3,807.36 1.27 kernel annual dairy revenue 654.10 2.72*** total household income 3,216.00 1.38

Treated Control (n) (n)

2.00-2.05 1.35-1.40 1.60-1.65 1.20-1.25

186 186 186 186

738 738 738 738

2.55-2.60 – 2.10-2.15 –

97 97 97 97

242 242 242 242

3.60-3.65 – 2.50-2.55 –

89 89 89 89

494 494 494 494

1 ATT

= average treatment effect on the treated. and *** are significant at the 5 and 1% levels, respectively. The z-values for the ATTs are based on bootstrapped standard errors with 500 replications.

2 **

We also note that while hubs are supposed to enhance access to inputs and services and could thus lower production costs for households, we believe that higher dairy revenues for participating households are largely due to higher prices since as can be seen in Figure 5, dairy revenues are largely driven by milk sales.

revenues (USD)

When we narrow our focus to only those households in the catchment of hubs supplying exclusively to processors, the treatment effect of participation in dairy hubs on dairy revenue is even larger. For this category of households, participation in dairy hubs yields an impact on annual dairy revenues to the tune of USD 1,673. These findings are in contrast with Navarro et al. (2015) which found for the case of Peru that informal markets tend to offer higher profits per litre of milk than formal channels. Similarly, average treatment effects on treated for household income are larger for this sub-sample of households. This effect is,

1000 900 800 700 600 500 400 300 200 100 0

from milk sales

from value addition annual revenue

Figure 5. Components of dairy revenues. International Food and Agribusiness Management Review

103

Rao et al.

Volume 19, Issue 4, 2016

however, insignificant. This is possibly due to the value of dairy revenue relative to total household income. As shown in Figure 4, dairy revenue for households in pure processor linkage is fairly low compared to total household income (1000=5 USD to 9,742 USD). This is comparison to the whole sample where the gap is relatively not wide (964 USD to 7,292 USD). We make a further comparison between participants and non-participants for households in the catchment of ‘mixed-linkage’ hubs. The lower panel of Table 3 shows that participation in dairy hubs for this subsample of households also yields positive and significant effects on dairy revenues, albeit by far lower magnitude relative to ‘pure processor’ linkage. Participation in this category of hubs leads to an increase in dairy revenues by 747 USD on average. As already explained, while price offers by processors may be low relative to alternative outlets, stability in prices ensures that overall annual returns are higher for households supplying pure processor-linked hubs. This is in comparison to mixed linkage approach where wide fluctuations in price offers by non-processor outlets may erode gains from windows of higher prices leading to low prices on average and hence low annual revenues. Treatment effects on household income are, however, insignificant for this sub-sample. Finally, we test for statistical difference in the distributions of treatment effects between the two sub-samples. Figure 6 compares the distribution of the two treatment effects (for ‘pure processor’ and ‘mixed linkage’ sub-samples). Kolmogorov-Smirnov test confirms that the cumulative distribution function (CDF) of treatment effects for ‘pure processor’ linkage statistically dominates the CDF of treatment effects for ‘mixed-linkage’. Sensitivity analyses The main weakness of PSM relies on the fact that program participation is only explained by observed (observable) covariates. The approach would therefore be effective as long as bias from unobserved covariates remains minimal. However, if program participation variables that are usually used to balance the treated and comparison sub-samples are incomplete, PSM results can be biased. It is therefore critical that factors driving participation in the program (especially dropped nonparticipant are carefully investigated and are included in the modelling of PSM to the extent possible. Including more observed variables ensures that matched treated and untreated observations are as similar as possible. This follows from the assumption that some of the unobserved factors may be related to many matching variables included in the PSM modelling, which allows for reduction of the potential bias that emanated from omission of unobserved variables. We then test for bias reduction (BR) by estimating a bias reduction index as follows: pure processor

pixed linkage

1.0

probability

0.8 0.6 0.4

KS-Statistic = 0.00134 (P=0.000)

0.2 0.0 -5,000

0 5,000 10,000 average treatment effect on treated (for dairy revenue)

15,000

Figure 6. Cumulative distribution of treatment effects on dairy revenues by market linkage. KS = KolmogorovSmirnov test. International Food and Agribusiness Management Review

104

Rao et al.

Volume 19, Issue 4, 2016

(

BR = 100 1â&#x20AC;&#x201C;

Biasafter

Biasbefore

) (3)

BR values greater than 20% are considered large â&#x20AC;&#x201C; indicating substantial reduction of bias achieved via matching (Rosenbaum and Rubin, 1985). Additionally pseudo R-squared should be fairly low after matching to ensure that there are minimal systematic differences in the distribution of covariates between treated and comparison groups, i.e. the two groups are more similar thus allowing for unbiased comparison of outcome variable. Another issue with PSM is that it requires large sample of non-participants from which matches are drawn. It is needed so that enough variation is provided in the representative sample. Otherwise interpretation of treatment effect results will mislead policy implications. We show results of this BR test in Table 4, which reveals that the variance of treatment status explained by covariates declined substantially after matching. Similarly likelihood ratio test (P-value) shows that the joint significance of covariates on treatment status cannot be rejected before matching, while it is rejected after matching. The joint insignificance of covariates together with the low pseudo R-squared after matching imply that there is no systematic differences in the distribution of covariates between groups after matching. Our matching is therefore based on fairly similar observations of matched and comparison groups with minimal bias if any. While test results in Table 4 have shown that our matching procedure was successfully able to balance the distribution of observed characteristics, hidden bias may still arise if there are unobserved variables that simultaneously affect assignment into treatment. Matching estimators are not robust to such hidden bias. We therefore test for potential hidden bias from farmer heterogeneity due to unobserved variables using the bounding approach suggested by Rosenbaum (2002) and explained in the analytical framework. Assuming

P-value of LR (unmatched)

P-value of LR (matched)

74.2 74.2 75.1 75.1

0.55 0.55 0.55 0.55

0.05 0.05 0.04 0.04

0.000 0.000 0.000 0.000

0.503 0.503 0.770 0.770

21.5 21.5 21.5 21.5

11.0 11.0 14.4 14.4

48.8 48.8 33.0 33.0

0.564 0.564 0.564 0.564

0.165 0.165 0.126 0.126

0.000 0.000 0.000 0.000

0.187 0.187 0.544 0.544

47.2 47.2 47.2 47.2

21.8 21.8 12.2 12.2

53.8 53.8 74.2 74.2

0.682 0.682 0.682 0.682

0.241 0.241 0.108 0.108

0.000 0.000 0.000 0.000

0.163 0.163 0.960 0.960

LR = Likelihood ratio test. International Food and Agribusiness Management Review

105

Pseudo R (matched)

8.5 8.5 8.2 8.2

Pseudo R (unmatched)

32.9 32.9 32.9 32.9

% bias reduction

Median absolute bias (after matching)

Matching algorithm Outcome Whole sample nearest neighbour annual dairy revenue total household income kernel annual dairy revenue total household income Processor-linked hubs nearest neighbour annual dairy revenue total household income kernel annual dairy revenue total household income Mixed-linkage hubs nearest neighbour annual dairy revenue total household income kernel annual dairy revenue total household income

Median absolute bias (before matching)

Table 4. Indicators of covariate balancing before and after matching.1

Rao et al.

Volume 19, Issue 4, 2016

two individuals have the same observed covariates z (as implied by the matching procedure), the two matched observations would differ in their odds of participating in the dairy hubs only by the difference in unobserved covariates, which is measured by the parameter Γ. The test procedure involves changing the level of Γ and deriving the bounds on the significance levels of the ATT under the assumption of endogenous self-selection into dairy hub participation. This allows for identification of the critical levels of Γ at which the estimated ATT would become insignificant. Results of this test are shown in the fifth column of Table 3. Using the example of dairy revenues for the pure processor linkage, the critical values for hidden bias (Γ) are 2.10-2.15 with KBM and 2.55-2.60 with NNM. The lowest value of Γ=1.6 implies that individuals that have the same z-vector would have to differ in their odds of participation in dairy hubs by at least a factor of 1.6 (60%) in order to render the ATT for dairy revenues insignificant. Even though unobserved variables may play a certain role, it is very unlikely that they would influence the odds of participation in dairy hubs to such a great extent.

5. Conclusions Dairy activities account for a significant proportion of household income in the milk producing zones. However, potential for increased dairy incomes is compromised by many smallholders operating below capacity, largely occasioned by limited access to services, essential inputs and business support. Consequently, milk prices are usually low while input costs are high, thus constraining milk profit margins. These limitations have motivated various market linkage mechanisms aimed at increasing participation by households in output markets while also linking households to input and service markets. Dairy business hubs is one such imitative currently implemented under the EADD project. By bulking and chilling milk through the hubs, farmers can bargain for higher prices from milk processing companies. Additionally, the hub offers a onestop source of essential dairy-related inputs and services such as feeds, drugs, breeding, animal health and extension services, usually under flexible payment arrangements. While milk prices may sometime be lower in hubs than other outlets, the possibility of accessing inputs and services cost effectively and under flexible payment arrangements makes hubs a preferred market access mechanism for many households. These gains translate to increased production levels and enable farmers to cost-effectively produce higher volumes of milk, thus leading to increased dairy income. The effects may also spill over to total household income. In this study, we have analysed the effects of dairy households’ participation in the hub as implemented under the EADD project. First, our findings show that participation by households in dairy hubs significantly increases both dairy revenues and household income, irrespective of the market linkage pursued by respective dairy hubs. Subsample analyses (market-linkage sub-samples) reveal even higher effects for households participating in dairy hubs that sell milk exclusively to processors. This is in comparison to households that supply milk to dairy hubs following a mixed marketing approach. Another important result relates to the multiplier effect of dairy income on total household income, possibly due to investment of returns from dairy in other revenue generating enterprises at household level. Yet processors and other formal outlets often have difficulties buying milk from smallholders due to their scattered and remote locations. Appropriate measures need to put in place to eliminate some of the barriers that smallholder experience in trying to access processor outlets. Morgan (2009) underscores the need for collective approaches involving cooperatives as a means to accessing the more formal milk markets including processors. However, the success of such collective approaches hinges crucially on good governance, limited state management and the fit of respective collective action approaches to cultural and socio-economic context. More flexible approaches involving business-oriented farmer groups operating outside the tenets of cooperative laws could be more successful in linking smallholders to formal markets including processors. The EADD project recognizes this fact and has therefore encouraged formation of the flexible and business-oriented producer organizations. Producer organisations (POs) at the centre of dairy International Food and Agribusiness Management Review

106

Rao et al.

Volume 19, Issue 4, 2016

hubs decide on their market linkage strategy, whether to sell exclusively to processors or to diversify to different milk buyers, based on their internal business strategy as well as their external environment. Based on these results, development agencies and facilitators should be encouraged to work closely with POs to identify the most appropriate market linkage strategy, including focusing on dairy processors, to maximise impact on dairy farmers income. In spite of the impressive effects of household participation in pure processor-based linkages, caution needs to be taken to avoid monopolistic tendency that emerge as consolidation occurs in the processing sector. Increased consolidation often leads to competition among large processors and occasionally to lower prices paid to producers (Morgan, 2009). To safeguard against such ills, farmer organizations/dairy hubs should pursue more formal contracts with processors stipulating purchase prices and payment schedules. Some more advanced dairy hubs could also integrate vertically by establishing their own processing units, which would give farmers even guarantee better prices. Finally, in light of the apparent benefits and given the low levels of farmer participation dairy hubs currently, future work should seek to understand incentives/disincentives for farmer participation in the hub arrangements.

Supplementary material Supplementary material can be found online at https://doi.org/10.22434/IFAMR2014.0177. Table S1. Propensity score model.

Acknowledgements This article is based on the research conducted on the East Africa Dairy Development program project and we thank all partners and donors. It was also partly funded by the Livestock and Fish program of the CGIAR and we therefore thank all donors that globally support our work through their contributions to the CGIAR system (http://www.cgiar.org/about-us/our-funders). We extend our appreciation to the project team and the dairy producers and Producers Organizations for their cooperation during the research. The usual disclaimers apply.

References Ali, A. and A. Abdulai. 2010. The adoption of genetically modified cotton and poverty reduction in Pakistan. Journal of Agricultural Economics 61: 175-192. Amorim, P., H. Meyr, C. Almeder and B. Almada-Lobo. 2013. Managing perishability in production-distribution planning: a discussion and review. Flexible Services and Manufacturing Journal 25: 389-413. Barrett, C.B. 2008. Smallholder market participation: concepts and evidence from eastern and southern Africa. Food Policy 33: 299-317. Caliendo, M. and S. Kopeinig. 2008. some practical guidance for the implementation of propensity score matching. Journal of Economic Surveys 22: 31-72. Dehejia, R.H. and S. Wahba. 2002. Propensity score-matching methods for nonexperimental causal studies. Review of Economics and Statistics 84: 151-161. Duncan, A., N. Teufel, K. Mekonnen, V. Singh, A. Bitew and B. Gebremedhin. 2013. Dairy intensification in developing countries: effects of market quality on farm-level feeding and breeding practices. Animal 7: 2054-2062. Faltermeier, L. and A. Abdulai. 2009. The impact of water conservation and intensification technologies: empirical evidence for rice farmers in Ghana. Agricultural Economics 40: 365-379. Fischer, E. and M. Qaim. 2012. Linking smallholders to markets: determinants and impacts of farmer collective action in Kenya. World Development 40: 1255-1268.

International Food and Agribusiness Management Review

107

Rao et al.

Volume 19, Issue 4, 2016

Heckman, J. and S. Navarro-Lozano. 2004. Using matching, instrumental variables, and control functions to estimate economic choice models. Review of Economics and Statistics 86: 30-57. Holloway, G., C. Nicholson, C. Delgado, S. Staal and S. Ehui. 2000. Agroindustrialization through institutional innovation transaction costs, cooperatives and milk-market development in the East-African highlands. Agricultural Economics 23: 279-288. Imbens, G.W. 2004. Nonparametric estimation of average treatment effects under exogeneity: a review. Review of Economics and Statistics 86: 4-29. Jayne, T.S., D. Mather and E. Mghenyi. 2010. Principal challenges confronting smallholder agriculture in Sub-Saharan Africa. World Development 38: 1384-1398. Maddala, G.S. 1983. Limited-dependent and qualitative variables in econometrics. Cambridge University Press, Cambridge, UK. Morgan, N. 2009. Models and opportunities for smallholder dairy producers in Asia: lessons learned. Smallholder dairy development: lessons learned in Asia. Available at: http://tinyurl.com/oxpjjnd. Navarro, E.F., G. Faure, E. Cortijo, E. De Nys, J. Bogue, C. GĂłmez, W. Mercado, C. Gamboa and P.-Y. Le Gal. 2015. The impacts of differentiated markets on the relationship between dairy processors and smallholder farmers in the Peruvian Andes. Agricultural Systems 132: 145-156. Njuki, J., S. Kaaria, A. Chamunorwa and W. Chiuri. 2011. Linking smallholder farmers to markets, gender and intra-household dynamics: does the choice of commodity matter&quest. European Journal of Development Research 23: 426-443. Rosenbaum, P.R. 2002. Observational studies. Springer Science+Business Media, New York, USA. Rosenbaum, P.R. and D.B. Rubin. 1985. Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. American Statistician 39: 33-38. Smith, J.A. and P.E. Todd. 2005. Does matching overcome Lalondeâ&#x20AC;&#x2122;s critique of nonexperimental estimators? Journal of Econometrics 125: 305-353. Thornton, P. and M. Herrero. 2001. Integrated crop-livestock simulation models for scenario analysis and impact assessment. Agricultural Systems 70: 581-602. Thorpe, W., H. Muriuki, A. Omore, M. Owango and S. Staal. 2000. Dairy development in Kenya: the past, the present and the future. Paper presented at the Annual Symposium of the Animal Production Society of Kenya (APSK), Nairobi, Kenya, 22-23 March 2000. Available at: http://tinyurl.com/olo2ex6. Wiggins, S., J. Kirsten and L. LlambĂ. 2010. The future of small farms. World Development 38: 1341-1348.

International Food and Agribusiness Management Review

108

OPEN ACCESS International Food and Agribusiness Management Review Volume 19 Issue 4, 2016; DOI: 10.22434/IFAMR2015.0109 Received: 14 July 2015 / Accepted: 10 September 2016

Fertilizer freight rate disparity in Brazil: a regional approach CASE STUDY Lilian M. de Limaa, Lilian de Pelegrini Eliasb, José V. Caixeta-Filho c, and Jamile de Campos Coletib aAssistant

Professor, Escola Superior de Agricultura Luiz de Queiroz (ESALQ), University of São Paulo (USP), Padua Dias, 11 Avenue, 13400-970, Piracicaba/SP, Brazil

bDoctoral

student and master in Economics Development, Economy Institute (IE), University of Campinas (UNICAMP), Pitágoras street 353, 13083-857 Barão Geraldo, Campinas/SP, Brazil

cFull

Professor, Escola Superior de Agricultura Luiz de Queiroz (ESALQ), University of São Paulo (USP), ESALQ-LOG (coordination), Padua Dias, 11 Avenue, 13418-900 Piracicaba/SP, Brazil

Abstract An increase in Brazilian agricultural product exportation with a concurrent increase in the use of fertilizer has put pressure on the country’s already overtaxed transportation system and expanded the number and intensity of transportation bottlenecks, especially during the grain harvest and planting seasons. Problems with the transportation system have led to an increase in fertilizer transportation costs and a disproportionate increase in fertilizer’s share of total agricultural production costs, highlighting the need to discover the most economic fertilizer transportation routes. Our research found a significant variance in fertilizer transportation costs among different Brazilian transport regions, referred to as transport corridors in this study. Literature on the subject has found that regional fertilizer shipping price variations are often contingent on the presence of shipping intensive industries, ports and storage centers. Using a comparative analysis based on an econometric model, this study examines the effect of intra-regional fertilizer transportation routes on shipping costs and clarifies the dynamics of fertilizer transport in Brazil. Keywords: fertilizer, transport, econometric, freight JEL code: Q13, C10, R10, R40 Corresponding author: jose.caixeta@usp.br

109

De Lima et al.

Volume 19, Issue 4, 2016

1. Introduction and background According to the Brazilian Ministry of Agriculture (MAPA, 2016), Brazilian agribusiness exports accounted for 45.9% of total Brazilian exports between December 2014 and November 2015. Brazilian agribusiness accounted for an average of 23.2% of the country’s total gross national product between 2004 and 2013 (CEPEA, 2016). Over those 10 years, the agriculture sector was the largest contributor to both the country’s balance of trade and its gross national product. The main destinations for these exports were the European Union, the United States, China, Russia, Argentina, Japan, Iran, and Venezuela. By weight, soybean and its derivatives are the most exported Brazilian agricultural products and are usually shipped from the ports of Paranaguá, Santos and Rio Grande do Sul (MAPA, 2016; Ribeiro et al., 2009). According to data published by the government’s Companhia Nacional de Abastecimento (CONAB, 2016), the 2015/2016 Brazilian grain harvest was an estimated 210.5 million MT. The United States Department of Agriculture (USDA, 2016) estimated Brazilian soybean production from the 2015/2016 crop at 100 million MT, almost the half of the country’s grain production, 50.4% higher than from the 2011/2012 crop and only 6.5% lower than soybean production in the United States, the world’s principal soybean producer (CONAB, 2016). The exportation of Brazilian agriculture products has significantly increased over the last 10 years. Soybean exports alone have risen over 600% since 1997. According to the Secretaria do Comércio Exterior (Brazilian Secretariat of Foreign Trade (SECEX), 2016), Brazil exported 8.3 million MT of soybeans in 1997; in 2015 Brazilian soybean exports reached 54.3 million MT. The increase is continuing: soybean exportation rose 18.9% between 2014 and 2015. With the bulk of Brazil’s soybean production located in the landlocked Centro-Oeste (Midwest) region (Izumi, 2012), the rapid rise in international sales has highlighted the country’s logistic deficit, especially in the product transport segment. In June of 2015, 23.9% of the price of soybeans went to pay transportation costs. Each year, it takes more time and more money to transport grains from Brazil’s Midwest to international markets (Oliveira, 2011), making logistic research extremely relevant. The key to logistics success is managing difficulties, mainly brought about by inclement weather, poor infrastructure and seasonal flux. The logistics chain can be divided into many segments, such as transportation, storage, material handling, protective packaging, acquisition, planning, and information collection. The chain’s objective is the efficient transport of the demanded product to the right place at the right time and in the right condition while minimizing the total cost of operation. During the harvest and post-harvest periods, inadequate transportation and loading services and ineffective product packaging often disrupt the shipping chain’s dynamic, opening the door to significantly higher producer transportation costs. For agricultural products and their inputs, efficient logistics is essential to the maintenance of a competitive pricing structure. Because agriculturally oriented products have low value, the cost of transportation is an important component of total product price, which differentiates them from high added-value products. To get a sense of the extent of the logistics component of Brazilian agricultural costs, this study will analyze fertilizer transportation pricing over the multi-year period of greatest agricultural product exportation. The aim of this study is to better understand the dynamics of fertilizer logistics in Brazil and, with the aid of a multiple linear regression econometric model, to ascertain the impact of shipping route selection on fertilizer transportation pricing. In order to obtain greater adjustment of the model, explanatory variables other than the transport corridor were included. These additional variables represent the price of diesel, distance traveled (km), and binary variables for the periods when the fertilizer was transported. The estimated coefficients of these additional explanatory variables serve as complementary results without representing the central objective of this work.

International Food and Agribusiness Management Review

110

De Lima et al.

Volume 19, Issue 4, 2016

It should be noted that this study does not aim to estimate the cost of fertilizer freight transport. The study makes use of a multiple regression model employing the ordinary least squares (OLS) estimation method to conduct a descriptive analysis of selected variables relating to the transport corridors.

2. Logistical context The Council of Supply Chain Management Professionals (CSCMP, 2016) writes that ‘Logistics management is that part of supply chain management that plans, implements, and controls the efficient, effective forward and reverse flow and storage of goods, services and related information between the point of origin and the point of consumption in order to meet customers’ requirements.’ Logistics includes all production handling activities and information processing to, from and between participants in a supply chain. The goal of logistics is to make products and services available where they are needed and when they are desired (Bowersox and Closs, 2004: p. 13). Fulfilling this goal is an important part of controlling company costs and enhancing life in the country as a whole. The Council report reveals that total US business logistics activities represented 17.9% of US GDP in 1980 and 8.3% of US GDP ($1.45 trillion) in 2014. In Brazil, business logistics represented 11.5% of GDP in 2012 according the Brazilian Institute of Logistics and Supply Chain (ILOS, 2016). The recognition of logistics as a component of business structure is recent. Modern logistics, which comprises the logistics of the transformation process in specific sector companies, began after the Second World War. The postwar period demanded that industry quickly fill the gap between increasing demand and reduced consumer supply by taking advantage of the idle capacity at industrial plants and innovating new production processes. This larger-scale production required tight and flexible integration between manufacturing segments and strategic planning so that stock could be quickly turned into key elements, from which arose standardization (Novaes, 2007). Strategically designed logistics extends beyond punctual optimization; it includes the incorporation of competitiveness as a differentiator. A properly organized logistics structure gives coherence to the whole production system and, as such, has become essential to business success. Logistics determines what will be produced in what quantity, organizes the supply of raw materials, and ensures the successful, timely delivery of the product to the consumer: the key to sales success. The complexity of logistics is addressed in the concept of total cost. All the processes involved in logistics should be analyzed in a unified way. Comprehensive analysis was made possible due to the development of sophisticated information technology and management acumen. Logistics decisions should cover the entire production chain and the market in which it operates, seeking the best balance between service, end user satisfaction, and process costs. The growing relevance of logistics in total production cost and in the design of the production process makes investigation of the logistics dynamic underlying each product’s production and sale essential. This is particularly true when it comes to agricultural inputs, such as fertilizer, in a country like Brazil, a country of continental proportions and high demand. Fertilizer transportation According to Michelon (2007), trucking does have some great advantages over other means of transport due to its scheduling flexibility and ease of cargo combination. The timing of freight pick-up and dispatch can usually be adjusted to meet the customer’s needs much more easily when the freight is hauled over a road transport system than when moved by waterway or rail. If transported by road, cargo can be accepted for shipment, loaded, combined with other cargo, and expeditiously delivered, giving trucking companies an advantage over other transport modes when working in the spot freight market. The ability to combine International Food and Agribusiness Management Review

111

De Lima et al.

Volume 19, Issue 4, 2016

cargos and make advantageous use of the spot market gives a roadway shipping company many opportunities to operate at full capacity on both outbound and return legs. There are many ways to transport final product and raw materials to and from Brazilian agriculture, but the predominant grain and agricultural input transport mode is via the country’s road system (Scherer and Martins, 2004). Lima (2001) calculated that 2/3 of all freight in Brazil is carried over roadways. Unfortunately, road transportation in Brazil is hampered by limited infrastructure. For grains, this is especially evident during the peak harvest season. The Brazilian road system’s failings lead to seemingly contradictory transportation problems: serious congestion and a shortage of adequate vehicles. These difficulties are reflected in extremely high seasonal freight rates that are most evident in the spot market. Other forms of transport, such as railways and waterways, are in development in Brazil; but these lower cost alternatives are not yet available. At this time, the only way to reduce transportation costs is through efficient use of the road system. Freight transport via a road system is also differentiated from most other freight transport modalities in that it has lower fixed costs and higher variable costs. Freight transport over long distances by roadway is quite a bit costlier than moving the same freight by train or waterway. Freight movement usually makes up about 60% of all logistics costs in developed countries (Rodrigues, 2007). The costs of moving freight linked with shipping distance are considered variable costs, and all costs that arise independent of shipping distance are considered fixed costs (Lima, 2001). Correa Jr. and Caixeta-Filho (2003) and Lima (2001) describe the main variables that influence freight rates, which can be divided into six categories, as shown in Table 1. The fixed costs of freight transportation in Brazil may be less constant than in the developed world. First, there is the rather high rate of inflation that causes shipping costs to often change rather quickly and disrupt planning. Second is the environmental push: the Brazilian government is adding taxes and road use regulations to encourage the use of more sustainable transportation alternatives (Steadieseifi et al., 2014). The variable costs depend on issues influenced by the differentiated price conditions in each region, such as seasonality, infrastructure, and the potential for ‘return freight’ (back loading). The following section gives Table 1. The main variables that influence freight rates (adapted from Correa Jr. and Caixeta-Filho, 2003; Lima, 2001). Variable costs related to travel and distance traveled

Fixed costs related to shipping company operation

Product handling facilities and peculiarities

• fuel • oil • tire • lubricants • washing • road-use taxes • tolls • other maintenance

• taxes • insurances • licensing • depreciation • facilities • staff (driver) • administration • business taxes • financing Infrastructure

• type and dimensions of cargo • load risk (flammable, toxic or theft prone) • operating costs • vehicle specificity (refrigerated, tanker, grain or fruit hauler)

Market conditions • seasonality • possibility of return freight

Organizational

• regional peculiarities that includes road conditions • traffic

• competition or synergy with other transportation modes • tolls and working scales along the route • lead time to delivery

International Food and Agribusiness Management Review

112

De Lima et al.

Volume 19, Issue 4, 2016

a broad overview of transport dynamics in the Brazilian fertilizer sector and a more detailed discussion of the return freight concept. Market conditions: seasonality and return freight In this study, ‘seasonality’ refers to changes in freight rates as determined by seasonal transportation demand. Seasonal fluctuations in the demand for fertilizer transportation indicate that fertilizer freight logistics are very different from that of other agricultural products. Fertilizer deliveries in Brazil never stop but oscillate within a narrow range and follow a different schedule relative to grains. Fertilizer shipments are at their lowest in April and highest in September. Deliveries of 1.1 million metric tons (MMT) of fertilizer were made in April of 2010, 3.9 MMT in September of 2014; and 1.4 MMT in April of 2015, the least monthly amount delivered that year. The seasonal fluctuation in demand for soybean transportation is enormous: minimal in December and January, peaking in May and June. In April 2016 10.1 MMT of soybeans were exported, 26 times the January 2016 amount (394 thousand metric tons (TMT)). The disparity in the amounts of fertilizer and soybean transported and the timing of that transportation opens opportunities to transport fertilizer back to grain producing regions as return freight, which should benefit both the shipper and the shipping company. Return freight is that cargo that returns to the shipping service’s initial debarkation point. If a good is shipped from point X to point Y, the return freight would be that cargo that returns to point X from point Y. This return freight is also referred to as the ‘backload.’ When the soybean harvest in central west Brazil is at its height, the amount of freight to be shipped by roadway to export facilities exceeds the normal demand for freight to be shipped from those facilities back to the harvest area. This is a common occurrence at some stage of most of the country’s grain harvests, soybean being used as an example because it is the most exported crop. To secure the grain hauling truck’s return from the export facility during the harvest, grain producers may be obliged to pay for transport of their outbound freight and the return of an empty truck to the harvest area. This translates into extremely inflated freight rates for the shipment of just harvested grains to export facilities, especially to the Brazilian ports of Santos and Paranaguá. Any additional payment received for the shipment of a return cargo is of benefit to the shipping company. Carrier agents consider return freight to be a compensation; and the price for grain transportation can be adjusted by the probability of return freight, which often leads to negotiation between both shipper and shipping company as a part of price formation (Oliveira et al., 2010). Fertilizer is a very good return freight candidate, giving it an important role in the return freight calculation. Fertilizer is bulky, relatively imperishable, necessary and imported into Brazilian ports in great quantity. Other products also lend themselves to this strategy: limestone, cement, soybean meal, wheat, bagged sugar, sorghum, citrus pulp, seeds, gypsum, industries’ finished products and construction materials, such as bricks and tiles. The combination that occurs most frequently in Brazil is outbound with soybean and return with fertilizer (Michelon, 2007). The Brazilian soybean harvest takes place from late January to May, with the first very small quantities exported in late January and early February. As noted earlier, as the harvest reaches its peak, there is an increase in demand for one-way grain transportation services. Although this is a time of low fertilizer use, it is also a time when scheduling fertilizer as return freight is most economically advantageous. At this point, the fertilizer shipper must decide if it is rational to make use of the return freight option or wait and ship during the planting season when the cargo will used almost immediately. By taking advantage of the return freight option, shipping companies can reduce round trip expenses and grain producers and fertilizer suppliers can reduce shipping costs. When the shipping company is making a International Food and Agribusiness Management Review

113

De Lima et al.

Volume 19, Issue 4, 2016

profit even if the truck returns to its point of origin empty, return freight is transformed into an opportunity to increase shipping company profits and lower the shippersâ&#x20AC;&#x2122; costs. Oliveira et al. (2010) note that return freight can be considered an opportunity cost. Product shippers seeking to minimize costs and shipping companies seeking to maximize profits are aligned in their desire to arrange a return freight. Fertilizer market According to Brazilâ&#x20AC;&#x2122;s National Association for Fertilizer Promotion (ANDA, 2014), the country used 30.2 million metric tons of fertilizers in 2015, 23% more than used in 2007 (24.6 MMT), the first year ANDA compiled fertilizer utilization data. Since 2004, Brazilâ&#x20AC;&#x2122;s fertilizer segment has experienced an average annual growth rate of 3.0%; although, the segment stagnated in 2005, 2008, 2009 and 2015, most likely due to an increase in international fertilizer prices and maritime freight costs in those years. A survey conducted by the Brazilian Federal University of Rio de Janeiro (UFRJ) in conjunction with the Brazilian Chemical Industry Association (ABIQUIM, 2009) found that the products most imported into Brazil were raw materials used in fertilizer production. Between 2004 and 2015, fertilizer use in Brazil increased 32.7% while domestic production decreased 6.8% (ANDA, 2014). The gap between domestic production and consumption was filled by imports, with the importation of fertilizer and fertilizer components increasing by approximately 62.4% between the 2004 and 2015 (ANDA, 2014). Tavares and Haberli Jr. (2011) note that between 2007 and 2010, 95% of the potassium chloride used as fertilizer in Brazil was imported. Figure 1 shows the seasonal changes in fertilizer delivery, domestic production and importation. Brazilian fertilizer shipments begin to increase during the second half of the first semester, when planting of the summer grain crop begins, with greatest overall demand in the second semester. Considering that imported fertilizer and fertilizer components make up such large percentage of Brazilian fertilizer consumption, it is hard to overstate the importance of the transportation infrastructure responsible for moving these goods in the final cost of agricultural products (Teixeira, 2013). Seasonality also affects most agricultural products. Figure 2 shows the seasonal fluctuation in soybean exportation. While fertilizer importation and consumption are concentrated in August, September and October, soybean exportation is shown to be concentrated in April, May and June. 4.4 Quantity (million metric tons)

4.0 3.6 3.2 2.8 2.4 2.0 1.6 1.2 0.8 0.4 2008.1 2008.4 2008.7 2008.10 2009.1 2009.4 2009.7 2009.10 2010.1 2010.4 2010.7 2010.10 2011.1 2011.4 2011.7 2011.10 2012.1 2012.4 2012.7 2012.10 2013.1 2013.4 2013.7 2013.10 2014.1 2014.4 2014.7 2014.10 2015.1 2015.4 2015.7 2015.10

0.0

Domestic production of intermediate fertilizers Imports of intermediate fertilizers Fertilizers delivered to the final consumer

Figure 1. Fertilizers delivered to the final consumer, domestic production of intermediate fertilizers, and imports of fertilizer production inputs in million metric tons (ANDA, 2014). International Food and Agribusiness Management Review

114

10 9 8 7 6 5 4 3 2 1 0

2002.1 2002.3 2012.5 2012.7 2012.9 2012.11 2013.1 2013.3 2013.5 2013.7 2013.9 2013.11 2014.1 2014.3 2014.5 2014.7 2014.9 2014.11 2015.1 2015.3 2015.5 2015.7 2015.9 2015.11

Volume 19, Issue 4, 2016

Quantity (million metric tons)

De Lima et al.

Figure 2. Million metric tons soybeans exported (SECEX, 2016).

3. Econometric approaches in the transport sector According to Correa (2001), mathematical modeling is widely used in studies addressing transport issues. Econometric analysis has been particularly useful to identify key factors influencing freight rates and to estimate transportation demand functions. Thompson (1960), for example, found that the cost to ship chemicals in the United States was directly related to the distance traveled. The same author highlights the logarithmic form log-log as a good alternative when modeling the non-linear relationship. Kerr (1972) used OLS analysis via multiple regression techniques to study railway freight rates for products with different characteristics in the United States. The author considered miles driven, the overflow rate, and load weight as explanatory variables. Binkley and Harrer (1981) analyzed the determinants of marine grain shipping prices using two linear models estimated by OLS. One of their models examined the average effect of the following explanatory variables on the freight rate: travel distance and travel distance squared, ship’s size, and the size of the ship to the square of the distance, the transported volume, and binary variables to reflect the season of shipment, types of transport contract (‘free discharge’ and ‘gross terms’), and if the carrier was a US-flagged vessel. Hauser (1986) derived a single road shipping function using an OLS regression dealing with ten functions related to the producer’s cost to transport by roadway and length of route in the United States. The author concluded that due to intense competitiveness within the grain road transport industry, the freight rate is equivalent to shipping company operating costs plus a 2% profit margin. Prentice and Benell (1992), on the other hand, used a multiple linear regression model to estimate the utility of American transportation companies when transporting loads with different attributes (origin, need for refrigeration, destination, loading/unloading duration). They note that the transportation of red meat from Canada to the United States was the ‘desirable’ return freight. Beilock et al. (1996) developed a study to identify the determinants of the road freight rate for the flow of international goods in Europe and Western Asia. The factors, identified as rate determinants, were number of borders crossed, road conditions, and the use of ferries to cross waterways. The authors point out that although loading ability is an important factor in determining the freight rate, this was not considered in the study because the data refer to a category of carriers that do not seek return cargo. Garrido and Mahmassani (2000) developed a predictive model of transport demand using the variability of demand as a function of time and space. The authors employed a multinomial probit model and a MonteCarlo simulation to evaluate the likelihood of the multinomial probit model. The model developed by the

International Food and Agribusiness Management Review

115

De Lima et al.

Volume 19, Issue 4, 2016

authors considered the demand for shipping as a stochastic process, identified by an econometric model with a probability distribution function and interaction between the alternatives considered. Many of the key explanatory variables and mathematical approaches used in the cited studies are similar. Multiple regression models, OLS models, and autoregressive models are particularly evident among the more often used mathematical techniques, and are most probably selected according to data characteristics and the methodological predilection of those involved in research. The inference here is that there is no preferable technique but that researchers must use methodology that best suits the study’s characteristics and their own preferences. Some factors that appear more frequently in freight transportation studies are distance traveled, characteristics of the cargo’s point of origin and destination, loading and unloading times, load type and the value of the product carried. It was decided that our study would employ a multiple linear regression model estimated using the OLS method to evaluate variables that affect fertilizer freight rates, especially trip routing through selected transport corridors. Variables were selected based on related literature and data availability. It is noteworthy that multiple linear regression is a statistical tool that has wide application in the social sciences, especially in fields related to management, economics, and sociology (Hoffmann, 2015). This technique is concerned with the study of the dependence of a variable, the ‘dependent variable,’ on other variables; the ‘independent variables,’ with the objective of evaluating and/or predicting the average (population) or the average value of the dependent variable in terms of the known values of the explanatory variables (Oliveira, 2014). However, it should be clarified that while a regression analysis considers the dependence of one variable in relation to another, it need not imply a causal relationship. The success of any multiple linear regression analysis depends on the relationship between the dependent variable and explanatory variables as well on the availability of appropriate data and an adequate, suitable theoretical construction.

4. Methodology and specification of data This study examines road freight rates when fertilizer is shipped through different Brazilian regions, represented by transit corridors, and evaluates the impact of four variables on these rates. The main criterion for the choice of regions was that the selected routes in the region’s road transportation corridor attend to significant fertilizer supply and demand. This entails that routes connect cities serving as hubs for fertilizer blending operations, shown in Table 2, ports, which are often adjacent to fertilizer blending industries, and grain production areas, shown in Table 3. The differences among the five corridors’ freight rates will also give an indication of the effect of return freight on fertilizer transportation costs. This study uses multiple linear regression to determine the impact of variables related to five Brazilian regions on the fertilizer freight rate. The five regions are represented by five different transportation corridors. Table 2. Number of fertilizer manufacturing operations in or near cities located in the transport corridors (RAIS, 2012). State1

City

Fertilizing manufacturers

State

City

Fertilizing manufacturers

PR MT MG GO PR SP

Paranaguá Rondonópolis Uberaba Catalão Curitiba São Paulo

27 20 18 12 10 10

SP SP SP PR MG GO

Campinas Ribeirão Preto Cubatão Maringá Uberlândia Anápolis

8 8 7 6 6 6

1 MT

= Mato Grosso; MG = Minas Gerais; GO = Goiás; PR = Paraná; SP = São Paulo. International Food and Agribusiness Management Review

116

De Lima et al.

Volume 19, Issue 4, 2016

Table 3. Cities centered in Brazil’s biggest maize and soybean producing areas by megaton of production in 2012 (IBGE, 2012). City

State1 Maize2

Soybean2

City

State Maize

Soybean

Sorriso Jataí Rio Verde Sapezal Nova Mutum Lucas do Rio Verde Campo Novo do Parecis Maracaju Nova Ubiratã Primavera do Leste Diamantino

MT GO GO MT MT MT MT MS MT MT MT

1.961.880 863.100 907.500 1.130.326 1.107.481 716.550 1.063.800 615.000 890.988 744.000 873.600

Itiquira Querência Campo Verde Campos de Júlio Sidrolândia Dourados Ipiranga do Norte Santa Rita do Trivelato Montividiu São Gabriel do Oeste Vera

MT MT MT MT MS MS MT MT GO MS MT

615.000 629.640 882.126 590.700 563.565 358.800 280.000 494.748 466.095 368.880 349.800

1 2

1.998.402 1.221.000 1.070.000 817.004 775.720 1.089.710 597.000 918.000 608.405 588.748 447.400

642.600 558.780 284.272 494.712 521.515 611.850 606.600 388.500 413.400 489.000 474.800

MT = Mato Grosso; GO = Goiás;. MS = MatoGrosso do Sul. Quantity in million metric tons.

The dependent variable (FREIGHT) was defined as the average price in Brazilian Reais per metric ton (R$/ MT) charged by transportation companies to transport bulk fertilizer by roadway over routes in our selected transport corridors and is contingent on the cargo’s origin and destination, the period of fertilizer flow (month and year), and the length of the route (km). These data are based on a set of information consisting of 14,878 observations from January 2010 through March 2014. The observations are grouped by freight origin and destination into five sets, with each set indicative of routes in one of the transport corridors. Table 4 lists the corridor classification details. Each observation relates to one route, and a particular route may be used by different shipping companies. These data came from the Information System for Freights (SIFRECA), a monitoring and freight price data collection system under the direction of the Agroindustrial Logistics Research and Extension Group (ESALQLOG) from the Escola Superior de Agricultura Luiz de Queiroz/USP. SIFRECA is specifically concerned with road freight rates for the transport of Brazilian agricultural and agri-business products and collects data using periodic surveys of producers, processors, and traders. It should be noted that data used to calculate the average and nominal cost for bulk fertilizer road transport excludes information from self-employed carriers and tax and insurance figures and that ‘cost’ is the value paid to the transportation company by the agent in need of transportation services. For reasons of confidentiality, the ESALQ-LOG research group does not provide information broken down by transportation company, nor does their available database contain data more current than the period covered by our study. It was not possible to update the data. In addition to the freight rate over a particular route (R$/t) and the pairs of origin-destination cities that make up that route, the SIFRECA database also provided a measure for distance covered linked with each observation, represented by the variable (KM) in the model. It should be noted that in one month, for example, the same route can be observed with different freight rates for fertilizer transport because the transportation was provided by different carriers. The purpose of this study is to assess the impact of road transport corridors on the price of fertilizer shipping. For a better specification and robustness of analysis, other explanatory variables were included: distance, diesel price and period the fertilizer was transported.

International Food and Agribusiness Management Review

117

De Lima et al.

Volume 19, Issue 4, 2016

Table 4. Routes classification according to fertilizer transport corridors (ESALQ-LOG, 2014). Corridor A (COR_A)

Origin

Corridor B (COR_B)

Origin Destination

Corridor C (COR_C)

Origin

Corridor D (COR_D)

Origin

Corridor E (COR_E)

Origin

Destination

Routes connecting Paranaguá (port) and the city of Curitiba (fertilizer industries in close proximity to Paranaguá) to grain producing areas in Brazil’s Midwest: State of Mato Grosso: Sorriso, Sapezal, Nova Mutum, Lucas do Rio Verde, Campo Novo do Parecis, Nova Ubiratã, Primavera do Leste, Diamantino, Querência, Campo Verde, Campos de Júlio, Ipiranga do Norte, Santa Rita do Trivelato, Brasnorte, Sinop, Tapurah, Santo Antônio do Leste, Itiquira, and Vera; State of Mato Grosso do Sul: Maracaju, Sidrolândia, Dourados, São Gabriel do Oeste, Ponta Porã and Rio Brilhante; State of Goiás: Jataí, Rio Verde, Cristalina, Montividiu and Chapadão do Céu Routes connecting Paranaguá (port) to fertilizer industries in the State of Paraná: Curitiba and Maringá; State of Mato Grosso: Rondonópolis; State of Goiás: Catalão and Anápolis; State of Minas Gerais: Uberaba and Uberlandia; State of São Paulo: São Paulo, Campinas, Ribeirão Preto and Cubatão Routes connecting Santos (port), Guaruja (port) and Cubatao (12 km from Santos and home to fertilizer industries) to grain producing areas in Brazil’s Midwest: State of Mato Grosso: Sorriso, Sapezal, Nova Mutum, Lucas do Rio Verde, Campo Novo do Parecis, Nova Ubiratã, Primavera do Leste, Diamantino, Querência, Campo Verde, Campos de Júlio, Ipiranga do Norte, Santa Rita do Trivelato, Brasnorte, Sinop, Tapurah, Santo Antônio do Leste, Itiquira, and Vera; State of Mato Grosso do Sul: Maracaju, Sidrolândia, Dourados, São Gabriel do Oeste, Ponta Porã and Rio Brilhante; State of Goiás: Jataí, Rio Verde, Cristalina, Montividiu and Chapadão do Céu Routes connecting Santos (port), Guarujá (port) and Cubatão (fertilizer industries and adjacent to the port of Santos) to fertilizer industries in the State of Paraná: Paranaguá, Curitiba and Maringá; State of Mato Grosso: Rondonópolis; State of Goiás: Catalão and Anápolis; State of Minas Gerais: Uberaba and Uberlândia; State of São Paulo: São Paulo, Campinas, Ribeirão Preto and Cubatão Routes connecting fertilizer industries in the State of Goiás: Catalão and Anápolis; State of Minas Gerais: Uberaba and Uberlândia to grain producing regions in Brazil’s Midwest: State of Mato Grosso: Sorriso, Sapezal, Nova Mutum, Lucas do Rio Verde, Campo Novo do Parecis, Primavera do Leste, Diamantino, Querência, Campo Verde, Campos de Júlio, Brasnorte, Sinop, Tapurah, and Vera; State of Mato Grosso do Sul: Maracaju, Dourados, and Rio Brilhante; State of Goiás: Jataí, Rio Verde, Cristalina, Montividiu and Chapadão do Céu

Description of variables In the empirical model used in this study, the fertilizer freight rate (R$/t)1 was considered to be function of the following variables: ■■ Five binary variables associated with the five selected road transportation corridors: COR_A, COR_B, COR_C, COR_D, and COR_E. These corridors are used to transport the great majority of fertilizer used in Brazil. Data regarding fertilizer flow in the five corridors’ and nominal2 freight transportation prices were provided by ESALQ-LOG (ESALQ/USP). The corridors are defined in Table 4.

1 The fertilizer freight rate used in the model is in R$/MT not in US$/MT because the behavior of the Real$ is more relevant to freight rates in Brazil

than the US$ when trying to understand the Brazilian fertilizer freight market’s dynamic. 2 The rationale for this is due to the absence of a specific deflator for freight and little influence from inflation (given by general price inflation) on freight values, since a considerable part of these values correspond to previously contracted amounts and therefore are fixed for a period.

International Food and Agribusiness Management Review

118

De Lima et al.

■■

■■ ■■

Volume 19, Issue 4, 2016

Three binary variables, T1, T2 and T3, associated with periods of fertilizer transport. T1 represents the three month period of March through May; T2 represents the seven month period of June through December and T3 represents two month period of January through February (the whole year was considered). Mean distance (km) of travel, described as variable (KM); source: ESALQ-LOG (ESALQ/USP). Mean diesel price – Brazil (R$/liter)3, described as variable (DIESEL); source: Agência Nacional de Petróleo (ANP).

The variables’ descriptive statistics The figures in Table 5 show that approximately 61% of the observed fertilizer freight traffic was in Corridor A, that the mean price of diesel (R$/ltr) was R$ 1.84 or US$ 0.5624, that the mean distance traveled over the observed routes was 1,521 km and that period when the most routes were traveled (53%) was from June through December (T2). During the T2 period, from June through December, vehicles transporting fertilizer traveled over 53% of the designated routes, the highest percentage of the three transit periods. This was certainly to be expected as the T2 period is both the longest period, seven months, and encompasses the grain planting season: the time of greatest fertilizer demand. 28% of the routes were used to transport fertilizer during the three month March through May T1 period. These are the soybean harvest months and would be the most economically advantageous months for fertilizer to be used as return freight. Fertilizer was transported over 18% of the routes during the January through February T3 period, a period of few grain shipments and the least advantageous for the use of fertilizer as return freight. Empirical model A multiple linear regression equation employing the OLS method is used to estimate the coefficients of the previously described variables (Koutsoyiannis, 1972). These coefficients are then used to capture the influence of each of the five transit corridors on the price to ship fertilizer. The additional explanatory variables (diesel price, period of fertilizer flow, kilometers traveled) were included to achieve better model performance and improve its robustness. 3 The diesel price used in the model is in R$/liter not in US$/liter because the Brazilian government sets diesel prices in the Brazilian domestic market. 4 Real$

to commercial dollar conversion, July 12, 2016: 1 US$ = R$ 3.2750 (Banco Central do Brasil, 2016).

Table 5. Description of the exogenous variables and descriptive statistics (data from 2014). Variables

Description

Mean

Standard deviation

COR_A COR_B COR_C COR_D COR_E DIESEL* KM* T1 T2 T3

1 if route belongs to corridor A, 0 if it does not 1 if route belongs to corridor B, 0 if it does not 1 if route belongs to corridor C, 0 if it does not 1 if route belongs to corridor D, 0 if it does not 1 if route belongs to corridor E, 0 if it does not Average diesel price (R$/liter), per month Distance in km, from each fertilizer route 1 if the route occurred in the months between March and May, 0 if it does not 1 if the route occurred in the months June and December, 0 if it does not 1 if the route occurred in the months January and February, 0 if it does not

0.61319 0.11964 0.09699 0.04934 0.12085 1.84 1,521.34 0.28687 0.53212 0.18100

0.48704 0.32455 0.29595 0.21657 0.32596 0.22 625.53 0.45231 0.49898 0.38503

Continuous variables International Food and Agribusiness Management Review

119

De Lima et al.

Volume 19, Issue 4, 2016

The proposed model follows the functional form, 5

i=1

j=6

lnY = α + ∑ βiXi + ∑ βjXj + β9lnX9 + β10lnX10 + ε where ■■ ■■ ■■ ■■ ■■ ■■ ■■ ■■ ■■

(1)

lnY corresponds to the natural logarithm of the fertilizer freight rate in R$/MT; i corresponds to the corridor in which the routes are classified – Corridor A (i=1); Corridor B (i=2), Corridor C (i=3), Corridor D (i=4) or Corridor E (i=5); j refers to the period of fertilizer transportation – T1 (j=6); T2 (j=7) e T3 (j=8). α, βi, βj, β9, β10 are the estimated model parameters; Xi refers to the binary variable corresponding to the i-th type of Corridor; Xj refers to the binary variable corresponding to the j-th fertilizer transport period; ln X9 refers to the natural logarithm of the average diesel price in Brazil (R$/liter); ln X10 refers to the natural logarithm of the distance of each fertilizer route (km); and ε corresponds to the random error (distribution N (0,1) was assumed.

The E-Views 6.0 (IHS Global Inc., Irvine, CA, USA) statistical program and the ‘R-Studio’ program (RStudio Inc, Boston, MA, USA) were adopted to estimate the regression parameters, to carry out tests, and to create graphics useful when analyzing the results.

5. Results and discussion ‘Soybean price’ was initially one of the explanatory variables added to improve the estimated model’s robustness but was excluded from the final model due to the presence of endogeneity, which is discussed in the following section. Presence of endogeneity The use of instrumental variables allows consistent estimations when the explanatory variables are correlated with a linear regression’s error term. In this situation, the linear regression can produce biased and inconsistent estimations. However, consistent estimates may also be obtained when an instrumental variable is available. As part of this analysis, each explanatory variable was tested for endogeneity using the test developed by Hausman (Gujarati, 2006: p. 605; Wooldridge, 2010: p. 495;). The following structural model was used to carry out this test: Y= β0 + β1X1 + β2X2 + β3X3 + u where ■■ ■■ ■■ ■■

(2)

Y refers to the fertilizer freight rate; X1 diesel price (in Brazil); X2 distance of each route considered in the data (km); X3 soybean price; and u model error term.

The following variables are considered exogenous and not included in the model above: the binary fertilizer transport periods (T1, T2 and T3), and the binary representing each corridor (Cor_A, Cor_B, Cor_C, Cor_D, Cor_E). Even in this context, for each individual assessment of the explanatory variables (X1, X2 and X3) in Equation 2, the other explanatory variables are considered exogenous. Thus, for the assessment X1 as endogenous, X2 and X3 are considered exogenous; for the assessment of X2 as endogenous, X1 and X3 are considered exogenous, and for the assessment of X3 as endogenous, X2 and X1 are considered exogenous. To analyze whether the diesel price variable (X1) is endogenous, the reduced form of X1 (the regression of X1 against any exogenous variables or predetermined variables including exogenous variables that were not International Food and Agribusiness Management Review

120

De Lima et al.

Volume 19, Issue 4, 2016

considered in the structural model (Equation 2)) is estimated, yielding residual û1. Then, û1 will be added as an explanatory variable in the structural model that includes X1, and the significance of the coefficient of û1 is checked. The coefficient was obtained from a t test using the OLS method. If the coefficient of û1 is statistically different from zero, i.e. significant at a specific level of significance, the null hypothesis that the coefficient of û1 is equal to zero is rejected; and it follows that X1 is endogenous. The same individual analysis is repeated for variables X2 and X3. ‘Regression 1,’ ‘Regression, 2,’ and ‘Regression 3’ correspond to regressions obtained by regressing the fertilizer freight rate (R$/t) against exogenous variables in the structural model (Equation 2), which are X1, X2 and X3 with their residuals û1, û2 and û3, respectively. It can be highlighted that: ■■ û1 corresponds to the residual vector obtained from the estimation of regression model 1 with X1 being the dependent variable against the explanatory variables X2, X3, the binary corridor variables Cor_A, Cor_B, Cor_C, Cor_D and Cor_E and the binary shipping period variables T1, T2 and T3; ■■ û2 corresponds to the residual vector obtained from the estimation of regression model 2 with X1 being the dependent variable against the explanatory variables X1, X3, the binary corridor variables Cor_A, Cor_B, Cor_C, Cor_D and Cor_E and binary shipping period variables T1, T2 and T3; ■■ û3 corresponds to the residual vector obtained from the estimation of regression model 3 with X1 being the dependent variable against the explanatory variables X1, X2, the binary corridor variables Cor_A, Cor_B, Cor_C, Cor_D and Cor_E and binary shipping period variables T1, T2 and T3. The residual coefficients considered as explanatory variables of regressions 1, 2 and 3 are shown in Table 6. According to the values shown in Table 6, the coefficient of û3 was statistically different from zero, i.e. significant at 1% significance; therefore, the null hypothesis that the coefficient of û2 is equal to zero is rejected and that X3 (soybean price) is endogenous. The coefficients û1 and û2 are not significant, indicating that the null hypothesis that the coefficients are equal to zero is not rejected and that variables X1 and X3 are exogenous (diesel price and distance, respectively). For the chosen instrumental variable to be considered adequate it needs to be correlated with the explanatory variable and not correlated with the error term (Wooldridge, 2010). When the variable ‘soybean price’ was found to be endogenous, various instrumental variables were analyzed; but no statistically significant results were generated. As an example, the ‘volume of exported soybeans’ and ‘volume of imported fertilizers’ were selected as acceptable instrumental variables correlated with soybean price. When comparing the adjusted coefficient of determination (adjusted R-square) of regressions containing either of these instrumental variables with results from the model without these variables, i.e. the model with only the variables diesel price, length of transited routes and the binary variables fertilizer transport period (T1, T2 and T3,) and corridor (COR_A, COR_B, COR_C. COR_D and COR_E), it was found that the model lacking an instrumental variable was the more significant in terms of higher values for the explanatory variables’ adjusted R-square coefficients and the direction of their signals, indicating that the model lacking instrumental variables was the more Table 6. Residual estimates û1, û2 and û3, as explanatory variables of regressions 1, 2 and 3, respectively with data from 2016. Residual estimates

Coefficient (estimates)

Statistic t

P-value1

û1 (Regression 1) û2 (Regression 2) û3 (Regression 3)

-5.28 6.2×10-4 -1.97

-1.23 0.987 -18.83

0.2176# 0.3237# 0.0000*

denotes significance at 1%; # not significant (significance higher than 10%).

International Food and Agribusiness Management Review

121

De Lima et al.

Volume 19, Issue 4, 2016

robust. For this reason, the endogenous variable ‘soybean price’ and all related instrumental variables were excluded from the specified model. It should be noted, that the focus of this research is to quantitatively describe the impact of the region (represented by fertilizer transportation corridors) on the fertilizer freight rate. The variables ‘soybean price,’ ‘fertilizer price,’ length of route’ and the binaries referring to ‘period of fertilizer transportation’ were only considered to make the model more robust and accurate. Analysis of residuals White’s test was applied to check for data heteroscedasticity. The null hypothesis that the variance is constant (homoscedasticity) was rejected at a 1% significance level, verifying the existence of heteroscedasticity for the initial estimated model. White’s robust correction was used because it adjusts the standard errors from model heteroscedasticity when, in practice, one does not know the pattern of heteroscedasticity. In addition, a successful corrective procedure (the natural logarithm) was used to modify the price of fertilizer transportation, the route length and the average price of diesel. White’s test was reapplied after these procedures, and the absence of heteroscedasticity was verified. The test resulted in a non-significant value (or significant at much higher than a 10% level) making it impossible to reject the null hypothesis that the residuals are homoscedastic. Table 7 shows the test statistic after correction procedures. Although not commonly used for cross section data, the Durbin Watson test was applied to check for the presence of residual autocorrelation (Gujarati, 2006), which was originally statistically calculated as 1.87. Inserting an autoregressive component freight rate (AR (1)), which was significant at 1%, resulted in a Durbin Watson test value of 2.05. The lower (dL) and higher (dU) limits found in the Durbin Watson Table of Critical Values (Gujarati, 2006) were 1.57 and 1.78, respectively. As the region between dU and (4-dU) is the region of no autocorrelation and the calculated value of 2.05 is within this region’s boundaries (1.78 and 2.22), the finding that there is no autocorrelation of residuals is confirmed. After assessing the variance inflation factor, the presence of multicollinearity between the estimated model’s explanatory variables was discarded. Following procedures proposed by Gujarati (2006), the calculated variance inflation factor values were below 10, as is shown in Table 8. Analysis of Figure 3A shows that there is no evidence that the errors are not following the normal distribution. The red line represents the normal. The higher the adhesion values of the series to the red line, the greater the evidence that the residuals’ distribution is normal. Figures 3B and 3C are histograms of residual values, with 3B being the frequency of the residual values in the 10 to 100 range and 3C the frequency of residual values in the 1000 to 8,000 range, the histogram limit. Table 7. White test result with research data from 2016.1 Observation × R-square

Prob. chi-square

21.10

0.9090*

1 After applying the continuous variables’ natural logarithm and using White’s robust correction; the prob. Chi-square is not significant

(significance higher than 10%).

Table 8. Values of variance inflation factor to evaluate multicollinearity between the explanatory variables with research data from 2016. Diesel price

Freight

1.25

3.25

2.36

International Food and Agribusiness Management Review

122

De Lima et al. B 100

1.00 0.75

0.00 -0.25 -0.50 -1.00 -4 -3 -2 -1 0 1 2 3 Quantities of Resid

60 40 20

-0.75 4

8 6 4 2 0 -2

7 Frequency (×1000)

0.25

0.50 Frequency

Quantities of Normal

Volume 19, Issue 4, 2016

-4

-2

6 5 4 2 0 -2 -4

4 3 2 1 0 -4

-2

2,500 5,000 7,500 10,000 12,500

Residual

Actual

Fitted

Figure 3. (A) Q-Q Normal Probability Plot of the residuals; (B) and (C) histograms of the residuals; (D) behavior of estimated series x compared with observed values. Based on research data from 2016. Figure 3D shows the fit of the regression values with the observed fertilizer freight rates (green and red lines) and with the regression residuals (blue line). The figure’s uppermost graphic indicates that the regression data are a good fit for most of the reporting period. The more the green line overlaps the red, the better adjusted the estimated equation is shown to be, i.e. the better its linkage with the observed data. The blue line in Figure 3D also represents the residuals and is used to show the difference between the observed series and adjusted series. The more this line ‘oscillates,’ the greater the evidence that what is not explained by the model is extremely random and of minimal importance. It can be seen in Figure 3B and 3C that the residuals have a normal distribution. Although the Jarque Bera test, which is an asymptotic normality test (Jarque and Bera, 1987), led to rejection of the null hypothesis of normality of errors (because the skew values=-8,839; kurtosis=282,109; sample size (n)=14,878; statistical JB=48,486,376 with a probability of zero significance, rejecting the null hypothesis). According to Oliveira (2014), even if this hypothesis is not validated, it is still possible to correctly infer when there is a large enough sample to bring the law of large numbers into play (Judge et al., 1988). Our study used a relatively large sample of 14,878 data points. It should be noted that routes were excluded when freight rates charged for shipments over these routes showed them to be discrepant (outliers). To determine if a route’s data were to be excluded, a simple arithmetic average of all the different routes fertilizer freight rates was calculated. If the freight rate for a route exceeded one standard deviation from the average, data for that route were disregarded. Data from economically relevant routes were not discarded. Relevancy was determined by frequency of use and volume of traffic. These specific values were not gathered for this research but were derived from the informed opinions of researchers, data provider groups, and market players. In addition, some of the data came from previously included contracted freight values, which justified some apparent dispersion Results The results from the final model’s multiple linear regression are presented in Table 9. The table shows the coefficient values and the respective t-statistic figures for each explanatory variable. Values for the variable Corridor C (Cor_C) and the January through February fertilizer transportation period (T3) are omitted from the table as that corridor and period were chosen to be control variables for specification purposes and for analysis of the model’s results. Table 9 shows that the F statistic has a fairly high value, indicating that at least one of the estimated coefficients is different from zero, which consequently indicates that at least one of the selected independent variables is significant. The significance of the F test was expected, since, at the least, the distance variable KM has a very clear relationship with freight rates (Corrêa Jr. and Caixeta-Filho, 2003). It is noteworthy that the coefficient of determination, R2, performed satisfactorily (0.62 approximately) indicating that the variables International Food and Agribusiness Management Review

123

De Lima et al.

Volume 19, Issue 4, 2016

Table 9. Coefficients estimation for the regression model’s explanatory variables with research data from 2016.1,2 Explanatory variable

Coefficients

t-statistic

Standard error

P-value

constant COR_A COR_B COR_D COR_E T1 T2 lnDIESEL lnKM AR(1)3 R-Squared adjusted R-squared observations F-statistic prob (F-statistic) Durbin-Watson

0.733722* 0.145553* 0.377956* 0.222336* 0.046231* -0.075630* -0.033489* 0.094049* 0.510943* 0.241275* 0.6157 0.6154 14,878 2,634.390* 0.0000 2.05

15.00025 19.52019 42.95309 15.81382 3.887272 -9.331273 -4.456568 4.121097 78.44784 7.320571

0.048914 0.007457 0.008799 0.014060 0.011893 0.008105 0.007515 0.022821 0.006513 0.032958

0.0000 0.0000 0.0000 0.0000 0.0001 0.0000 0.0000 0.0000 0.0000 0.0000

1 * denotes

significance at 1%.

2 The coefficients are valid for any monetary unit. The fertilizer freight price, the diesel price and the distance are in natural logarithm

from and result in elasticity values. 3 Insertion of the autoregressive term (AR (1)) to correct residual autocorrelation. Further interpretation of its coefficient was not conducted as its significance value was below 1% (Gujarati, 2006).

explain about 62% of the observed variation in freight rates. The R2 is equal to 0.62, which is considered a high R2 in economics and a sign that the model is properly designed. Relatively to the control Corridor (Cor_C) and keeping the other explanatory variables constant, fertilizer producers would find shipping through Cor_A to be 14.55% more expensive, shipping through Cor_B would be 37.79% more expensive, shipping through Cor_D would be 22.23% more expensive, and shipping through Cor_E would be 4.62% more expensive. Corridor C (origin at Cubatão and the ports of Santos and Guarujá with destinations in the grain producing regions of Brazil’s Midwest) was shown to be the lowest priced fertilizer transport corridor. This can be explained by the fact that this corridor is well positioned for fertilizer transport economically contracted as return freight due to the overabundance of grain being delivered from Brazil’s grain growing region to the two major ports. Relative to the fertilizer freights rates found during the ‘T3’ period (January thru February) and keeping the other explanatory variables constant, the freight rate for fertilizer transported in the T1 period (March thru May) would be 7.56% less and the freight rate for fertilizer transported in the T2 period (June thru December) would be 3.34% less (Table 9). The largest amount of fertilizer transported was during the grain harvest period (T1), followed by the grain planting period (T2). It seems reasonable that the fertilizer freight rate is lowest during the T1 period because this period sees the shipment of soybeans and soybean meal at its greatest; the number of trucks that need to be returned from ports to the Midwest’s vast producing region at its peak; and the demand for return freight is at its highest. Fertilizer transported in the T2 period will be less like to form part of a return freight scheme than in the T1 period as fewer soybeans are being transported to export facilities; however, soybeans are still being harvested and shipped during the period leaving the fertilizer as return freight option available, just to a lesser

International Food and Agribusiness Management Review

124

De Lima et al.

Volume 19, Issue 4, 2016

extent. Fertilizer freight rates are their highest in the T3 period. During this period, almost no soybeans are harvested or shipped and the fertilizer as return freight option is, for the most part, unavailable. Although the T2 period was the period with the highest frequency of actual route utilization (about 53%), attention should be paid to the fact that the number of routes utilized need not translate to amounts of product shipped as many routes traversed may not include large grain handling and export facilities or may offer only limited return freight opportunities. The T1 period has a lower value for route utilization than the T2 period, but during the T1 period more routes are likely to be carrying cargo as return freight from export facilities relative to the T2 period, which would translate to the use of fewer routes to transport large amounts of fertilizer. The estimated diesel coefficient, given by β9, is the elasticity of the fertilizer transportation price relative to the mean price of diesel. The coefficient’s value indicates that when the price of diesel (R$/liter) increases by 1%, the cost to transport fertilizer (R$/MT) increases by 0.094% if all other variables remain constant. The Department of Operating Costs, Technical, and Economic Studies of the Brazilian National Association of Transportation and Freight Logistics (DECOPE), which represents the Brazilian cargo transportation business, reported that the impact of a 1% change in diesel’s price could lead to increases of 0.1 to 0.34% in freight rates (NTC-Logistica, 2015). This minimal impact in 2014 may be related to slowdown in the Brazilian freight market that year. NTC-Logistica (2015) noted that there was excess shipping supply for most of 2014, causing some lag in freight pricing. Finally, the coefficient of the variable that measures distance traveled in km (β10) indicates that when the distance traveled (km) increases by 1%, the cost to transport fertilizer (R$/t) increases by 0.51% if all other variables remain constant. A large number of studies dealing with the structure of freight cost consider that distance transported is the main factor for determining transportation costs, regardless of the transportation mode employed. According to Correa Jr. and Caixeta-Filho (2001), increased shipping costs are an inherent result of increased distance transported as variable operating costs, such as fuel, oils, lubricants, and driver time, increase with an increase in distance traveled. Oliveira (1996) and Martins (1998) identified a close direct relationship between distance traveled and the cost of grain transport in the Brazilian state of Paraná. Figure 4 shows a direct relationship between distance transported and fertilizer transport costs. Although the relationship is not one to one, shipping cost tend to be higher the greater the distance transported. 300

Freight costs (R$/mt)

250 200 150 100 50 0

400

800

1,200 1,600 2,000 2,400 2,800 Distance (km)

Figure 4. Fertilizer freight costs (R$/MT) × distance (km) from research data of 2016. International Food and Agribusiness Management Review

125

De Lima et al.

Volume 19, Issue 4, 2016

6. Final considerations This study’s results show the effect of price variation in several shipping inputs on the final Brazilian fertilizer shipping costs paid by the agent that hired the transportation company and the relative cost difference when shipping fertilizer over different road transport corridors. The study also examined the shipping complex’s dynamic and noted the significance of return freight (backhauling) in the control of shipping costs. It was found that shipping fertilizer though transport corridors linking major Brazilian ports with grain growing regions showed a significant annual average cost advantage over transport corridors linking ports with inland fertilizer industries. This finding most likely indicates the positive benefits of fertilizer as return freight after grain delivery. Although it appears that the availability of return freight between ports and grain growing regions generates average annual transportation savings, there is no reason to assume this cost savings is spread evenly throughout the year. Fertilizer use and crop harvesting are not concurrent. Transportation price spikes often occur during peak harvest season when shippers must deal with vehicle shortages and transportation, loading, and unloading bottlenecks. It certainly appears that both grain and fertilizer storage facilities should be a priority for any large grain producer wanting to reduce transportation costs. Corridor C showed the lowest freight rate (R$) among the corridors and was used as the benchmark when comparing the different corridors’ freight rates. Routes in this corridor connect the ports of Santos and Guarujá and the fertilizer industries in Cubatão, 12 km from Santos, to Brazil’s most important grain growing region, the Midwest. The option to ship fertilizer as return freight through this corridor during the grain harvest should be readily available. Corridor E showed the second least expensive fertilizer freight rate, 4.62% above Corridor C’s. Corridor E is made up of routes between Brazilian fertilizer industries and the country’s large grain producing Midwest region. Interestingly, return freight should have had only a minor impact in lowering Corridor E’s freight rate, less than in all the other corridors, as none of the routes in corridor E pass near major soybean processing or direct international export facilities. It is assumed that Corridor E’s freight rate was greatly influenced by the transport of other products, which would have helped lower its freight rate in general. In addition, all the other corridors involve shipping to and from Brazilian ports. Freight rates in these corridors may have been negatively affected by logistical bottlenecks in the form of an overloaded transport system and congestion at the ports during the grain harvest, thereby reducing the impact of the return freight option relative to Corridor E. In contrast, Corridor B, which covers the routes connecting the port of Paranaguá with inland fertilizer industries, showed the highest fertilizer freight rate, 37.79% above control corridor C. The high freight rate may be due to a lack of the return freight option as this corridor is out of all grain shipping patterns, and there are relatively major infrastructure inadequacies throughout the corridor and at the port. Although the availability of return freight and infrastructure inequality are two justifiable rational for this rather large transportation cost divergence, there is certainly room for further study to better isolate its causes. Fertilizer freight rates in Corridors A and D were somewhat similar, 14.5 and 22.23% higher than corridor C, respectively. Corridor A connects the port of Paranaguá and fertilizer manufactures in the nearby city of Curitiba with cities in the grain producing region. Corridor D connects the ports of Santos and Guarujá and the fertilizer industries in Cubatão with inland fertilizer industries. The methodology applied to complete this study has been found to be valid and can be used to expand the analysis to other transport corridors, other modes of transportation, and other products. The methodology could be further developed through the use of a longer series, more current data, and the separation of freight rates by shipping company. In this context, further studies to capture the impact each company has International Food and Agribusiness Management Review

126

De Lima et al.

Volume 19, Issue 4, 2016

on generalized fertilizer freight rates could be carried out using the methodology applied in this study with the inclusion of fixed or random effects methods of analysis. Our study does have one rather significant caveat: since interactions between the demand and supply of grain transportation services are contingent on factors that may have different levels of effect from one year to the next, its results may not be applicable if future conditions, particularly climatic or economic conditions, change appreciably from those in effect during the study’s period. In order to more fully understand and manage the Brazilian transportation system’s cost structure, there is a need for further study to examine factors that have been barely touched on in this study. The more obvious of these factors involve logistical deficiencies, such as in the loading and offloading of goods at Brazilian export facilities and in the maintenance and design of Brazil’s roadway system. Other factors for future analyses could focus on the state of the Brazilian fleet of transport vehicles and the effect of highway privatizations on transportation costs. The Brazilian transportation system functions, goods move from one place to another, whether they are moved in a timely or efficient manner is a question still to be answered. By identifying the effect of various shipping variables on the cost to ship fertilizer, this study should help agricultural product producers and fertilizer manufactures plan the most economically advantageous product transportation schedule and determine the actual economic benefit of storage facility investments. The authors of this study hope that our results assist those who depend on the Brazilian transportation system control their transportation costs.

References ABIQUIM. 2009. A indústria química. Availabe at: http://www.abiquim.org.br. ANDA. 2014. Principais indicadores do setor de fertilizantes. Availabe at: http://tinyurl.com/jjb6kuz. Banco Central do Brasil. 2016. Availabe at: http://www.ipeadata.gov.br. Beilock, R., P. Boneva, G. Jotova, K. Kostadinova and D. Vassileva. 1996. Road conditions, border crossing and freight rates in Europe and Western Asia. Transportation Quarterly, 50: 79-90. Binkley, J.K. and B. Harrer. 1981. Major determinants of ocean freight rates for grains: an econometric analysis. American Journal of Agricultural Economics 63: 47- 57. Bowersox, D.J. and D.J. Closs. 2004. Logística empresarial: o processo de integração da cadeia de suprimento. Atlas, São Paulo, Brazil. CEPEA. 2016. PIB Agro CEPEA-USP/CNA. Availabe at: http://cepea.esalq.usp.br/pib. CONAB. 2016. Acompanhamento da Safra Brasileira. Availabe at: http://www.conab.gov.br. Corrêa Jr., G. 2001. Determinantes do preço do frete rodoviário para transporte de soja em grãos em diferentes regiões brasileiras: uma análise econométrica. Ph.D. diss., Escola Superior de Agricultura ‘Luiz de Queiroz’, Universidade de São Paulo, Piracicaba/SP, Brazil. Corrêa Jr., G. and J.V. Caixeta-Filho. 2003. Principais determinantes do preço do frete rodoviário para o transporte de soja em grãos em diferentes estados brasileiros: uma análise econométrica. Economia Aplicada 1: 89-211. CSCMP. 2016. Supply chain management definitions and glossary. Availabe at: https://cscmp.org/supplychain-management-definitions. Garrido, R.A. and H.S. Mahmassani. 2000. Forecasting freight transportation demand with the space-time multimodal probit model. Transportation Research Part B: methodological 34: 403-418. Gujarati, D.N. 2006. Econometria Básica. Campus Elsevier, São Paulo, Brazil Hauser, R.J. 1986. Competitive forces in the U.S. inland grain transportation industry: a regional perspective. Logistics and Transportation Review 22: 158-183. Hoffmann, R. 2015. Análise de regressão: uma introdução à econometria. Availabe at: http://www.producao. usp.br/handle/BDPI/48616. IBGE. 2012. Produção agrícola municipal (PAM). Availabe at: http://www.ibge.gov.br/home. ILOS. 2016. Custos logísticos. Availabe at: http://tinyurl.com/jhq62yk. International Food and Agribusiness Management Review

127

De Lima et al.

Volume 19, Issue 4, 2016

Izumi, A.S. 2012. Caracterização de comparação da construção de um armazém ‘dentro da porteira’ e ‘fora da porteira’ no estado de Goiás. ESALQ-LOG. Availabe at: http://tinyurl.com/jujb5mh. Jarque, C.M. and A.K. Bera. 1987. A test for normality of observations and regression residuals. International Statistical Review 55: 163-172. Judge, G.G., R.C. Hill, W.E. Griffiths, H. Lutkepohl, T.C. Lee. 1988. Introduction to the theory and practice of econometrics. John Wiley and Sons, Toronto, Canada. Kerr, J.D. 1972. Least square analysis of freight-rate anomalies. Australian Journal of Statistics 14: 63-67. Koutsoyiannis, A. 1972. Theory of econometrics. 2nd ed. Palgrave, Ontario, Canada. Lima, M.P. 2001. Custeio do transporte rodoviário. Ogerente. Availabe at: http://tinyurl.com/zskpmll. Ministério da Agricultura, Pecuária e Abastecimento (MAPA). 2016. Estatísticas. Availabe at: http://www. agricultura.gov.br/vegetal/estatisticas. Martins, R.S. 1998. Racionalização da infra-estrutura de transporte no estado do Paraná: o desenvolvimento e a contribuição das ferrovias para a movimentação de grãos e farelo de soja. Ph.D. diss., Escola Superior de Agricultura ‘Luiz de Queiroz’, Universidade de São Paulo, Piracicab/SP, Brazil. Michelon, E.R.S. 2007. A utilização de carga de retorno no transporte de soja: características, dificuldades e vantagens. ESALQ-LOG. Availabe at: http://tinyurl.com/zaews7e. Novaes, A.G.N. 2007. Logística e gerenciamento da cadeia de distribuição: estratégia, operação e avaliação. 3rd ed. Elsevier, Rio de Janeiro, Brazil. NTC-Logistica. 2015. Anuario-2014/2015. Availabe at: http://www.portalntc.org.br. Oliveira, A.L.R. 2011. O sistema logístico e os impactos da segregação dos grãos diferenciados: desafios para o agronegócio brasileiro. PhD. diss., Instituto de Economia (IE), Universidade de Campinas (UNICAMP), Campinas, Brazil. Oliveira, J.C.V. 1996. Análise do transporte de soja, milho e farelo de soja na hidrovia Tietê- Paraná. Ph.D. diss., Escola Superior de Agricultura ‘Luiz de Queiroz’, Universidade de São Paulo, Piracicaba/SP, Brazil. Oliveira, M.T. 2014. Distância psíquica e seus efeitos sobre o fluxo de exportações dos estados brasileiros. Ph.D. diss., Universidade de Coimbra (UC), Coimbra, Portugal. Oliveira, C.F, M J. Rosa and J.V. Caixeta-Filho. 2010. Estimativa da oferta de fertilizantes como carga de retorno no ambiente portuário brasileiro entre 2005 e 2009. Informações econômicas 40: 1-9. Prentice, B.E. and D. Benell. 1992. Determinants of empty returns by U.S. refrigerated trucks: conjoint analysis approach. Canadian Journal of Agricultural Economics 40: 109-127. Relação Anual de Informações Sociais (RAIS). 2012. Do Ministério do Trabalho e Emprego. Availabe at: http://www.rais.gov.br. Ribeiro, S., F.H. Mansano, A.H. Gameiro and R.L. Lopes. 2009. Custo do Transporte como Ferramenta de Gerenciamento Logístico para a Soja: o Caso da Rota Maringá – Paranaguá. Revista ADM.MADE 13: 87-100. Rodrigues, P.R.A. 2007. Introdução aos sistemas de transporte no Brasil e à logística internacional. Aduaneiras, São Paulo, Brazil. Scherer, A.A. and R.S. Martins. 2004. Atributos da prestação de serviços para operações logísticas de commodities agrícolas na visão dos embarcadores. Anais eletrônicos. Congresso Internacional de Pesquisa em Logística. Availabe at: http://ageconsearch.umn.edu/handle/56818. SECEX. 2016. Sistema de Análise das Informações de Comércio Exterior. Availabe at: http://aliceweb. desenvolvimento.gov.br. Steadieseifi, M., N.P. Dellaert, W. Nuijten, T.V. Woensel and R. Raoufi. 2014. Multimodal freight transportation planning: a literature review. European Journal of Operational Research 233: 1-15. Tavares, M.F.F. and C. Haberli Jr. 2011. O mercado de fertilizantes no Brasil e as influências mundiais. Central de Cases. Availabe at: http://tinyurl.com/z6m72r9. Teixeira, L.S. 2013. Caracterização dos fluxos de fertilizantes no Brasil. ESALQ-LOG. Availabe at: http:// tinyurl.com/hlfh5e2. Thompson, L. 1960. Freight rate equations. Industrial and Engineering Chemistry 52: 40A-44A. USDA. 2016. Crop production. Availabe at: http://usda.mannlib.cornell.edu. Wooldridge, J.M. 2010. Introdução a econometria: uma abordagem moderna. Cengage Learning, São Paulo, Brazil. International Food and Agribusiness Management Review

128

OPEN ACCESS International Food and Agribusiness Management Review Volume 19 Issue 4, 2016; DOI: 10.22434/IFAMR2015.0095 Received: 30 June 2015 / Accepted: 10 October 2016

Consumer perceptions of climate change and willingness to pay for mandatory implementation of low carbon labels: the case of South Korea RESEARCH ARTICLE Hyeyoung Kima, Lisa A. Houseb, and Tae-Kyun Kim aAssistant

Research Scientist and bProfessor, Food and Resource Economics Department, University of Florida, P.O. Box 110240, Gainesville, FL 32611, USA

cProfessor,

Department of Agricultural Economics, Kyungpook National University, 80 Daehakro, Bukgu, Daegu 41566, Republic of Korea

Abstract The purpose of this study is to examine consumer values for mandatory carbon labels incorporating South Korean consumers’ perceptions about climate change using conjoint analysis. In a face-to-face consumer survey, we asked about individuals’ perceptions of the impact of climate change on their personal lives to measure its effect on consumer preference for carbon labels. The results of ordered logit and conditional logit regressions showed that a significant preference for mandatory carbon labels reflected Koreans’ level of concern about climate change. As an increasing number of consumers feel the impact of climate change, the gap of willingness to pay between voluntary and mandatory low carbon labels is significant. Also, consumer perception of the impact of climate change on their personal lives was significantly influenced by the area in which the respondents’ lived. Keywords: carbon labeling, climate change, mandatory labels, awareness, WTP JEL code: D12, P36, Q54 Corresponding author: tkkim@knu.ac.kr

129

Kim et al.

Volume 19, Issue 4, 2016

1. Introduction Climate change due to high levels of greenhouse gas (GHG) emissions in the atmosphere has become an important issue in the world. Global warming threatens to raise sea levels, exterminate species, and threaten food security (Cox et al., 2000; Vitousek, 1994). In response, more than 160 international parties adopted the United Nations Framework Convention on Climate Change (UNFCCC) in 1992 as the global legal policy framework and agreed to the Kyoto Protocol in 1997 as realistic rules for implementation. More recently, the Intergovernmental Panel on Climate Change (IPCC) pointed out human activities as one of the main causes for global GHG emissions, especially industrial activities (IPCC et al., 2007). In response to this, countries, retail chains, and non-government organizations in the world have established carbon labels to inform consumers of the environmental impact of the products they consume. The main role of product labels is to turn ‘credence’ attributes into ‘search’ attributes. This new information may influence consumers when they make product decisions if consumers perceive the information as important. For example, nutritional labels represent the nutrient content of food and can influence consumers’ food purchases because of their perception of the health of the product based on the label information (Chang and Nayga, 2011; Kim et al., 2012; Teisl and Levy, 1997). Consumers who purchase products with carbon labels may obtain utility in terms of both public satisfaction through participating in the reduction of carbon emissions and reduction of global warming, as well as private satisfaction through consuming clean air in their daily lives (Cohen and Vandenbergh, 2012; Michaud et al., 2013). In some cases, low carbon products such as cars and appliances may directly increase the private utility of consumers by reducing their spending on gas or electricity. Carbon labels are voluntarily adopted in most countries. Gadema and Oglethorpe (2011) pointed out the limited effect of voluntary systems on the reduction of carbon emission. Although producers or retailers may be interested in the label, this will likely only occur if the label directly or indirectly relates to increasing profit. For example, Tesco, a retailer in the U.K. reported that they would stop presenting carbon labels because of the cost of maintaining the labels (Quinn, 2012). In South Korea, 13.6% of companies indicated that they had a program to reduce carbon emission or had a plan to do it (Ministry of Environment, 2007, 2008). However, most companies develop the program not to provide environmental information of the product to consumers but to meet the standards of importing countries for exporting their products to the counties. Because of growing interest in this issue, many researchers have attempted to understand consumer perception of carbon labeling and attitudes about climate change (Kemp et al., 2010; Kim, 2011; Upham et al., 2011). However, the studies have neglected to develop a linkage between consumers’ perceptions about climate change and consumer value for the carbon labels. In addition, few studies have measured the value of mandatory carbon labels, although previous studies established a consensus that voluntary carbon labels were not very effective at reducing GHG emission (Cohen and Vandenbergh, 2012; Gadema and Oglethorpe, 2011). Note that mandatory labels require companies to produce products with reduced carbon emissions by law. Therefore, this study aims to further the literature regarding Korean consumers, as well as by investigating the difference in consumer values between mandatory and voluntary labels by incorporating individual perceptions about the impact of climate change on their personal lives. In this study, a choice experiment (CE) analysis was used to estimate the value of carbon labels depending on levels (measured and low carbon labels) and types of implementation (voluntary and mandatory). We expected consumer preference for carbon labels to vary based on their individual attitudes toward climate change. The impact of climate change on individuals’ lives was measured to predict the effect of individuals’ attitudes toward climate change on preference. Consumers who strongly perceive the impact of climate change may be willing to pay extra to purchase products with low carbon labels and may prefer mandatory carbon labels.

International Food and Agribusiness Management Review

130

Kim et al.

Volume 19, Issue 4, 2016

We selected fresh apples to derive policy implications of availability of low carbon production in agricultural products. We selected this product because apples are one of top three fresh fruits consumed in South Korea and they can be purchased year round. Even though the Ministry of Environment, who is administrating the carbon labeling, stated that basic agricultural products (in their original form or having undergone only primary processing) are not targeted for carbon labeling, the Ministry of Agricultural, Food and Rural Affairs have conducted several pilot projects for basic agricultural products to test implementation of low carbon labels since 2012. The following sections will present an overview of carbon labels, examine the measured values of carbon labels according to previous studies, and present the research method, findings, and conclusions of this study.

2. Overview of carbon labels Global carbon labels Introducing carbon labeling has symbolic meaning in developing world economies which implies that international leaders seek not only quantitative growth but qualitative growth of the world economy by considering global environment change. Unlike Eco Labels, which provide qualitative emission information of products, most carbon labels are designed to show quantitative emissions of carbon or GHG equivalent while a product is grown, manufactured, transported, used, and disposed. Despite controversial problems related to interpretation of the numerical values on carbon labels (Upham et al., 2011), carbon emissions play a role in informing consumers of the environmental implications of their economic activities. Selected carbon labels are shown in Table 1. The first carbon labeling in the world was the Carbon Footprint label created by the Carbon Trust in the UK in 2006. The Carbon Trust offered two types of Carbon Footprint labels, Reducing CO2 Label and CO2 Measured Label. The Reducing CO2 Label certified that companies committed to reduce the level of CO2 emissions resulting from the production and distribution of the products. The Carbon Trust required re-certification of the Reducing CO2 Label every two years. To recertify, companies must prove that they have reduced the amount of CO2 emissions. The CO2 Measured Label only indicates that the footprint of the products is accurately measured. Both certifications must meet the requirements in the PAS 2050 and/ or the WBCSD-WRI GHG Protocol Product Standard. Supermarket channels in European countries created carbon labels to inform consumers of the environmental impact of the products in response to global trends. Carbon labels may also promote a positive image of the supermarket by showing their desire to take care of the environment. In France, Casino initiated a carbon labeling program called Indice Carbone in 2011 which provided quantitative CO2 emissions, recycling Table 1. Selected carbon labels by country. Carbon labels by country Reducing CO2 Label

Indice Carbone

Climatop

Certified CarbonFreeâ&#x201E;˘

CFP mark

Country Label

France

Swiss

USA

Japan

Web source

carbontrust.com

produits-casino.fr

climatop.ch

carbonfund.org

cfp-japan.jp

International Food and Agribusiness Management Review

131

Kim et al.

Volume 19, Issue 4, 2016

information and additional information about the environmental impact of use and disposal of products. In Switzerland, Migros introduced a carbon label called Climatop comparing carbon emissions to that of similar items. Products displaying Climatop indicate that the product’s emissions are 20% lower those of its counterparts within the same product category. In the USA, Carbonfund.org, a nonprofit provider of climate solutions, created a label called Certified CarbonFree in 2007. Products obtain the CarbonFree label when they meet the standards of PAS 2050:2008, ISO Standard 14044:2006 or WBCSD-WRI Greenhouse Gas Protocol for corporate GHG reporting. Also, products are qualified for the CarbonFree product certification program as long as the products have received the Carbon Trust and the Carbon Pollution Reduction Scheme developed by the Australian Government. Aside from CarbonFree, the Energy Star label provides energy efficiency information for appliances. Murray and Mills (2011) estimated that Energy Star appliances are associated with carbon emission reductions of about 1.1 million metric tons per year. In Japan, the Japan Environmental Management Association for Industry (JEMAI) started new carbon footprint of products (CFP) programs based on ISO 14067 in 2012. The main features of CFP programs are CFP-product category rule certification, CFP verification, and verification of Emission Factors conducted by third party experts. Carbon labels in South Korea South Korea’s GHG emissions accounted for 1.3% of the world total in 2005. However, South Korea’s emissions increased 71.6% on a per capita basis over the period of 1990 to 2005 which is far outstripping the OECD average of 2.1% (OECD, 2010). In 2008, the president proclaimed ‘low carbon/green growth’ as the vision to guide Korea’s development during the next 50 years, and in 2009, South Korea voluntarily set a goal of cutting its GHG emissions by 30% relative to a ‘business as usual’ baseline by 2020 (OECD, 2010). In line with the national strategy for Green Growth, the government has encouraged firms to develop green technologies and is also aware of the importance of consumer participation. As such, South Korea introduced a carbon labeling system for daily household supplies and home appliances in February 2009 in order to reduce GHG emissions by leading consumers to low carbon products and by encouraging companies to develop new technology which can reduce emission levels. Under the administration of the Ministry of Environment, the Korea Environmental Industry and Technology Institute (KEITI) issues two different levels of carbon labeling: carbon emission certification (CEC) (Figure 1A) and low carbon product certification (LCPC) (Figure 1B). The original labels are translated to English in Figure 1. Similar to the CO2 Measured Label of the Carbon Trust, the CEC is issued if products are officially examined for emission levels and meet standard GHG emission levels. The LCPC is issued for a product which already obtained the CEC if the company successfully develops techniques to reduce a certain amount of GHG emissions to produce that product. The LCPC is similar to the Reducing CO2 Label of the Carbon Trust. Currently, 371 products obtained LCPC out of 1,640 total certified products (Korea Environmental Industry and

Figure 1. Translated carbon labels in South Korea. (A) carbon emission certification; (B) low-carbon product certification (adapted from Korea Environmental Industry and Technology Institute: http://tinyurl.com/zd8pubz).

International Food and Agribusiness Management Review

132

Kim et al.

Volume 19, Issue 4, 2016

Technology Institute: www.edp.or.kr). This indicates that companies are less likely to invest in developing new technology to reduce GHG emissions.

3. Measuring values of carbon labels CEs have widely been used to measure consumer willingness to pay (WTP) for attributes of products. In particular, CEs have been used to examine values of different types of labels such as nutritional labels, health labels, ingredient labels, GMO labels, country of origin labels and food mileage labels. Although Lusk and Schroeder (2004) pointed out the hypothetical bias in CEs, CEs are relatively cost effective and enable large coverage compared to experimental auctions, which may reduce hypothetical bias. Also, Lusk and Schroeder did not find a statistical difference of marginal WTPs between CEs and experimental auctions. Studies based on consumer surveys have showed that consumers positively valued low carbon emissions and had higher WTP for low carbon products than high carbon products. Many studies have focused on measuring WTP for non-agricultural products such as green power (Borchers et al., 2007; Clark et al., 2003), appliances (Sammer and Wüstenhagen, 2010) and automobiles (Hidrue et al., 2011), while several studies have examined the value of low carbon emission agricultural products. Michaud et al. (2013) examined WTP for roses to determine consumer values of carbon labels for non-food agricultural products. They conducted a discrete CE with real purchases of roses associated with an ecolabel and a carbon footprint label in France. The results of the mixed logit model indicated that consumers significantly valued environmental attributes and the value of a low-carbon footprint was considerably greater than eco-labels (low fertilizer). The premium for roses with a low-carbon footprint was approximately 2.4 times larger than eco-labeled roses. Aoki and Akai (2013) and Loureiro et al. (2002) used food products to compare WTP for environment friendly foods. Aoki and Akai performed a real CE to compare consumer WTP for the reduction of CO2 emissions for Satsuma mandarin oranges based on consumer attitudes toward the environment in Japan. They used three environmental factors: environmental consciousness (EC), environmental knowledge (EK) and environmental behavior (EB) in daily life. Consumers were categorized into two groups with high and low attitudes of EC, EK and EB. The results of the random parameter logit model indicated that only environmental consciousness led to significant differences in respondents’ purchase behavior of oranges based on carbon emission levels. Consumers belonging to a high EC group were willing to pay over 2.2 times higher than consumers in a low EC group for the reduction of 1g of CO2 emissions per orange. Loureiro et al. (2002) measured American consumers’ WTP for eco labeled apples using a double-bounded logit model. They used the eco-label certified by The Food Alliance, a non-profit third-party certifying organization based in Portland, Oregon. However, the mean premium for eco-labels was low, only 5% of market prices. This study indicates the importance of consumer’s recognition of the labels. As presented at www.ecolabelindex.com, more than 150 institutes certify eco-labels on food. Although many eco-labels have been developed, only a few were recognized. Low accessibility to labels may lead consumers to place a low value on this information. In South Korea, to the best of our knowledge, few studies have measured WTP for carbon labels on foods. According to survey results by KEITI, only one third of consumers in Korea indicated willingness to purchase products with carbon labels when the price was 5% higher than market prices (KEITI, 2009). Therefore, this study will measure Korean consumers’ WTP for fresh apples produced ecologically. In particular, this study aims to measure the value of mandatory implementation of carbon labels.

International Food and Agribusiness Management Review

133

Kim et al.

Volume 19, Issue 4, 2016

4. Study framework and methodology A framework of climate change perceptions and carbon labeling preferences Previous carbon labeling studies focused on understanding consumers’ attitudes and perceptions about climate change and measured WTP for carbon labels. This study asked, in what circumstance do consumers perceive GHG emission as an important attribute? The study tested whether consumers who perceive the effect of climate change on their lives would place greater value on the information on carbon labels. Connections between the two sectors will provide quantitative information about the value of GHG emission according to qualitative personal perceptions. In a conditional logit model, invariant variables such as demographics and attitudes are implemented as interaction terms with attributes of the choices to avoid constant variations within the group. The number of estimated parameters is determined depending on the size of invariant variables and the number of attributes of the choices. As invariant variables and attributes that can be interacted with the invariant variables become larger, the model loses a degree of freedom. If an interesting invariant variable is not exogenous, the interaction terms violate the assumption that covariates are exogenous in the population. The instrument variable (IV) estimate is an appropriate method to figure the issues in general applications. However, the IV estimate is complicated to use when there are multiple endogenous variables from interactions with an endogenous invariant variable. Also, the duplications of observations lead to invalid standard errors of the IV estimation. Therefore, we used two-step approaches using ordered logit models and conditional logit models, as described in Figure 2. Figure 2 frames the modeling of individuals’ perceptions of climate change and apple purchases based on various grades, prices and carbon labels. The probability that respondents agreed or strongly agreed that climate change influenced their personal lives was calculated based on the switched signs of marginal probability and the differences of marginal effect. The predicted probability was included in the apple purchase model as interactions with carbon labels. The interaction terms indicate preferences for carbon labels of individuals who perceived the impact of climate change on their personal lives.

• Causes of climate change • Demographics • Living area

What apples will you purchase?

Conditional logit model Climate change influence my life Ordered logit model Attributes of apple • Grades • Prices • Carbon labels

1=Strongly disagreed 2=Disagreed 3=Neither disagreed nor agreed 4=Agreed 5=Strongly agreed

Interactions Carbon labels × prob (≥4) Prob (≥4)

Figure 2. Modeling perception of climate change and apple purchases.

International Food and Agribusiness Management Review

134

Kim et al.

Volume 19, Issue 4, 2016

Application of ordered logit model for climate changes and personal lives Participant perceptions about the impact of climate change on their personal lives were measured with 5-point scales (1 was strongly disagree and 5 was strongly agree, J=5). For convenience, define W as respondents’ characteristics such as age, gender, income, education, residency (Seoul or not) and attitudes about human activity being the cause of climate change. From the estimated ordered logit model, we will have (J-1) unknown thresholds (αj) and regression parameters (β) associated with W. J

J-1

∑ α + W’β + ε; Y = ∑ j I (α j=1

j=1

j-1

< Y* ≤ αj) (1)

where Y* is the latent variable measuring individuals’ perceived impact of climate change on their personal lives, and Y is the indicated ordered response. The thresholds αj are cut-points on the latent variable used to differentiate changing points given that all the predictor variables are set at zero. The sign of the regression parameter β can be immediately interpreted as determining whether or not the latent variable increases with the regressor. The corresponding estimated thresholds (αj) and linear combination of between estimated parameters and variables (W’β) enter the ordered logit model with j indicating the five perception levels to obtain predicted probability: Prob (Y ≤ j|W) =

exp (αj + W’β) (2) 1 + exp (αj + W’β)

Prob (Y > j|W) = 1 – Prob (Y ≤ j|W) (3) Applications of conditional logit model for effect of climate change on carbon labels The experiments are based on random utility theory and are consistent with Lancaster’s theory of utility maximization which states that consumers demand attributes embodied in a good (Louviere et al., 2000). Let Uik be the ith individual’s utility of choosing kth alternative. The total utility can be divided into two components of a systematic component, Vik, and a random component, εik: Vik + εik. To indicate invariant variables and attributes of choice, the system component can be expressed: Vik = α’Χik + βk’Zi where Χik is a vector of alternative attributes for individual i and Zi is a vector of interaction terms between product attributes and individual characteristics. A conditional logit model was estimated using CE information. Assuming the random component is independently identically distributed with type I extreme value (Gumble) distribution leading to a logit model formulation (McFadden, 1974). The probability of consumer i choosing alternative k out of K options is: Prob (i=k) =

exp(Vik)

∑ exp(Vik)

(4)

k=1

The systematic component includes five dummy variables, superior quality, good quality, CEC, LCPC and mandatory low-carbon certification (MLCC) of product attributes and prices. We set marketable quality and no carbon labeling as baselines. Also, the model includes three interactions between individuals’ perception of the impact of climate change on their personal lives and three carbon labels. Based on signs and degree of marginal effect of each level, we used the predicted probability of agreeing and strongly agreeing to the statement as an indicator of respondents’ attitudes about climate change. The systematic component model also includes a constant specific to the alternative ‘None of these’ to capture the average effect on utility when consumers do not purchase. The model was estimated using maximum likelihood estimation. From

International Food and Agribusiness Management Review

135

Kim et al.

Volume 19, Issue 4, 2016

estimated parameters, consumers’ WTP was calculated with the negative of the ratio of attribute coefficients to price coefficients indicating the marginal rates of substitution (Louviere et al., 2000).

5. Consumer survey of climate change and carbon labels A consumer survey was conducted with primary shoppers in households living in urban areas in South Korea. The survey questionnaire was designed for face-to-face interviews and consisted of three parts. In the first part, participants were asked to indicate their perception and attitude toward climate change. In the second part, participants indicated their socio-economics, gender, age, level of education attained, income and living area. In the third part, participants were presented a panel explaining the definition and purpose of carbon labeling, including the current voluntary system and what the mandatory system could be. The panel did not describe other benefits of the mandatory carbon labeling system. Participants’ self-evaluation of the benefits of a mandatory system will be reflected on purchase decisions in CE questions. Lastly, participants faced six CE questions. We designed the CE using three attributes of fresh apples: price, quality and carbon labeling to measure WTP for apples. Four price levels represent retail prices for 5 kg of Fuji apples (the most popular size) depending on the quality: ₩ 20,000, ₩ 30,000, ₩ 40,000 and ₩ 50,0001. South Korea grades quality standards of fresh apples with three classes depending on external (size, shape, color, etc.) and internal (sweetness, juiciness, etc.) attributes: superior, good and marketable quality. A market price range of good quality apples was from ₩ 25,000 to 46,000 in 2012, based on the price information provided by the Korea Agro-Fisheries and Food Trade Corp. Two levels of current carbon labeling (CEC and LCPC) and a MLCC were included in the choice experimental design, which also included a no carbon labeling baseline. A large number of hypothetical apples could be constructed to make two alternatives using three attributes and their various levels. Based on a D-efficiency criterion, 144 alternatives provided 100% efficient design. Despite the high efficiency, the number of designs will create many different versions of survey questions. The number of choice sets is related to quality of information (Johnson and Orme, 1996) and participants’ burden to respond. Optimal profiles of the attributes were drawn using an orthogonal design, from which 36 profiles were obtained, which achieved a D-efficiency score of 98.6. Although this number is significantly reduced, the 36 profiles are still a large number to evaluate because the respondents also answer other questions2. The 36 profiles were randomly sorted into six blocks, as shown in Supplementary Table S1. Six versions (one with each block) were randomly distributed to respondents. In each choice set, respondents will select one of two hypothetical apples or select a ‘None of these’ option, as shown in Table 2. This feature ensured that respondents were never forced to purchase an apple. Following the pre-test, respondents spent 15 to 20 minutes to complete the final questionnaire. The survey was conducted by trained interviewers in Daegu, which is the third largest city of six metropolitan cities in Korea, and in Seoul, which is the capital and the biggest city in Korea, in August 2012. We recruited 1 The

average exchange rate in August of 2012 was used to convert the unit (₩ 1,133 = 1 US$). (2006) indicated the importance of appropriate choice tasks to obtain reliable answers. Johnson and Orme (1996) found no evidence of increasing random error within about 20 choice tasks. 2 Orme

Table 2. A translated question example from the choice experiment. Grade Carbon labels Price I would choose…

Apple 1

Apple 2

good mandatory low carbon certification ₩ 40,000/5 kg

superior Carbon emission certification ₩ 30,000/5 kg

□

International Food and Agribusiness Management Review

136

none of these

□

Kim et al.

Volume 19, Issue 4, 2016

eight interviewers who were college students with experience in consumer surveys and trained them to understand the purpose of the study, interview skills about choice experimental questions and to explain the carbon labeling system. A pretest with trained interviewers was conducted for two days, August 23 and 24, 2012. After pretest, the questionnaire and survey procedures were finalized. The interview was conducted in randomly selected grocery stores and respondents were also randomly selected from the selected grocery stores. Of 280 consumers initially requested to participate the survey, we obtained 186 valid observations providing a 6.9% of margin of error under 5% significance levels using stratified sampling without replacement.

6. Sample descriptions and attitudes toward climate changes A summary of respondents’ demographic characteristics are shown in Table 3 alongside census data. Average age of respondents was very close to the census, while respondents were more likely to be female, educated and residents of Seoul compared to the census. Since the target samples were primary shoppers in households, the high percentage of female respondents was acceptable. To improve representation of certain groups in the population, we used post-stratification weights to adjust the distribution gaps of education and residency. Respondents’ average attitudes toward climate change are shown in Table 4. The majority of respondents felt the effect of climate change in their lives. Over 57% of respondents agreed (strongly agree or agree) that climate change affected their personal lives. This implies that climate change is not only a national issue, but a private issue. Over 60% of respondents indicated a major cause of climate change is human activity, compared to natural causes. Many individuals indicated willingness to change their behavior and activities to reduce climate change (about 78%). However, over 60% of respondents pointed out that there are many external factors which make their effort difficult. Overall, approximately 90% of respondents agreed that efforts to reduce climate change are very urgently required and 91% indicated that they strongly concerned or concerned climate change. This finding is very similar to the study result from the Ministry of Environment that has showed over 90% of consumers were concerned climate change and perceived the conditions as serious (Ministry of Environment, 2007, 2008)3. To reduce climate change, about 87% of respondents agreed to reduce consumption of products which cause environmental pollution and about 55% of respondents indicated that they would pay more for products which reduce climate change.

3 We calculated statistical power using the answers of respondents’ concerns about climate change compared to the result of national study conducted

by the Ministry of Environment (2007) in order to alleviate concerns about the sample size. The power of the study is 0.9656 indicating that the probability of rejecting the null hypothesis while the alterative hypothesis is true was over 96% given the sample size.

Table 3. Sample descriptive statistics and variable description. Variable

Variable description and code1

Sample n=186

average age gender (%) income (%)

respondent age 1 if male 1 if m onthly household income was ₩ 4 million and over education (%) 1 if respondents completed 2 yr or 4 yr college Seoul (%) 1 if respondents live in Seoul

Seoul n=107

South Korea Census2

39.9 32.3 51.6

42.2 34.6 53.3

36.7 29.1 49.4

38.9 50.1 40.0

75.8 57.5

71.0 100.0

82.3 0.0

51.8 43.53

1 Alternative

code for the dummy variables, gender, income, education and Seoul is ‘0’. census statistics are based on 2012. 3 Percentage of population living in Seoul of seven metropolitan cities in 2012. 2 All

International Food and Agribusiness Management Review

137

Daegu n=79

Kim et al.

Volume 19, Issue 4, 2016

Table 4. Attitudes about climate change. Variable

Variable description and code1

climate change

5 if respondents strongly agreed that climate change influence my daily life 4 if respondents agreed 3 if respondents neither agreed nor disagreed 2 if r espondents disagreed 1 if respondents strongly disagreed 1 if respondents believed a major cause of climate change is human activity not natural causes change 1 if respondents will change their behavior and activities to reduce climate changes. 1 if r espondents perceived that there are extra factors which make their effort difficult. 1 if the effort to reduce climate change is urgent. 1 if I will reduce consumption of products which cause environmental pollution to reduce climate change. 1 if I will pay more to purchase products which reduce climate changes. 1 if r espondents strongly concern or concern about climate changes

human

behavior extra factors effort consumption

pay more concerns about climate change 1 Alternative

Sample (%) Seoul (%) n=186 n=107

Daegu (%) n=79

6.5

9.4

2.5

51.1 25.3 16.1 1.1 60.2

58.9 28.0 3.7 0.0 66.4

40.5 21.5 32.9 2.5 51.9

77.9

85.9

67.1

62.9

65.4

59.5

88.7 88.7

91.6 90.7

84.8 86.1

56.9

56.1

58.2

90.9

92.5

88.6

code for the dummy variables is ‘0’.

7. Estimated results and discussions The results of the estimated ordered logit models performed with survey data in Stata® 13 (StataCorp, College Station, TX, USA) are shown in Table 5. Looking at differences among demographics, perceptions of the impact of climate change on participants’ personal lives were not significantly different by age and education, but the perceptions were significantly different by gender and income. Male respondents were less likely to feel the impact of climate change on their lives compared to female respondents. Respondents who earned over the average national household income were more likely to feel the impact of climate change on their lives. Aside from demographics, respondents’ living area was significantly different from zero at the 5% level. Participants living in Seoul were more likely to perceive the impact of climate change on their lives. This result suggests that people living in Seoul were influenced by environmental concerns and is also consistent with the different levels of environment issues by region (Lee et al., 2011). Although respondents who believed human activities are a major source of climate change were more likely to perceive the impact of climate change on their lives, the estimated parameter was not significant. The result indicates that causes of climate changes are not major factors explaining perceptions of the impact of climate change on respondents’ personal lives. Marginal effects of variables are shown in the last five columns of Table 5. The marginal effects indicate the change of probability for a particular answer given a one unit change of a covariate. For the statement of the impact of climate change on their personal lives, male respondents were 10.9% more likely to disagree than females, and respondents living in Seoul were 18.8% less likely to disagree than respondents living in Daegu. The signs of marginal effects changed at the level that respondents agreed to the statement that climate change influences their personal lives (j=4). In addition, the biggest difference in marginal effect occurred between the third and fourth level. Respondents living in Seoul were 33% (sum of marginal probability of ‘agree’ International Food and Agribusiness Management Review

138

Kim et al.

Volume 19, Issue 4, 2016

Table 5. Estimated results of likelihood of climate change impacting daily lives.1 Coefficient age gender income education Seoul human cut1 cut2 cut3 cut4 n F(6, 180) 1 **

0.027 -0.789** 0.555* -0.540 1.380** 0.484 -3.474** 0.078 1.419 4.763**

Std. err. 0.017 0.347 0.331 0.401 0.349 0.386 1.058 0.864 0.901 1.004

Marginal effect P (Y=1)

P (Y=2)

P (Y=3)

P (Y=4)

P (Y=5)

0.000 0.005 -0.003 0.002 -0.009 -0.003 –

-0.003 0.109** -0.070 0.062 -0.188** -0.063

-0.003 0.079** -0.060* 0.062 -0.131** -0.052

0.005 -0.160** 0.109* -0.098 0.267** 0.097

0.001 -0.032** 0.025 -0.028 0.061** 0.021 –

186 6.26 (P-value<0.05)

and * indicate that the coefficients are significantly different from zero at 5% and 10% levels, respectively.

and ‘strongly agree’) more likely to perceive the impact of climate change on their lives than respondents living in Daegu. Also, male respondents were 19% (sum of marginal probability of ‘agree’ and ‘strongly agree’) less likely to perceive the impact than female respondents. We calculated the predicted probability of individuals’ perception of the impact of climate change on their personal lives for agreeing or strongly agreeing to the statement, Prob (Y≥4) because the changes of marginal effect were not only the greatest between the Likert scales of 3 and 4, but the effect sign changes at this point. Figure 3 includes the distribution of the predicted probability by respondents’ living area. The mean of predicted probability is 57.0%, ranging from 11.9 to 92.2%. That is, on average, 57.0% of individuals perceived the impact of climate change on their personal lives. However, the bimodal distribution of predicted probability indicates that consumer perception of the impact of climate change on their personal lives was polarized and the average does not appropriately represent the whole distribution. Overall, respondents’ living area explained much of the distinction. Therefore, we will compare consumers’ WTP at the lower quartile (Q1=36.8%, i.e. 25th percentile) and upper quartile (Q3=77.6%, i.e. 75th percentile) of the distribution to represent two areas, along with the average in the second stage. The lower and upper quartiles also represent the average predicted probability for respondents who lived in Seoul (72%) and who lived in Daegu (37%).

Live in Seoul

20%

Don't live in Seoul Min: 11.9 Lower quartile: 36.8 Median: 59.0 Upper quartile: 77.6 Max: 92.2 Overall mean: 57.0 Live in Seoul: 72.0 Don't live in Seoul: 36.6

15% 10% 5% 0%

<20

20-30 30-40 40-50 50-60 60-70 70-80 80-90 Predicted probability (%)

90+

Figure 3. Distribution of predicted probability by respondent’s living area, P(Y≥4).

International Food and Agribusiness Management Review

139

Kim et al.

Volume 19, Issue 4, 2016

The estimated results of the conditional logit model are shown in Table 64. The sign of an alternative specific constant indicated respondent utility decreased when they selected not to purchase apples. All estimated parameters were statistically significant at 10% or less and have expected signs, positive for quality and carbon labeling, and negative for prices. This indicates that the attributes are important factors when consumers purchase apples. That is, higher quality apples and products with carbon labels were preferred, while expensive apples were not preferred. Also, all signs of the interactions were positive, which indicated that consumers who perceived the impact of climate change on their personal lives were more likely to select products with carbon labels or produced in an environmentally friendly way, compared to respondents who did not. In other words, consumer perception about climate change is an important factor for the success of low carbon labels. To compare consumer utility levels in monetary terms, WTPs for carbon labels were calculated as a ratio of attribute coefficients to price coefficients, as shown in the last column of Table 6. Standard errors were measured with the delta method. The values for carbon labels were compared to no carbon labels and the changes of WTPs over probability were also specified. Higher levels of WTP reveal higher utility levels for consumers when they consume the products. As the probability that consumers perceived the impact of climate change on their personal lives increased, WTPs for carbon labels increased. In particular, the gap between voluntary and MLCC increased as the probability that consumers perceived the impact of climate change on their personal lives increased. At the mean of the predicted probability (P=57.0), an average WTP for a carbon emission certification was ₩ 14,894/5 kg (about 43% of average market price, ₩ 35,000/5 kg), an average WTP for a low carbon product certification was ₩ 20,234/5 kg (about 58% of average market price), and an average WTP for MLCC was ₩ 23,429/5 kg (about 67% of average market price). As pointed out above, the lower and upper quartiles represent average predicted probability of respondents’ perception about climate change by living area. The predicted probability of respondents living in Seoul was highly distributed in the upper quartile and of respondents living in Daegu was mostly distributed in the lower quartile. Consumers who were in the upper quartile of predicted probability were willing to pay ₩ 3,916~6,455/5 kg more for carbon labels than consumers in the lower quartile of predicted probability. As the carbon labels are restricted, the differences of WTP between the lower and upper quartiles are larger. 4 We

considered the random parameters logit model but we did not find significant heterogeneity of individuals.

Table 6. Estimated results of conditional logit model and willingness to pay for carbon labels.1,2 ASC good quality superior quality CEC LCPC MLCC price P(Y≥4)×CEC P(Y≥4)×LCPC P(Y≥4)×MLCC LR χ2 (10) log likelihood

Coefficients Std. err.

Willingness to pay for carbon labels by quartiles (₩/5 kg)

-4.635** 1.136** 1.872** 1.188** 1.762** 1.817** -0.00013** 1.210* 1.384* 1.995** 1,432.39 -509.858

Lower quartile (Q1) CEC 12,956** (1,882.16) LCPC 18,018** (1,817.25) MLCC 20,234** (2,076.32)

0.320 0.155 0.174 0.453 0.440 0.498 0.000 0.734 0.731 0.836 P-value<0.05

P3 14,894** (1,501.97) 20,234** (1,563.08) 23,429** (1,777.71)

1 **

Upper quartile (Q3) 16,872** (1,941.49) 22,495** (2,100.44) 26,689** (2,385.46)

and * indicate that the values are significantly different from zero at 5% and 10% levels, respectively. ASC = Alternative specific constant; CEC = carbon emission certification; LCPC = low carbon product certification; MLCC = mandatory low-carbon certification. 3 Mean of predicted probability of consumers perceived the impact of climate change on their personal lives. 2

International Food and Agribusiness Management Review

140

Kim et al.

Volume 19, Issue 4, 2016

The marginal WTPs and confidence intervals by quality levels and carbon labels measured with the delta method are shown in Table 7. Consumers were willing to pay about ₩ 9,011/5 kg extra to purchase good quality apples instead of apples that were marketable quality. To purchase superior quality, consumers were willing to pay an extra ₩ 5,837/5 kg over good quality. Differences of market price between good quality and marketable quality were, on average, ₩ 6,798 to 10,197/5 kg in 2012 and 2013, respectively (Korea Agricultural Marketing Information Service: www.kamis.co.kr). The marginal WTP, ₩ 9,011/5 kg, was within the boundary of the difference of market prices. Average respondents were willing to pay an extra ₩ 5,339/5 kg to purchase apples with reduced carbon emission (i.e. LCPC) labels rather than carbon measured (i.e. CEC). When the low carbon labels were mandatory (MLCC), they were willing to pay an extra ₩ 8,534/5 kg to purchase apples produced with low carbon emission compared to carbon emission certification. In addition, marginal WTPs at the upper quartile were greater than those at lower quartile. The significance of the marginal WTPs varies depending on the probability of respondents perceiving that climate change impacts their personal lives. Figure 4 includes marginal WTP for carbon labels across the predicted probability of the effect of climate changes on personal lives. The shaded area indicates significant marginal WTP. Marginal WTPs from carbon emission certification to low carbon product certification were significant, with the predicted probability ranging from 19 to 100%. Approximately 98.4% of respondents were in this range. Marginal WTPs from low carbon emission certification to MLCC were significant, with the predicted probability ranging from 44 to 100%. Approximately 68.3% of respondents were in this range. As consumers perceived the effect of climate change on their lives, consumers significantly preferred MLCC to voluntary low carbon certification. This implied that consumers were ultimately interested in the Table 7. Marginal willingness to pay.1 Marginal WTP (95% confidence interval), $/5kg By apple grades marketable to good quality good to superior quality By carbon label levels2 CEC to LCPC LCPC to MLCC

Lower quartile

Average

Upper quartile

– –

9,011** (6,773~11,258) 5,837** (3,411~8,264)

– –

5,062** (1,110~9,014) 2,217 (-1,256~5,689)

5,339** (2,240~8,439) 3,195** (597~5,793)

5,623** (1,214~10,032) 4,193** (520~7,867)

1 ** 2

indicates that the values are significantly different from zero at 5% levels. WTPs are given in South Korea Won. On average, exchange rates were ₩ 1,133 = 1US$; August 2012.

Carbon emission certification to low carbon product certification ₩/5 kg

Low carbon product certification to mandatory low carbon certification

7,000

P=57.0%

Q1=36.8%

Q3=77.6%

6,000 5,000 4,000 3,000 2,000 1000 0

Figure 4. Marginal willingness to pay for carbon labels by probability levels. International Food and Agribusiness Management Review

141

Kim et al.

Volume 19, Issue 4, 2016

reduction of carbon emission to alleviate the impact of climate change on their personal lives. In particular, consumers who live in a vulnerable area to climate change significantly preferred not only reduction of carbon emissions to certification of carbon emission but mandatory low carbon labels. The premium for carbon certification and low carbon labels from this study was 30~40% in the lower quartiles and 43~54% in the upper quartiles of the predicted probability. The dispersion (quartiles) of the predicted probability of the impact of climate change on their personal lives was closely related to respondent living area (Figure 3) and levels of environmental disorder or air pollution. One critical environmental disorder in South Korea is increasing air pollution. In particular, increasing concerns about particulate matter (PM) stimulate consumers to fear for the air safety. Noll et al. (2007) and Vogt et al. (2011) found a positive relationship between carbon emission and PM. The negative impact of PM on human health has been reported (Dockery and Pope, 1994; Englert, 2004; Harrison and Yin, 2000) and the United States Environmental Protection Agency also warns of the potential health problems of PM for human lungs and heart (U.S. EPA, 2004). Jang et al. (2012) found that the level of PM10 varied by geographical area in South Korea: of 10 major cities, Seoul ranked the third highest in PM10 levels. Consumers who live in Seoul breathed PM10 at least 30 days (about a month) per year between 2001 and 2008 (Lee et al. 2011). Because of the nature of air pollution, these people may strongly feel the necessity of curbing environmental disorder and place a high evaluation on the low carbon labels, much as if they represent a private attribute.

8. Conclusions Voluntarily implemented carbon labels have not appeared to lead companies to develop technology to reduce carbon emissions. To improve the effect of carbon labels, making low carbon labels mandatory has been suggested. A prior condition to the success of the program is consumer value for carbon labels. Although previous studies have found positive consumer value for products with low carbon emission labels, the carbon labels seem to play a less important role in consumer’s decision making in the real market. This study has focused on investigation of factors that would influence consumers’ value for carbon labels along with individual perceptions about the impact of climate change on their personal lives. Based on a consumer survey with CEs, this study compared consumer preferences between mandatory and voluntary carbon labels while considering individuals’ perceptions of the impact of climate change on their personal lives. An ordered logit model was used to measure the probability of perceptions about the impact of climate change on individuals’ personal lives. The predicted probability was implemented in a conditional logit model to measure consumer preference for carbon labels based on levels of perception. This study showed that consumers in South Korea generally recognized climate change in their lives and perceived the necessity of some effort to reduce climate change. Also, consumer perceptions about the impact of climate change on their personal lives was an important indicator to predict the success of carbon labels in the market. Respondents who more perceived the impact of climate change on their personal lives (at upper quartile, 77.6%) were willing to pay ₩ 561~2,538/5 kg extra to purchase the same apples with low carbon emission labels as compared to respondents who weakly perceived the impact of climate change (at lower quartile, 36.8%). Consumers who were more exposed to the risk of environmental disorder tended to significantly prefer mandatory low carbon emission labels and strongly feel their impact. In this study, we found that consumers’ living area played a crucial role. In other words, consumers living in a vulnerable area to climate change had high value for low carbon emission labels and mandatory low carbon emission labels. A limitation of this study is the relatively small sample size and underrepresented distribution of respondents’ education levels and living area. Although the margin of error located an acceptable range, the statistical power of the survey is high and post-stratification was used for adjusting sample distribution, the concerns about the sample still remain. Future research could investigate consumer perceptions of mandatory versus voluntary labels both in Korea with a larger sample size, as well as other countries to further our understanding of this topic. International Food and Agribusiness Management Review

142

Kim et al.

Volume 19, Issue 4, 2016

Supplementary material Supplementary material can be found online at https://doi.org/10.22434/IFAMR2015.0095. Table S1. Profiles of hypothetical apples.

References Aoki, K. and K. Akai. 2013. Do consumers select food products based on carbon dioxide emissions? IFIP Advances in Information and Communication Technology 398: 345-352. Borchers, A.M, J.M. Duke and G.R. Parsons. 2007. Does willingness to pay for green energy differ by source? Energy Policy 35: 3327-3334. Chang, H.H. and R.M. Nayga. 2011. Mother’s nutritional labels use and children’s body weight. Food Policy 36: 171-178. Clark, C.F., M.J. Kotchen and M.R. Moore. 2003. Internal and external influences on pro-environmental behavior: participation in a green electricity program. Journal of Environment Psychology 23: 237-246. Cohen, M.A. and M.P. Vandenbergh. 2012. The potential role of carbon labeling in a green economy. Energy Economics 34: s53-s63. Cox, P.M., R.A. Betts, C.D. Jones, S.A. Spall and I.J. Totterdell. 2000. Acceleration of global warming due to carbon-cycle feedbacks in a coupled climate model. Nature 408: 184-187. Dockery, D.W. and C.A. Pope. 1994. Acute respiratory effects of particulate air pollution. Annual Review of Public Health 15: 107-32. Englert, N. 2004. Fine particles and human health – a review of epidemiological studies. Toxicology Letters 149: 235-242. Gadema, Z. and D. Oglethorpe. 2011. The use and usefulness of carbon labeling food: a policy perspective from a survey of UK supermarket shoppers. Food Policy 36: 815-822. Harrison, R.M. and J. Yin. 2000. Particulate matter in the atmosphere: which particle properties are important for its effects on health? Science of the Total Environment 249: 85-101. Hidrue, M.K., G.R. Parsons, W. Kempton and M.P. Gardner. 2011. Willingness to pay for electric vehicles and their attributes. Resource and Energy Economics 33: 686-705. Intergovernmental Panel on Climate Change (IPCC). IPCC Core Writing Team. 2007. Contribution of working groups I, II and III to the forth assessment report of the intergovernmental panel on climate change. IPCC (Ed.), Climate change 2007: synthesis report, Geneva, Switzerland. Jang, J.H., H.W. Lee, and S.H. Lee. 2012. Spatial and temporal features of PM10 evolution cycle in the Korea peninsula. Journal of Environmental Science 21: 189-202. Johnson, R.M. and B.K. Orme. 1996. How many questions should you ask in choice-based conjoint studies? Sawtooth Software Inc., Sequim, WA, USA. Kemp, K., A. Insch, D.K. Holdsworth and J.G. Knight. 2010. Food miles: do UK consumers actually care? Food Policy 35: 504-513. Kim, H., L.A. House, G. Rampersaud and Z. Gao. 2012. Front-of-package nutritional labels and consumer beverage perceptions. Applied Economic Perspectives and Policy 34: 599-614. Kim, J.S. 2011. Consumer awareness about climate changes and food safety. Issue and Focus 117, Korea Institute for Health and Social Affairs. Korea Environmental Industry and Technology Institute (KEITI). 2009. Public perceptions about carbon footprint of products, Seoul, South Korea. Lee, S., C.H. Ho and Y.S. Choi. 2011. High-PM10 concentration episodes in Seoul, Korea: background sources and related meteorological conditions. Atmospheric Environment 45: 7240-7247. Loureiro, M.L, J.J. McCluskey and R.C. Mittelhammer. 2002. Will consumers pay a premium for eco-labeled apples? Journal of Consumer Affairs 36: 203-219. Louviere, J.J., D.A. Hensher and J.D. Swait. 2000. Stated choice methods: analysis and application. Cambridge University Press, New York., NY, USA.

International Food and Agribusiness Management Review

143

Kim et al.

Volume 19, Issue 4, 2016

Lusk, J.L. and T.C. Schroeder. 2004. Are choice experiments incentive compatible? A test with quality differentiated beef steaks. American Journal of Agricultural Economics 86: 467-482. McFadden, D. 1974. Conditional logit analysis of qualitative choice behavior. In Frontiers in Econometrics, edited by P. Zarembka. Academic Press, New York, NY, USA. Michaud, C., D. Llerena and I. Joly. 2013. Willingness to pay for environmental attributes of non-food agricultural products: a real choice experiment. European Review of Agricultural Economics 40: 313-329. Ministry of Environment. 2007. National awareness on climate changes in South Korea (1st survey). South Korea Ministry of Environment, Sejong City, South Korea. Available at: http://tinyurl.com/j67we49. Ministry of Environment. 2008. National awareness on climate changes in South Korea (2nd survey). South Korea Ministry of Environment, Sejong City, South Korea. Available at: http://tinyurl.com/hfh57q4. Murray, A.G. and B.F. Mills. 2011. Read the label! Energy star appliance label awareness and uptake among U.S. consumers. Energy Economics 33: 1103-1110. Noll, J.D., A.D. Bugarski, L.D. Patts, S.E. Mischler and L. McWilliams. 2007. Relationship between elemental carbon, total carbon, and diesel particulate matter in several underground metal/non-metal mines. Environmental Sciences and Technology 41: 710-716. Organisation for Economic Co-operation and Development (OECD). 2010. Economic Surveys: Korea 2010. OECD Publishing, Paris, France. Available at: http://www.oecd.org/korea/45432048.pdf. Orme, B.K. 2006. Getting Started with Conjoint Analysis: Strategies for Product Design and Pricing Research. Research Publishers LLC, Madison, WI, USA. Quinn, I. 2012. ‘Frustrated’ Tesco ditches eco-labels. The Grover. Available at: http://tinyurl.com/zh2g7p6. Sammer, K. and R. Wüstenhagen. 2010. The influence of eco-labelling on consumer behaviour, results of a discrete choice analysis for washing machines. Business Strategy and the Environment 15: 185-199. Teisl, M. and A. Levy. 1997. Does nutritional labeling lead to healthier eating? Journal of Food Distribution Research 28: 18-27. Upham, P., L. Dendler, and M. Bleda. 2011. Carbon labeling of grocery products: public perceptions and potential emissions reductions. Journal of Cleaner Production 19: 348-355. U.S. Environmental Protection Agency (U.S. EPA). 2004. Air quality criteria for particulate meter (Final report, Oct 2004). EPA 600/P-99/002F-bF. Washington DC, WA, USA. Vitousek, P.M. 1994. Beyond global warming: ecology and global change. Ecology 75: 1861-1876. Vogt, M., E.D. Nilsson, L. Ahlm, E.M. Mårtensson and C. Johansson. 2011. The relationship between 0.522.5 μm aerosol and CO2 emissions over a city. Atmospheric Chemistry and Physics 11: 4851-4859.

International Food and Agribusiness Management Review

144

OPEN ACCESS International Food and Agribusiness Management Review Volume 19 Issue 4, 2016; DOI: 10.22434/IFAMR2015.0179 Received: 14 September 2015 / Accepted: 27 September 2016

Economic feasibility of tobacco leaves for biofuel production and high value squalene RESEARCH ARTICLE Aleksandre Maisashvili a, Henry L. Bryantb, and James W. Richardsonc aResearch

scientist, bAssociate professor, and cRegents professor, Agricultural and Food Policy Center, Department of Agricultural Economics, Texas A&M University, 2124 TAMU, 77843-2124 College Station, TX, USA

Abstract In this paper, we estimate the economic feasibility of biofuel production and high value squalene from tobacco-based biomass. Pro-forma stochastic financial statements were constructed and the feasibility of multi-year financial projects were evaluated. The results suggest the commercialization of biofuel production from tobacco is not economical in the short run, but may have a potential in the long run. The results also indicate the economic feasibility of high value squalene is viable under the certain conditions. Keywords: biofuels, tobacco, squalene, project feasibility analysis, multi-year simulation models JEL code: C15, Q13, Q16, Q42 Corresponding author: amaisashvili@email.tamu.edu

ÂŠ 2016 Maisashvili et al.

145

Maisashvili et al.

Volume 19, Issue 4, 2016

1. Introduction In this article we investigate the economic and financial feasibility of biofuel and high value squalene production from tobacco leaves. Using a tobacco plant as a feedstock has been motivated by recent interest in investigating renewable fuel feedstocks. In 2004 the U.S. government terminated the federal tobacco quota program (Capehart, 2004). This safety net used a production control and price support system to guarantee farmers a minimum price at the beginning of the season. The termination of the program had a substantial impact on tobacco farmers. The 2012 Census of Agriculture found that the number of tobacco farms decreased by 82% from 56,977 in 2002 to 10,014 in 2012. Elimination of the safety net put tobacco farmers into an uncertain market with risky prices (USDA, 2012). In terms of acreage, the harvested area decreased by 30% from 2004 to 2016 (USDA, 2016). Alternative use of a tobacco plant, specifically for renewable fuel, has been discussed in the literature (Adrianov et al., 2009; Giannelos, 2002; Usta, 2005; Usta et al., 2011). Over the last few years, research institutions, such as Tyton BioEnergy Systems, Synthetic and Systems Biology Innovation Hub, NUP/UPNA-Public University of Navarre, the IdAB-Institute of Agrobiotechnology, and Lawrence Berkeley National Laboratory, have begun to engineer and modify the plant to use it as a fuel feedstock. The advances in the research show promising potential for tobacco growers, but these institutions have not addressed the key issue yet, whether tobacco farming as a feedstock is economically viable or not? This study contributes to the literature by developing a probabilistic financial model, based on the Monte Carlo simulation approach, to assess the economic feasibility of tobacco leaves as a renewable fuel feedstock. Currently, two main products can be produced from a tobacco plant if used as a feedstock, a finished motor fuel, and high value squalene oil. Both of these products are highly demanded, as outlined in the next two sub sections. Tobacco as a renewable fuel Rising energy costs and policies to reduce dependence on foreign energy supplies has dramatically increased the domestic production of plant-based fuels as an alternative fuel source. Fossil fuels are becoming scarce resources and their prices will again increase gradually. Moreover, the energy sector is one of the largest contributors to greenhouse gas emissions, and a contribution to climate change. Global demand for energy is expected to increase by at least 50% over the next 20 years, with this increase mainly driven by rapid population growth and industrial development in developing countries (Brackmort, 2012; Liu et al., 2014). Alternatives to fossil fuels have been sought for decades. Biofuels have been considered one the most promising possibilities for solving the problem. The U.S. is searching for solutions through biofuels because these fuels are renewable, which means they can be constantly replenished and, in contrast to nonrenewable sources, wonâ&#x20AC;&#x2122;t eventually dwindle. Moreover, the United States is committed to increase the amount of biofuels it uses from 9 to 36 billion gallons by 2022 (Brackmort, 2012). Currently, biofuels are produced from different types of biomass, including sugar-cane, corn, vegetable oil, algae, and wood, to name a few. Ethanol production is the major plant based-biofuel that has been used as an alternative fuel in the United States and Brazil. With the current available technology, virtually all ethanol is derived from corn in the United States. Because corn is also a food source, use of corn for ethanol production will affect consumers through food price increases. Considering a nonfood alternative for biofuel production is critical. Producing fuels and fuel-like precursors from farm-ready and non-food crops have been motivated by the U.S. Department of Energy as well. The tobacco plant is a potential source of renewable energy. In particular, tobacco leaves contain hydrocarbon molecules that can be converted into a fuel that can be used as a substitute for a petroleum fuel.1 The tobacco leaves have at least four advantages in terms of biofuel production:

1 The supply chain for tobacco-based renewable fuels is going to closely resemble to an existing supply chain for corn ethanol. The chain is composed

of three main segments, tobacco biomass production and harvesting, squalene extraction, and refining to finished motor fuel. Tobacco production and harvesting will be similar to the conventional tobacco farming and the differences are outlined in the article. By-products, such as distillerâ&#x20AC;&#x2122;s dried grains with solubles, will not be fed to animals but will be treated as residual stalks that are potential inputs for the pulp and paper industry.

International Food and Agribusiness Management Review

146

Maisashvili et al.

Volume 19, Issue 4, 2016

1. Unlike other biofuels made from corn, soybean, and other crops, tobacco as a biofuel feedstock would not directly affect a major food source. 2. Tobacco generates multiple harvests per year and produces a large biomass. 3. Tobacco is amenable to genetic engineering, which means that in the long-run its leaves can store more hydrocarbon molecules than are currently present. 4. Tobacco cultivation has been widely known and it is grown in large tracts throughout the South and Eastern U.S., as well as more than 100 countries. Tobacco as a high value squalene The main hydrocarbon of interest contained in a tobacco leaf is squalene. Squalene is an isoprenoid compound structurally similar to beta-carotine and it is an intermediate metabolite in the synthesis of cholesterol. Squalene has clinical applications, mainly used as an adjunctive therapy in a variety of cancers (Kelly, 1999). Brito et al. (2011) also discuss the importance of squalene use in emulsion adjuvants. Squalene is also shown to have antioxidant, drug carrier, detoxifier, skin hydrating, and emollient characteristics (Kim and Karadeniz, 2012). Demand for squalene is high and according to Markets and Markets, a research firm, the squalene market is estimated to reach $211 million by 2021 and is projected to grow annually by 10.2% between 2016 and 2021 (Markets and Markets, 2016). Squalene extracted from the tobacco leaves does not have to be used for biofuel production and can be commercialized directly, given its clinical uses. Objective of the study The objective of the study is to estimate the economic feasibility of both biofuel production from tobacco leaves and high value squalene production. First, we analyze biofuel and squalene markets independently. Following, we discuss major requirements for tobacco-based biofuel production to become financially feasible. To achieve this objective, the sub-objectives are to: 1. Model the cost of tobacco production for biomass. 2. Estimate the cost of squalene extraction from the leaves. 3. Estimate the cost of squalene refining for the finished motor fuel. 4. Estimate the probability of profitable production of finished biofuels and high value squalene from tobacco biomass, and report pro-forma financial statements to analyze the overall economic feasibility of the two products.

2. Literature review Research related to feedstock feasibility for biofuels literature is very broad, and extensive. Many feedstock alternatives have been studied from an economic standpoint, such as, algae, sweet sorghum, sugar cane, energy cane, switchgrass, corn starch, corn stover, to name a few (Monge et al., 2014; Outlaw et al., 2007; Palma et al., 2011; Rezenede and Richardson, 2015; Richardson and Johnson, 2014, 2015; Richardson et al., 2007a,b, 2010, 2014a,b). Comparing a tobacco-based feedstock feasibility with the feasibility studies of other feedstock materials are not applicable, as they depend on different production, price, yield, and, most importantly, technology assumptions. Research related to biofuel and squalene production from tobacco feedstock is limited. Because the concept of using a tobacco plant as a feedstock is still in its infancy, the area of research is a new concept. At this time there does not exist a single study that addresses the economics of renewable fuel production from the tobacco plant or the production of high value squalene from tobacco leaves. Some studies have analyzed the chemical properties of producing biofuel from tobacco seeds (Giannelos, 2002; Usta, 2005; Usta et al., 2011). Although these authors found that the seed oil can be used as a fuel for diesel engines, tobacco plants yield a modest amount of seeds, making a biofuel production unattractive from a financial standpoint. Adrianov et al. (2009) explored engineering approaches to enhance the oil content in tobacco leaves for biofuel production. Typical tobacco plant leaves contain 1.7 to 4% of oil on a dry weight basis. Tobacco plants were engineered International Food and Agribusiness Management Review

147

Maisashvili et al.

Volume 19, Issue 4, 2016

and gene modification increased oil content to 6.8%. Adrianov et al. (2009) strengthen the argument that the tobacco plant can be genetically modified to produce more hydrocarbon molecules. They emphasize that because of its biomass potential and the possibilities of further metabolic engineering, tobacco represents an attractive and promising energy plant which could also serve as a model for utilization of other high-biomass plants for oil production. However, this hypothesis has not been verified from an economic standpoint. All of these studies discuss the potential of the tobacco plant for being a next generation biofuel crop. However, none of these studies actually describe the economic feasibility of the feedstock to produce biofuels. The current study aims to bridge the gap and focus on economic feasibility. The study will be useful to investors or scientists looking for a new biofuel feedstock or new materials for squalene production.

3. Data and methodology Economic feasibility studies including risk have proven to be a powerful tool in business valuation (Johnson et al., 2016; Kwak and Ingall, 2007; Monge et al., 2014; Outlaw et al., 2007; Palma et al., 2011; Rezende and Richardson, 2015; Richardson and Johnson, 2014; Richardson et al., 2007a,b, 2014a,b). We employ a stochastic cash flow and net present value (NPV) approach for the analysis.2 A stochastic NPV approach has been widely used as one of the main methodologies to define the overall viability of a risky project (Monge et al., 2014; Outlaw et al., 2007; Palma et al., 2011; Richardson et al., 2007a,b). NPV is also considered to be a simple and intuitive approach for lay investors (Monge et al., 2014). Therefore, the NPV probability distribution estimated using a Monte Carlo simulation method, and the underlying probability distributions for stochastic variables (yields, input prices and output prices) will be used to conduct the feasibility studies for biofuel and squalene. The estimation of the underlying probability distribution for NPV is critical in the estimation of the probability of success, where economic success is defined as a positive NPV criteria.3 Assumptions for estimating the probability distributions for the random variables (yields and prices) are discussed in the subsequent sections. Tobacco biomass yield Tobacco biomass yield and squalene production are critical stochastic variables for the two economic simulations. Tobacco for biofuel feedstock production requires more dense planting than traditional tobacco production for conventional use, so historical yield data cannot be used directly. Because there is no market for tobacco yield that requires dense planting, the yields for biomass tobacco were simulated using the GRKS distribution.4 The assumed minimum, middle, and maximum yield values were obtained from the trials conducted by Mundell and Chambers (2011) at the Kentucky Research Institute. Their results indicated that tobacco yield ranged between 55,000 to 150,000 pounds per acre. Assuming 80% moisture content, this translates into approximately 11,000 to 30,000 pounds per acre on a dry matter basis. Given limited yield data is available, the GRKS distribution was used to simulate yield with a minimum of 55,000, middle of 90,000, and maximum of 150,000 pounds per acre. A 6% growth rate was incorporated each year to account for technological increases in yield over time. Using the inverse transform sampling method,5 the tobacco yield is distributed as follows (the values are in thousands of pounds and the sign, â&#x20AC;&#x2DC;~â&#x20AC;&#x2122;, denotes a stochastic variable):

~ Yield

tobacco

~ GRKS(55, 90, 150)

(1)

2 A stochastic

NPV approach has advantages when compared to conventional NPV methodology as the former allows us to combine the shapes of probability distributions for variables we cannot observe, such as: profits, costs, or prices so the decision maker can make better decisions. 3 A positive NPV indicates that the internal rate of return exceeds the investorsâ&#x20AC;&#x2122; discount rate. For the present study the discount rate is set at 7% as suggested by others (Richardson et al., 2007a; Palma et al., 2011; Monge et al., 2014). 4 The GRKS is a parametric, two-piece normal distribution and has been used extensively for studying project feasibility analysis (Monge et al., 2014; Outlaw et al., 2007a,b; Palma et al., 2011; Richardson et al., 2007a,b). The GRKS is similar to the triangle distribution and has the same parameters: minimum, middle, and maximum values. However, it is different from the triangle distribution because it accounts for uncertainties that go below and above the assumed minimum and maximum values, respectively about 5% of the time. 5 Inverse transform is a method for pseudo-random number sampling to generate sample numbers at random from any probability distribution given its cumulative distribution function.

International Food and Agribusiness Management Review

148

Maisashvili et al.

Volume 19, Issue 4, 2016

Squalene yield Extraction volume of squalene can be estimated with a high degrees of precision after its content is determined because squalene is a non-volatile compound. As mentioned previously, Adrianov et al. (2009) was able to increase the oil content in the tobacco plant up to 7% in a short time span, so a high squalene content in the long run is a realistic assumption. Currently, tobacco contains about 2-4% of squalene on a dry matter basis, and, as previously mentioned, several research institutions aim to achieve as high as 20% yield content per dry matter. The information was verified as well through personal communication with tobacco researchers at Synthetic and Systems Biology Innovation Hub (Synthetic and Systems Biology Innovation Hub, personal communication, April 2015). We used two different approaches for squalene production. For the tobacco biofuel market, squalene content was assumed to be 20% on a dry matter basis. For the high value squalene market, three cases were considered. The first case uses a conservative approach and assumes that current squalene content is 2% per dry matter, experiences modest growth over time and reaches 20% in 8 years and stays the same thereafter. The second scenario assumes that squalene grows moderately faster and reaches 20% in 6 years and stays the same thereafter. The third scenario analyzes an aggressive growth and assumes that squalene content will reach 20% per dry matter in 5 years. Squalene growth assumptions, used in the three scenarios, are modest given that scientists expect to reach 20% goal in the short period of time (Synthetic and Systems Biology Innovation Hub, April 2015, personal communication). Changes in squalene content will mainly affect the cost of refining. Production and extraction costs will basically remain unchanged, because expenses are heavily dependent on the biomass production. Diesel, gasoline, and squalene prices Diesel and gasoline values were simulated to incorporate risk for final products. Although there are many ways to simulate fuel prices, a Markov process was shown to accurately project prices, as well as providing dynamic information about the market (Aleksandrov et al., 2013; Gonzales et al., 2005; Mostafei et al., 2011). In a Markov process, only the present value of a variable is relevant for predicting its value in the future. Following the basic generalized Wiener process as outlined in Hull (2006), the prices of gasoline and diesel are simulated as follows: lnPt+1 – lnPt = σε(Δt)1/2 (2) where, lnPt is a natural logarithm of the gasoline or diesel price observed in period t; σ is the annualized variance of diesel or gasoline price obtained from past observations; Δt is a small change in period time t; and ε is a correlated standard normal deviate obtained by multiplying a factored correlation matrix between gasoline and diesel prices by a vector of standard normal deviates. Taking lnPt to the right hand and exponentiating both sides, the stochastic process follows as: Pt+1 = Ptexpσε√Δt (3) The simulated prices are path dependent, where: Pt+2 = E(Pt+1)expσε√Δt (4) Simulating the squalene prices is more challenging because a long history of data is not available, compared to fuel prices. Traditionally, shark livers are processed to obtain squalene, which has increased shark hunting. The extraction of squalene from vegetable sources has been motivated by environmental concerns, surrounding shark hunting (Markets and Markets, 2016). Not until recently, Amyris, Inc. (Emeryville, CA, USA) has started selling commercial quantities of squalene derived from crushed sugarcane. Currently, squalene sells for $ 30 per liter, which translates into $ 114 per gallon (Ciriminna et al., 2014). However, determining the International Food and Agribusiness Management Review

149

Maisashvili et al.

Volume 19, Issue 4, 2016

market value of squalene in the future is highly uncertain. The GRKS distribution was used to simulate squalene prices given the limited information available. Similar to squalene yield scenarios, three cases were considered for simulating the prices as well. In the first, the minimum, middle, and maximum values were chosen to be $ 92, 100, and 114 per gallon, respectively.

~ Price

squalene

~ GRKS(92, 100, 114)

(5)

Note, that the first scenario uses a conservative approach. In particular, we used a current market price ($ 114 per gallon) to be the maximum value in the distribution. The second and third scenarios are even more conservative, and they incorporate price decay functions to account for possible supply growth from plant based squalene and environmental regulations regarding squalene production from sharks. In particular, the second scenario assumes that squalene value will decline gradually, on average, by 5% per year over the next 10 years. The third scenario is the most pessimistic, and assumes that squalene value, on average, declines by 10% per year over the next 10 years. Tobacco production cost The values used in the production budget are based on projected input prices and recommended production practices suggested by the Virginia Cooperative Extension (2012) and North Carolina State University Cooperative Extension (2012). The information provided by the guidelines was adapted according to the production characteristics associated with more dense planting for biomass production, rather than for human use. The variable production costs are composed of two types of costs; pre-harvest costs and harvest costs. Additionally, production also includes fixed costs. The detailed costs are based on a machine harvest method and incorporate multiple harvests per year. Pre-harvest costs are typically cash expenses that must be paid annually to produce a crop of tobacco prior to harvest. Examples of pre-harvest variable costs include: plants, fertilizer, chemicals, machinery fuel and repairs, hired labor, machinery fuel, machinery repairs and maintenance, interest on the sum of cash costs for the time from planting to harvest, and labor expenses. These expenses appear also in the harvest variable costs section. The guidance on expenses related to machinery and labor were obtained from extension budgets and were adjusted according to the production sizes discussed in a subsequent section. Drying the tobacco biomass is a major post-harvest cost. Moisture content of the feedstock can be anywhere from 70-80% depending on field and weather conditions. The maximum optimal moisture content for harvest was assumed to be 80%. Fixed costs are expenses that result from the ownership of land, equipment and buildings. Examples of these costs include depreciation, property taxes, interest, and insurance. Detailed information on fixed expenses were incorporated in the model. For illustration, the adjusted mean production costs based on 90,000 pounds of green biomass per acre are presented in Supplementary Table S1. For input prices that have limited historical data (i.e. herbicides, fungicides), we employed a GRKS distribution to simulate those costs. For input prices that have relatively large history of data (i.e. fertilizer prices), we employed a multivariate empirical distribution (MVE) proposed by Richardson et al. (2000). A MVE is a non-parametric distribution and accounts for inter-temporal (across time) and intra-temporal (across variables) correlation among the random variables. Squalene extraction and refining costs We employed a supercritical fluid extraction (SFE) method to calculate the cost of squalene extraction from the tobacco biomass. The literature suggests that SFE is a proven method of extraction (Anderson, 2011; Bhattacharjee and Singhal, 2003; Bhattacharjee et al., 2012; Cathpole et al., 1997; Mercer et al., 1999; International Food and Agribusiness Management Review

150

Maisashvili et al.

Volume 19, Issue 4, 2016

Reverchon, 1997; Rizvi et al., 1986; Vazquez et al., 2007). Detailed information for the cost of extraction, squalene recovery, and the cost of an SFE plant was obtained through personal communication with Cybertech Engineering. Supplementary Table S2 summarizes the required inputs and deterministic total cost of squalene extraction per acre. Squalene is too heavy to be used directly in gasoline because it has 30 carbon atoms in its backbone. Tracy et al. (2011) describe squalene as a promising potential precursor to high-octane gasoline because the carbon backbone is branched. The authors describe that branched hydrocarbons are known to have octane numbers superior to linear molecules. The authors also show that catalytic cracking is a natural route to generate gasoline from squalene. Their results indicated that the gasoline obtained from squalene has high octane numbers. In particular, the research octane (RON) and motor octane numbers (MON) were 96.5 and 84.6, respectively. The result meets the required minimums of 91 RON and 82 MON for gasoline. Cost of refining is calculated by breaking down the total cost of a gallon of gasoline to obtain the fraction of refining from the total cost. The cost of gasoline is broken down into four components: crude oil, refining cost plus profits, distribution and marketing cost, and taxes. Crude oil margin is the difference between the monthly averages of the composite refinery acquisition cost, which is basically the average price of crude oil paid by oil refineries. Refining cost and profits is the difference between the monthly average spot price of gasoline after it exits the refinery and the average price of crude oil purchased by refineries. Distribution and marketing costs and profits are the difference between the average retail price of gasoline as computed from U.S. Energy Information Administrationâ&#x20AC;&#x2122;s (EIA) weekly survey and the sum of the other two components. The fourth component is taxes, which is a monthly national average of federal and state taxes applied to gasoline and diesel. According to EIA, the cost breakdown of gasoline and diesel as of May, 2016 are represented in Supplementary Table S3. Based on the information provided in Supplementary Table S3, the approximate cost of refining can be estimated accounting for the fraction of diesel and gasoline that could be refined from squalene. Following Tracy et al. (2011), Hillen et al. (1982) and Banerjee et al. (2002), the refined squalene is assumed to yield 65% gasoline and 35% diesel. This leads to an estimated cost of refining to be approximately $ 0.46 per gallon.6 Financial statements The simulated random variables for yields and prices were used to construct the stochastic pro-forma financial simulation model to simulate a 10-year time horizon. Tobacco is cultivated primarily on small and medium-sized farms. The most common size of a tobacco farm is from 4 to 12 acres (Huntrods, 2012; USDA, 2006). Therefore, we constructed the financial statements for a 5 acre hypothetical farm in Barren county, Kentucky. Stochastic input and output variables were simulated according to the aforementioned distributional assumptions. Several key output variables (KOV) were calculated to assess the overall financial and economic feasibility of producing biofuels from squalene from tobacco. Through Monte Carlo simulation, the model estimates a probability distribution for each of the KOVs such as, present value of ending net worth, net cash income, net cash flow, net worth, and NPV. The distributions of these KOVs have been used for analyzing the feasibility of alternative feedstocks for renewable fuels (Monge et al., 2014; Outlaw et al., 2007; Richardson and Johnson, 2014; Richardson et al., 2007a, 2014a,b). To estimate the distributions for the KOVs, the model must include variables in an income statement, a cash flow statement, and a balance sheet. The NPV is calculated as follows:

~ ~ EndingNetWorth â&#x2C6;&#x2018; ( Dividends ) +( ) (1+r) (1+r) 10

NPV = â&#x20AC;&#x201C; (BeginningNetWorth) +

t=1

6 The

fuels once refined are equivalent to gasoline and diesel, and they can enter the current fuel pipeline without modification.

International Food and Agribusiness Management Review

151

(6)

Maisashvili et al.

Volume 19, Issue 4, 2016

Equation 6 is calculated by adding the present value of ending net worth and total discounted dividends, and subtracting the beginning net worth. Annual dividends (farmer withdrawals) are calculated as a fraction of the beginning net worth and, as a bonus, from positive annual net cash income (NCI). If the value of NPV is greater than zero, the business is considered an economic success (Palma et al., 2011; Richardson et al., 2012). To calculate the beginning net worth, the summation of the initial assets is required. The main assets are tobacco drying equipment, the SFE plant, and the harvesting machinery. It was assumed that a share of the initial capital needed to operate the business is financed. The loan share was assumed to be 60% for all the equipment, as suggested by Richardson et al. (2012), yielding a beginning liability (details discussed in the balance sheet section). The beginning liability is subtracted from the initial asset value, and any initial cash reserves are added, to yield beginning net worth. Income statement Among the three financial statements required for the analysis, estimating the pro-forma income statement is the first step. The revenues are the product of stochastic diesel and gasoline prices, and stochastic fuel production. The revenue for high-value squalene production is estimated from stochastic squalene production and stochastic squalene price. Additionally, we incorporated the revenues generated from lignocellulose feedstock which is a residual of the biomass after oil has been extracted. Tobacco residuals (tobacco stalks) are referred to be promising materials for pulp and paper industry (Shakhes et al., 2011). We assume that the selling price of the feedstock will average $ 35 per dry short ton (McAloon et al., 2000). Revenues generated from the tobacco stalks does not have a significant impact on the analysis. The expenses are composed of pre-harvest and harvest expenses, tobacco drying expenses, squalene extraction cost, land rent, total operating expenses, and the interest expenses. The interest expenses consist of the interest from the original loan, the interest on the operating loan, and the interest on any cash flow deficit from the previous year. The operating loan is meant to cover a portion of production, operating, and fixed expenses. The portion of fixed operating expenses covered annually was obtained from Richardson et al. (2012). NCI equals total revenue minus total cash expenses.

~ ~ ~ NCI = Revenue – Expenses (7) t

Cash flow statement The main KOV in the pro forma cash flow statement is ending cash (EC). EC is the difference between cash inflows and outflows.

Ending Casht = Cash Inflowst – Cash Outflowst (8) Cash inflows are the summation of NCI from the income statement, beginning cash, and interest earned on positive cash reserves from the previous year.

Cash Inflowst = Beginning Casht + NCIt + Interest Earnedt (9) Cash outflows are the summation of principal payments on the original loan, the fully paid deficit loan from the previous year’s cash flow deficit, if any, dividends, and income tax. If NCI is positive, we assume that the owner receives a dividend equal to 5% of net cash income.

Cash Outflowst = Principal Paymentst + Deficit Loant-1 + Dividendt + Income Taxt (10)

International Food and Agribusiness Management Review

152

Maisashvili et al.

Volume 19, Issue 4, 2016

Income taxes are estimated using the internal revenue system corporate income tax schedule. The taxable income is estimated by subtracting the annual depreciation from NCI plus interest earned. If the difference is positive, income taxes must be paid, otherwise it is zero. Balance sheet Annual net worth comes from the pro-forma balance sheet and is the difference between of the stochastic value of assets and liabilities. The annual value of assets equals cash reserves plus the value of plant equipment and land. Total capital assets (beginning net worth) for the process unit are $ 850,000 and, as mentioned previously, will be financed with 40% equity and 60% debt at a 7% interest rate over a 10 year period. The initial capital investment in assets includes SFE plant, machinery harvester, and the drying equipment. Annual EC reserves are conditional on positive EC coming from the cash flow statement. The annual value of liabilities is calculated by summing total deficit loans and the equity loan balance.

Liabilitiest = Original Loant + Deficit Loant (11) The deficit loan is acquired to cover negative EC and, condition on the next yearâ&#x20AC;&#x2122;s EC, is fully paid in one year.

~ Deficit Loan = t

{

~ ~

|Ending Casht|, Ending Casht < 0 (12) 0, Ending Casht > 0

The equity loan balance is the remaining balance in a given year after paying the principal and interest from the previous year.

~ Original Loan = t

{

Original Loant-1 â&#x20AC;&#x201C; Loan Principalt, t>0 (13) Total Investment Ă&#x2014; Fraction Financed, t = 0

4. Results and discussion Economic feasibility of biofuel production Diesel, gasoline, squalene, biomass yield and the production costs were simulated with their distributional assumptions each year of the 10-year horizon. The model was simulated for 500 iterations to generate sufficient observations to estimate reliable distribution for the KOVs.7 Stochastic pro-forma financial statements were constructed and summary statistics of the KOVs for biofuel production from tobacco are presented in Table 1. Results indicate that the potential production of biofuel from tobacco is uneconomical, even if the average squalene yield is 20% per dry matter. The summary statistics shown in Table 1 indicate the average net cash income, (averaged across 10 years), is negative for every iteration, given the minimum and the maximum values are $ -108,000 and -97,000, respectively. The average EC (averaged across 10 years) is also negative, and the range of the distribution is between $ -810,000 and -761,000. Large cash outflows are mainly caused by the increasing short term debt generated due to negative EC balance observed in previous years. As a result of increasing carry-over debt, the liabilities exceed the assets, yielding negative ending net worth by terminal year. The range of average ending net worth in the last year is between $434,000 and -385,000. The average NPV is highly negative, yielding a zero probability of economic success.

7 The sampling procedure was the Latin hypercube, which allows for more complete simulation over the uniform (0,1) distribution space, so a smaller

number of iterations is necessary when compared to the less efficient Monte Carlo sampling procedure.

International Food and Agribusiness Management Review

153

Maisashvili et al.

Volume 19, Issue 4, 2016

Table 1. Summary statistics of the key output variables obtained from pro-forma financial statements for tobacco biofuel market.

Mean Standard deviation Coefficient of variation Minimum Maximum

Average net cash income

Average ending cash

Average ending net worth

Net present value

-103,528 2,070 -2.00 -108,821 -97,982

-784,492 9,541 -1.22 -810,153 -761,147

-409,171 9,541 -2.33 -434,832 -385,826

-862,987 9,590 -1.11 -887,506 -837,297

The primary reason behind the unfavorable economic results is low prices for diesel and gasoline compared to the production and oil extraction costs of biomass.8 Given NPV is negative, the model was solved to estimate the per unit subsidies on diesel and gasoline that would make NPV zero. The subsidies, on average, are $ 15/gallon for diesel and $ 25/gallon for gasoline. Thus, it is not likely that renewable fuels from tobacco will be economically viable, at least in the short run. Economic feasibility of high value squalene market Nine scenarios were analyzed to reflect assumptions about potential yield increases and price decreases that will occur if the supply of squalene increases. The scenarios are summarized in Table 2. The summary statistics for selected KOVs for the high value squalene market are presented in Table 3. The average NCI is positive for all nine scenarios. Probability of a positive average NCI is 100% for all scenarios except SC8P10, which has a 99% chance of a positive NCI, and the mean value is much lower when compared to other scenarios. This result is not surprising given that SC8P10 assumes a modest growth of squalene content per dry matter and a sharp decline in squalene price over time. Comparing average NCI across squalene improvement rates for a given price decrease of 5% (SC8P5, SC6P5, SC5P5) shows that increasing the technological advances will significantly improve the economic viability as average NCI increases from $ 66,000, to $ 112,000, to $ 133,000, as the time to reach 20% squalene is reduced from 8 to 5 years, respectively. However, one must not ignore the price effect of increased supply, as evidenced by scenarios SC6P10, SC6P5, SC6P0, where average NCI increases from $ 66,000, to $ 112,000, to $ 169,000, respectively, as price decline rates are smaller and smaller. Judging the project feasibility solely on the average NCI is unrealistic as it does not capture the actual cash on hand a farmer will have left after paying the dividends, the principal payments, and the carry over debt payments (if any) over time. 8 The

average sale price for gasoline was $ 2.27/gallon and for diesel it was $ 2.32/gallon.

Table 2. Construction of nine scenarios for analyzing the feasibility of high value squalene market. Scenario

Name

Squalene content

Price response

1 2 3 4 5 6 7 8 9

SC8P10 SC8P5 SC8P0 SC6P10 SC6P5 SC6P0 SC5P10 SC5P5 SC5P0

20% in 8 years 20% in 8 years 20% in 8 years 20% in 6 years 20% in 6 years 20% in 6 years 20% in 5 years 20% in 5 years 20% in 5 years

-10%/year -5%/year -0%/year -10%/year -5%/year -0%/year -10%/year -5%/year -0%/year

International Food and Agribusiness Management Review

154

Maisashvili et al.

Volume 19, Issue 4, 2016

Table 3. Summary statistics of key output variables for nine scenarios for high value squalene market.1 SC8P10 SC8P5 Average net cash income mean 26,245 66,480 st.dev 9,859 13,389 CV 38 20 min -1,542 31,127 max 61,485 115,136 P (NCI>0) 99.7% 100.0% Average EC mean -282,757 -187,801 st.dev 36,762 40,780 CV -13 -22 min -396,389 -324,803 max -181,946 -76,275 P (EC>0) 0.0% 0.0% Present value of ending net worth mean 1,416 145,434 st.dev 39,469 44,123 CV 2,787 30 min -112,097 11,491 max 104,890 259,434 P (PENW>0) 51.2% 100.0% P (PENW>BNW) 0.0% 0.0% 1

SC8P0

SC6P10 SC6P5

SC6P0

SC5P10 SC5P5

SC5P0

117,554 17,819 15 69,506 182,065 100.0%

66,028 12,817 19 27,478 109,193 100.0%

169,854 20,759 12 113,822 245,541 100.0%

86,639 133,356 13,350 16,673 15 13 43,737 85,821 132,554 194,848 100.0% 100.0%

192,694 22,009 11 132,963 275,245 100.0%

-89,349 -117,930 -16,000 93,751 44,446 46,352 48,745 53,625 -50 -39 -305 57 -245,163 -270,661 -185,489 -94,732 33,796 5,089 116,622 244,440 2.1% 0.2% 37.7% 95.2%

-14,778 88,406 50,712 53,490 -343 61 -185,497 -98,582 120,900 238,441 39.6% 93.9%

202,632 59,606 29 -5,349 376,363 99.8%

298,022 51,951 17 146,841 434,837 100.0% 2.6%

236,333 45,738 19 71,421 358,722 100.0% 0.0%

547,983 62,097 11 361,037 725,839 100.0% 99.0%

158,174 45,342 29 6,662 275,118 100.0% 0.0%

112,425 15,754 14 67,583 168,541 100.0%

304,456 48,094 16 142,612 432,262 100.0% 1.6%

468,325 59,044 13 296,064 638,368 100.0% 86.8%

380,757 50,024 13 207,263 521,164 100.0% 31.4%

NCI = net cash income; EC = ending cash; PENW = present value of ending net worth; BNW = beginning net worth.

The average EC is negative for all scenarios except for SC6P0, SC5P5, and SC5P0. The results show that the average EC is highly responsive to both number of years it takes to reach 20% squalene content on a dry matter basis, and on the squalene price. Comparing average EC across squalene growth rates for a given price decrease of 5%, the probability of obtaining positive EC increases from 0 to 94% (SC8P5, SC6P5, SC5P5). On the other hand, assuming 20% squalene is achieved in 6 years, and comparing across squalene price decline rates, the probability of positive EC increases from 0.2 to 95% (SC6P10, SC6P5, SC6P0). The third part of Table 3 summarizes the present value of ending net worth (PENW) for the nine scenarios. The average PENW is marginally positive for SC8P10, and significantly positive for the other eight scenarios. PENW is highly responsive to both, squalene content per dry matter and squalene price. In particular, assuming a 20% squalene content per dry matter is reached in 8 years, the probability of a positive PENW increases from 51 to 100% as the average squalene price drops from 10 to 0% per year (SC8P10, SC8P5, SC8P0). Similarly, the probability of a positive PENW is 100% as long as the 20% squalene content per dry matter is reached in 5, 6, or 8 years, holding the average price drop fixed at 5% per year (SC8P5, SC6P5, SC5P5). The results also indicate the probability of PENW being greater than the beginning net worth (BNW) is less than 40% for all scenarios besides SC6P0 and SC5P0. The probability of PENW being greater or equal than the BNW is 86 and 99% for SC6P0 and SC5P0, respectively. Combining all of the information from the three pro-forma financial statements is necessary to determine the probability of economic success. The probabilities of obtaining a positive NPV across the 9 scenarios are presented in Table 4. The results indicate both the squalene growth rate and the price decline rate are critical components in the analysis. In particular, assuming the squalene content reaches 20% in 8 years, the probabilities of obtaining positive NPVs improve from 0 to 81% as price reductions go from 10 to 0%. However, if average squalene price drops, on average, by 10% per year, the maximum probability of economic International Food and Agribusiness Management Review

155

Maisashvili et al.

Volume 19, Issue 4, 2016

Table 4. Probabilities of obtaining a positive net present value for nine scenarios for the high value squalene market. Price decline rate

Squalene content per dry matter

10%/year 5%/year 0%/year

20% in 8 years

20% in 6 years

20% in 5 years

0% 1% 81%

2% 86% 100%

40% 99% 100%

success is 40%, given that 20% squalene content per dry matter is achieved in 5 years. Two scenarios have 100% probabilities of economic success. These scenarios require 20% squalene content per dry matter to be achieved either in 5 years or in 6 years, and assumes that the average squalene price remains unchanged over the 10-year time horizon. In addition to looking at the average NPV values, it is important to look at the NPV distribution to assess the risk component for different scenarios. Figure 1 shows the cumulative distribution functions estimated from the simulation results to contrast the feasibility probabilities of the three scenarios using the NPV. In particular, Figure 1 shows the impacts of yearly price decline rates assuming that 20% squalene content can be achieved in 5 years (SC5P10, SC5P5, SC5P0). The probabilities of the project not being feasible (NPV<0) can be identified in the graph where the sigmoidal curves intercept the vertical axis. Therefore, the probabilities of the project being feasible (NPV>0) are estimated by subtracting the intercept from one. As shown in Figure 1, assuming squalene price, on average, remains unchanged, the probability of the project being feasible is 100% (SC5P0). Even if the average squalene price declines by 10% per year, there is still a 40% chance of obtaining a positive NPV, as long as 20% squalene content is achieved in five years (SC5P10). As mentioned previously, the squalene market growth rate is projected to be greater than 10% per year, meaning that the annual price decline of 10% is perhaps too conservative. The most realistic scenario among the three is SC5P5, assuming that squalene price, on average, declines by 5% per year and 20% squalene content can be achieved in 5 years (SC5P5). The probability of obtaining a positive NPV under this scenario is 99%, showing a promising potential for tobacco growers in the future.

SC5P10

SC5P5

SC5P0

1.0 0.9 0.8

Probabilities

0.7 0.6 0.5 0.4 0.3 0.2 0.1

-300

-200

0.0

-100 0 100 Net present value (Ă&#x2014;1000 $)

200

300

400

500

600

Figure 1. Cumulative distribution function approximation of the net present value for high value squalene market by altering price decline rates and assuming 20% squalene content can be achieved in 5 years.

International Food and Agribusiness Management Review

156

Maisashvili et al.

Volume 19, Issue 4, 2016

Economic feasibility of biofuel production with sensitivity analyses The previous two sections analyzed financial and economic feasibilities of a tobacco-based biofuel and squalene production independently. As discussed above, under current available technologies, producing and marketing only biofuel from the biomass is not profitable. The purpose of this section is to discuss the requirements for tobacco-based biofuel to become economically feasible. The process requires a significant portion of the biomass to be commercialized towards high value squalene market. In other words, part of the squalene extracted is refined and sold as a fuel, and the remaining portion is marketed as a high value product. As previously stated, the business is considered an economic success if the value of NPV is greater than zero. However, as outlined by Monge et al. (2014) ‘an investor might adopt a more stringent criteria to consider a technology feasible by avoiding the chances of a negative NPV as much as possible’. Richardson et al. (2012) and Monge et al. (2014) follow the criteria that a business is considered an ‘economic success’ if the chances of a positive NPV are 95% or higher. Under this criteria, we conducted sensitivity analyses on multiple parameters to assess their impacts on the probability of economic success. In particular, we performed sensitivity analyses on the different fractions of squalene and fuel marketed, as well as the implications of capital and operating expenses reductions on the technologies’ feasibility chances. We maintained all the main aforementioned assumptions. The details of the assumptions are as follows: tobacco leaves contain 20% squalene yield per dry matter, the biomass yield is distributed as presented in Equation 1, gasoline and diesel prices follow a distribution from Equation 3, and the squalene price is distributed as outlined in Equation 5. With regard to the portion of squalene refined and commercialized as fuel, two cases were considered. In case one, 70% of the squalene is refined and sold as a fuel and the remaining 30% is marketed as a high value product. In case two, the portions were assigned as 75-25% for fuel and squalene, respectively. Commercializing more than 75% squalene for the fuel market is not profitable. Table 5 shows the benefits of capital and operating expense reductions on the technologies’ feasibility chances under the two cases discussed above. Cost reductions of 25% are considered possible in the short term, whereas 50% cost reductions are considered more long-term scenarios (Monge et al., 2014). There are several cases under which the biofuel commercialization from a tobacco biomass is economically feasible (Table 5). Under all of the assumptions considered for the analyses, the project would become economically successful by reducing either the operating or capital expenses by 25 and 50%, respectively or possibly less. In particular, reducing both operating and capital expenses by 25% yields 99% probability of economic success, when considered 70-30% case for fuel and squalene, respectively. Reduction of capital expenses has much bigger impact on the chance of economic success when compared to reductions of operating expenses. When 75% of extracted squalene is refined and sold as a biofuel, economic success can only be obtained if capital expenses are reduced by 50% and operating expenses are reduced by at least 25%. Therefore, under these circumstances, biofuel production from tobacco biomass may have a potential in the long run. Table 5. Probabilities (in %) of obtaining a positive net present value by reducing the capital and operating expenses (CAPEX and OPEX) by different percentages, joint market. OPEX reduction (%)

0 25 50

70% fuel – 30% squalene CAPEX reduction (%)

75% fuel – 25% squalene CAPEX reduction (%)

0 0 54

13 99 100

100 100 100

0 0 0 0

International Food and Agribusiness Management Review

157

0 0 95

62 100 100

Maisashvili et al.

Volume 19, Issue 4, 2016

5. Conclusions The tobacco industry has been declining over the last several years. The removal of government support programs accelerated the reduction of tobacco farms and harvested acreage by over 82 and 30%, respectively. The objective of the study was to estimate the economic and financial feasibility of biofuel production and high value squalene production from tobacco plant leaves. Stochastic simulations were conducted to estimate the probability of profitable production of finished fuels and high value squalene from tobacco plant leaves. The biomass yield and squalene price were considered critical variables for these simulations. Stochastic pro-forma financial statements were simulated for a 10 year time horizon and NPV was calculated to assess overall feasibility of the project. The results showed that if the only output produced from the biomass is used to produce a finished motor fuel, the project is not economically viable, at least under current available technologies. Even with the strongest assumptions of 20% squalene per dry matter and an average yield with a 6% annual growth rate, the economic success of renewable fuel production from tobacco plant leaves is not economical. The average net cash income, the average EC, and the average ending net worth were negative for the 10-year time horizon. Probability of economic success, i.e. Prob.(NPV>0), was zero, meaning that use of tobacco leaves to only produce biofuel is not economically attractive, under the assumptions made in this study. Given the clinical applications of high value squalene, the hydrocarbon does not have to be refined for a fuel production. Nine scenarios were analyzed to assess the economic feasibility of producing high value squalene from tobacco leaves. Considering the project is economically feasible if the probability of obtaining a positive NPV is at least 95%, three scenarios out of nine yield favorable economic outcomes (SC5P5, SC5P0, SC6P0). The results showed that if the average squalene price remains unchanged, over the next 10 years, and squalene content reaches 20% per dry matter either in 5 or in 6 years (SC5P0, SC6P0), the project is economically feasible. Moreover, if squalene value drops by 5% per year, on average, and squalene content per dry matter reaches 20% in 5 years (SC5P5), the probability of economic success is about 99%. However, if squalene value drops, on average, by 10% per year, or the squalene content reaches 20% per dry mater in 8 years, the project is not economically feasible. It has been demonstrated that tobacco-based renewable fuel production can become economically viable by improving certain agronomic and technological factors without any government support. In particular, if biofuel production is the main target, some portion of squalene should still be marketed as a high value product. Even under these circumstances, economic feasibility can only be achieved under certain conditions, requiring reductions of capital and operating expenses by at least 25%. The results are relevant to potential investors who are considering tobacco as an alternative feedstock, and to farmers. We demonstrate that, under the assumptions made in the study, using tobacco as a biofuel feedstock, regardless of its high biomass potential, is still in its early stages and requires further technological and agronomic improvements before commercialization. Future research should focus on new insights regarding yield and biomass improvements, including analyses of squalene market with an updated price information, and reducing cost of processing.

6. Limitations of the study Although the empirical framework of the analysis provided useful implications, those results, however, suffer from important limitations that must be acknowledged and discussed. First, the study was based on lab experiments and bench scale extraction information with extrapolations of the processes to scale. Even though data are limited and assumptions must be made, studies such as this are necessary to evaluate alternative feedstocks, and to estimate how conditions will have to change to gain profitability and adoption. International Food and Agribusiness Management Review

158

Maisashvili et al.

Volume 19, Issue 4, 2016

Second, the study did not incorporate any possible governmental incentives. For example, possible renewable identification number credit and the second generation biofuel producer tax credit could be incorporated into the study. The RIN credit is a price mechanism that ensures the compliance of the renewable fuel standard by obligated parties (i.e. refiners, importers and blenders) and directly translates into a price premium for biofuel producers. To the best of authorsâ&#x20AC;&#x2122; knowledge, currently there are no renewable identification number prices published specifically for tobacco biofuel as of May, 2016. Third, we assumed a just in time delivery system and ignored any storage costs associated with either biofuel or squalene production. We also assumed that the SFE-plant was built just before the first year of operations. Fourth, we assumed that squalene from the biomass is extracted using a SFE process, and the cost associated with this type of extraction was simulated in the study. Different research laboratories, as mentioned in the article, use their own extraction technologies. Some of these technologies are patented and the companies do not disclose any detailed information regarding the cost of extraction or the capital costs. Hence, their values maybe higher or lower than the values used in this study.

Supplementary material Supplementary material can be found online at https://doi.org/10.22434/IFAMR2015.0179. Table S1. Deterministic costs of tobacco production based on average yield of 90,000 pounds of green biomass per acre. Table S2. Cost breakdown of squalene extraction based on average yield of 90,000 pounds of green biomass per acre. Table S3. Cost breakdown of gasoline and diesel as of May 2016.

References Adrianov, V., N. Borisjuk, A. Pogrebnyak, J. Brinker, S. Dixon and J. Spitsin. 2009. Tobacco as a production platform for biofuel: overexpression of arabidopsis DGAT and LEC2 genes increases accumulation and shifts the composition of lipids in green biomass. Plant Biotechnology Journal 8: 277-287. Aleksandrov, N., R. Espinoza and L. Gyurko. 2013. Optimal oil production and the world supply of oil. Journal of Economic Dynamics and Control 37: 1248-1263. Anderson, G.E. 2011. Edible oil processing, solvent extraction. Available at: http://tinyurl.com/zthv4wo. Banerjee, A., R. Sharma, Y. Chisti, U.C. Banerjee. 2002. Botryococcus braunii: a renewable source of hydrocarbons and other chemicals. Critical Reviews in Biotechnology 22: 246-279. Bhattacharjee, P. and R.S. Singhal. 2003. Extraction of squalene from yeast by supercritical carbon dioxide. World Journal of Microbiology and Biotechnology 19: 605-608. Bhattacharjee, P., D. Chattarjee and R.S. Singhal. 2012. Supercritical carbon dioxide extraction of squalene from Amaranthuspaniculatus: experiments and process characterization. Food Bioprocess Technology 5: 2506-2521. Bracmort, K. Meeting the renewable fuel standard (RFS) mandate forcellulosic biofuels: questions and answers. Congressional Research Service Available at: http://tinyurl.com/hhfwpnq. Brito, L.A., M. Chan, B. Baudner, S. Gallorini, G. Santos, D.T. Oâ&#x20AC;&#x2122;Hagan, and M. Singh. 2011. An alternative renewable source of squalene for use in emulsion adjuvants. Vaccine 29: 6262-6268. Capehart, T. 2004. Long-Lived Tobacco program to end. USDA. Available at: http://tinyurl.com/j3nbphp. Cathpole, O.J., J.C. von Kamp, and J.B. Grey. 1997. Extraction of squalene from shark liver oil in a packed column using supercritical carbon dioxide. Industrial and Engineering Chemistry Research 36: 4318-4324. Ciriminna, R., V. Pandarus, F. Beland, and M. Pagliaro. 2014. Catalytic hydrogenation of squalane to squalene. Organic Process and Research Development 18: 1110-1115.

International Food and Agribusiness Management Review

159

Maisashvili et al.

Volume 19, Issue 4, 2016

Hillen, L.W., G. Pollard, L.V. Wake and N. White. 1982. Hydrocracking of the oils of Botryococcus braunii to transport fuels. Biotechnology and Bioengineering 24: 193-205. Hull, J.C. 2006. Options, futures, and other derivatives. Pearson Education, Noida, India. Huntrods, D. 2012. Tobacco profile. Agricultural Marketing Resource Center. Iowa State University, Iowa, IA, US. Giannelos, P.N., F. Zannikos, S. Stournas, E. Lois and G. Anastopoulos. 2002. Tobacco seed oil as an alternative diesel fuel: physical and chemical properties. Industrial Crops and Products 16: 1-9. Gonzalez, A.M., A.M.S. Roque and J. Garcia-Gonzalez. 2005. Modeling and forecasting electricity prices with input/output hidden Markov models. IEEE Transactions on Power Systems 20: 13-24. Johnson, M.D., C. Rutland, J.W. Richardson, J. Outlaw and C. Nixon. 2016. Greenhouse gas emissions from U.S. grain farms. Journal of Crop Improvement 30: 447-477. Kelly, G.S. 1999. Squalene and its potential clinical uses. Alternative Medicine Review 4: 29-36. Kim, S.K. and F. Karadeniz. 2012. Biological importance and applications of squalane and squalene. Marine Medicinal Foods 65: 224-233. Kwak, Y.H. and L. Ingall. 2007. Exploring Monte Carlo simulation applications for project management. Risk Management 9: 44-57. Liu, Z., I. Atlman, I. and G.T. Johnson. 2014. The feasibility of co-firing in Missouri. Biomass and Bioenergy 69: 12-20. Markets and Markets. 2016. Squalene market by type (animal-sourced and vegetable-sourced) and end-use industry (cosmetics, pharmaceuticals and food) â&#x20AC;&#x201C; Global Forecasts to 2021. Available at: http:// tinyurl.com/gv26n58. McAloon, A., F. Taylor, W. Yee, K. Ibsen and R. Wooley, 2000. Determining the cost of producing ethanol from corn starch and lignocellulosic feedstocks. National Renewable Energy Laboratory. Golden, CO, USA. Mercer, P. and R. Armenta. Developments of oil extraction from microalgae. 2011. European Journal of Lipid Science and Technology 1: 1-9. Monge, J.J., L.A. Ribera, J.L. Jifo and J.A. da Silva. 2014. Economics of lignocellulosic ethanol production from energy cane and sweet sorghum in South Texas. Journal of Agricultural and Applied Economics 46: 457-485. Mostafei, H., S. Kordnoori and M. Ostadrahimi. 2011. Modelling the fluctuations of brent oil prices by a probabilistic Markov chain. Journal of Computations and Modeling 1: 17-26. Mundell, R.E. and O. Chambers. 2011. Evaluation of Nicotina. Kentucky Tobacco Research and Development Center for the production of plant-made pharmaceutical and industrial materials, University of Kentucky, Lexington, KY, USA. North Carolina State University Cooperative Extension. 2012. Flue-cured tobacco guide. Availabe at: http:// tinyurl.com/7wqyrdn. Outlaw, J.L., L.A. Ribera, J.W. Richardson, J. da Silva, H.L. Bryant and S.L. Klose. 2007. Economics of sugar-based ethanol production and related policy issues. Journal of Agricultural and Applied Economics 39: 357-363. Palma, M.A., J.W. Richardson, B.E. Roberson, L.A. Ribera, J.L. Outlaw and C. Munster.2011. Economic feasibility of a mobile fast pyrolysis system for sustainable bio-crude oil production. International Food and Agribusiness Management Review 14: 1-16. Reverchon, E. 1997. Supercritical fluid extraction and fractionation of essential oils and related products. Journal of Supercritical Fluids 10: 1-37. Rezende M. and J.W. Richardson. 2015. Economic feasibility of sugar and ethanol production in brazil under alternative future prices outlook. Agricultural Systems 138: 77-87. Richardson, J.W., B. Herbst, J. Outlaw and C. Gill. 2007a. Including risk in economic feasibility analyses: the case of ethanol in Texas. Journal of Agribusiness 25: 115-132. Richardson J.W and M. Johnson. 2014. Economic viability of a reverse engineered algae farm (REAF). Algal Research 3: 66-70. Richardson J.W. and M. Johnson. 2015. Financial feasibility analysis of NAABB developed technologies. Algal Research 10: 16-24. International Food and Agribusiness Management Review

160

Maisashvili et al.

Volume 19, Issue 4, 2016

Richardson, J.W., M. Johnson, R. Lacey, J. Ayler and S. Capareda. 2014b. Harvesting and extraction technology contributions to algae biofuels economic viability. Algal Research 5: 70-78. Richardson, J.W., M.D. Johnson and J.L. Outlaw. 2012. Economic comparison of open pond raceways to photo bio-reactors for profitable production of algae for transportation fuels in the southwest. 2012. Algal Research 1: 93-100. Richardson, J.W., M.D. Johnson, X. Zhang, P. Zemke, W. Chen and Q. Hu. 2014a. A financial assessment of two alternative cultivation systems and their contributions to algae biofuel economic viability. Algal Research 4: 96-104. Richardson, J.W., S.L. Klose, S.L. and W.A. Gray. 2000. An applied procedure for estimating and simulating multivariate empirical (MVE) probability distributions in farm-level risk assessment and policy analysis. Journal of Agricultural and Applied Economics 32: 299-315. Richardson, J.W., W. Lemmer, and J. Outlaw. 2007b. Bio-ethanol production from wheat in the winter rainfall region of South Africa: a quantitative risk analysis. International Food and Agribusiness Management Review 10: 181-204. Richardson, J.W., J.L. Outlaw and M. Allison. 2010. The economics of micro algae oil. AgBioForum 13: 119-130. Rizvi, S.S.H., J.A. Daniels, A.L. Benado and J.A. Zollweg. 1986. Supercritical fluid extraction: operating principles and applications. Food Technology 57: 57-64. Shakhes, J., M.A. Marandi, F. Zeinaly, A. Saraian and T. Saghafi. 2011. Tobacco residuals as promising lignocellulosic materials for pulp and paper industry. BioResources 6: 4481-4493. Tracy, N.I., D.W. Crunkleton and G.L. Price. 2011. Catalytic cracking of squalene to gasoline-range molecules. Biomass and Bioenergy 35: 1060-1065. U.S. Energy Information Administrationâ&#x20AC;&#x2122;s (EIA). 2016. Weekly petroleum status report. Available at: http:// tinyurl.com/hovgewl. United States Department of Agriculture (USDA). 2006. Tobacco Yearbook. Available at: http://tinyurl. com/ho4xt2f. United States Department of Agriculture (USDA). 2012. Census of agriculture. United States summary and state data. Available at: http://tinyurl.com/jm2u4xe. United States Department of Agriculture (USDA). 2016. National agricultural statistical service. 2016. Available at: http://tinyurl.com/zoekjbh. Usta, N. 2005. Use of tobacco seed oil methyl ester in a turbocharged indirect injection diesel engine. Biomass Bioenergy 28: 77-86. Usta, N., B. Aydogan, A.H. Con, E. Uguzdogan and S.K. Ozkal. 2011. Properties and quality verification of biodiesel produced from tobacco seed oil. Energy Conversion and Management 52: 2031-2039. Vazquez, L., C.F. Torres, T. Fornari, F.J. Senorans and G. Reglero. 2007. Recovery of squalene from vegetable oil sources using countercurrent supercritical carbon dioxide extraction. Journal of Supercritical Fluid 40: 59-66. Virginia Cooperative Extension. 2012. Flue-cured tobacco guide. Available at: http://tinyurl.com/gpw58do.

International Food and Agribusiness Management Review

161

OPEN ACCESS International Food and Agribusiness Management Review Volume 19 Issue 4, 2016; DOI: 10.22434/IFAMR2015.0195 Received: 26 October 2015/ Accepted: 26 September 2016

Ethanol and sugarcane expansion in Brazil: what is fueling the ethanol industry? CASE STUDY Ana Claudia Sant’Anna a, Aleksan Shanoyanb, Jason Scott Bergtoldc, Marcellus M. Caldasd, and Gabriel Grancoe aPhD.

candidate, Kansas State University, Department of Agricultural Economics, 400 Waters Hall, Manhattan, KS 66506, USA

bAssistant

Professor, Kansas State University, Department of Agricultural Economics, 304G Waters Hall, Manhattan, KS 66506, USA

cAssociate

Professor, Kansas State University, Department of Agricultural Economics, 307 Waters Hall, Manhattan, KS 66506, USA

dAssociate

Professor and ePhD. candidate, Kansas State University, Department of Geography, 118 Seaton Hall, Manhattan, KS 66506, USA

Abstract This case study describes Brazilian ethanol industry and strategic issues faced by sugarcane farmers and processors as a result of recent industry expansion into the states of Goias and Mato Grosso do Sul. It provides detailed description of the ethanol supply chain in Brazil from field to market and discusses market drivers influencing the industry. Shaped by government regulations, market liberalization, globalization, and technological change, the Brazilian ethanol industry provides a rich context for learning and applying strategic analysis tools. The case is designed to be used in a graduate or undergraduate agribusiness management or strategic management course. The specific teaching objective for this case is to refine and reinforce students’ understanding of industry analysis and the effect of market drivers on competitive forces in an industry. Students will be expected to conduct an industry analysis and provide strategy recommendations to managers of ethanol plants and farmers. The case study incorporates all of the essential information for students to understand the underlying economics of the ethanol value chain and how the external forces shape strategic growth opportunities. Keywords: sugarcane, ethanol, industry analysis, Brazil, Cerrado, case study, teaching case JEL code: Q13, Q16 Corresponding author: acsantanna@ksu.edu

163

Sant’Anna et al.

Volume 19, Issue 4, 2016

The future growth for transportation in Brazil should be from ethanol, but how long it takes us to get there, I really don’t know. – Soren Schroder, CEO of Bunge Limited1

1. Introduction Sugarcane production has expanded in Brazil from the North-Northeast regions of the country to the CenterSouth over the last decades, becoming a dominant crop in the state of Sao Paulo. Since the 2000s, there has been a significant expansion of sugarcane production into Goias and Mato Grosso do Sul fueling land use change in states historically known for livestock and soybean production (Granco et al., 2015). In order to guide the sugarcane expansion while protecting native biomes, the Brazilian government launched the Sugarcane Agro-ecological Zoning (ZAE CANA) (Barros, 2011). The ZAE CANA maps areas suitable for sugarcane production (Figure 1) accounting for weather and soil conditions, as well as environmental, social and economic aspects (Manzatto, 2009). In the states of Goias and Mato Grosso do Sul this zoning policy identifies over 22.6 million hectares as suitable for growing sugarcane (Manzatto, 2009). This geographic expansion in sugarcane production driven and accompanied by changes in government policies and regulations, technological innovations, domestic and global demand, have created unprecedented competitive dynamics in the Brazilian ethanol industry. Industry players at all levels of the supply chain have been forced to reevaluate their strategies in the face of changing industry dynamics. Managers of ethanol plants have to evaluate strategic implications of geographic growth on competitive dynamics and vertical coordination. Specifically, the first strategic decision faced by managers of mills as they expand into central part of Brazil is whether to install a facility in an area with established sugarcane production and, possibly, compete with existing mills for inputs or to locate in a new

1 Said

at the Goldman Sachs 18th Annual Agribusiness Conference in New York City in 2014. Quoted by Oil Price Information Service (OPIS).

Goias Mato Grosso do Sul Sao Paulo ZAE Cana Federal District

Figure 1. Map of Brazil with the states of Goias, Mato Grosso do Sul and Sao Paulo and the Sugarcane Agro-ecological Zoning (ZAE CANA) (adapted from Manzatto, 2009). International Food and Agribusiness Management Review

164

Sant’Anna et al.

Volume 19, Issue 4, 2016

area and have to invest in establishing a new procurement base. The second strategic decision is whether to procure sugarcane from farmers or to backward integrate into sugarcane production by renting or buying land. Farmers in Goias and Mato Grosso do Sul have to reevaluate their strategy in the face of increasing competition for land and labor. Specifically, farmers have to decide whether to become a part of a growing industry and enter into sugarcane production or to continue with their existing production systems. Further, if they decide to join the ethanol supply chain, then they have to make a decision on whether to produce sugarcane and supply to the plant or to rent out the land to the plant instead. In this background the government’s challenge is to guide the sugarcane expansion by minimizing environmental impact while avoiding creation of unnecessary barriers for industry players. Clearly, all of these decisions are interdependent and may result in a wide range of intended and unintended consequences for every player in the industry as evidenced by the comments and remarks of farmers and processors. Farmers in Mato Grosso do Sul have expressed concerns regarding the prices of land and labor. Specifically farmers have commented: ‘the arrival of the mill increases the prices of land’; ‘where the mill comes in, producers are left to fight over the lands that are left over’. Others complained about the shortage of labor after the arrival of the mil: ‘the demand for workers is higher than the supply’2. The management of the mills, on the other hand, considers these concerns as ‘a natural reaction to the arrival of a new crop 3‘ according to Pedro Mizutani, the then president of Cosan Sugar and Ethanol (O Estadão de S. Paulo, 2011). Furthermore, mills believe that through the expansion they ‘bring development and labor opportunities to the region’ in the words of SJC Bioenergia director Ingo Kalder (Siqueira, 2013). Careful industry analysis and the evaluation of strategic implications is warranted for both processors and farmers. Processors with regards to expanding capacity and securing a procurement base. Farmers, with regards to their production systems and marketing strategies.

2. Background and history of the Brazilian ethanol industry Sugarcane has been grown in Brazil since 1532 (BNDES, 2008). The rise in sugar demand and high price in the 16th century led to the first commercial production of sugar in Brazil. Financial support from the Dutch East India Company at that time allowed the sugar industry to rapidly expand. In the mid-17th century, sugar production declined, with the expulsion of the Dutch from the Northeast and consequent expansion of the sugar industry in the Caribbean (BNDES, 2008). Ethanol production began in 1905 along with the first experiments with ethanol fueled vehicles (BNDES, 2008). A maximum blend of 5% ethanol to petrol (E5) was introduced in Brazil in 1931 by the Decree Number 19.717. A year later the Department of Agriculture signed contracts to support sugar-mills in the production of pure alcohol and in 1933 the National Institute for Sugar and Alcohol (IAA) was created. The purpose of the IAA was to regulate and establish standards for the domestic sugarcane industry (Vieira et al., 2007). The IAA protected the industry by setting prices through quotas (Reinhardt et al., 2010). A boost to ethanol production came in the end of the 1960s and beginning of the 1970s. First due to the decline of sugar prices, one of the country’s main exports, that fell from US$1237/t in 1974 to US$ 172/t in 1978 (Melo, 1981). Second, due to oil imports that were equal to half of Brazil’s total value of the exported goods, causing a strain on the economy. In this context, the government launched the Brazilian National Alcohol Program (PROALCOOL) in November 1975 to promote large-scale production of ethanol as a substitute for gasoline (BNDES, 2008).

2 Responses

from survey applied in Mato Grosso do Sul and Goias in 2014 as part of the National Science Foundation (NSF) grant Collaborative Research: direct and indirect drivers of land cover change in the Brazilian Cerrado: the role of public policy, market forces, and sugarcane expansion for more detail see Sant’Anna et al. (2015). 3 Spoken in relationship to the restrictions faced in sugarcane expansion in Jatai, a county in the state of Goias.

International Food and Agribusiness Management Review

165

Santâ&#x20AC;&#x2122;Anna et al.

Volume 19, Issue 4, 2016

The PROALCOOL consisted of three phases: (1) the introduction of production subsidies and establishment of an E20 blend (20% of anhydrous ethanol in gasoline); (2) launch of ethanol-only cars in 1979; and (3) a program phase out through gradual elimination of subsidies. Additional incentives of the program included: providing low interest loans for companies willing to enter the ethanol industry (1980-1985), setting ethanol prices at the pump to 59% of the gasoline price, taxation of gasoline, and reduction of value added tax for ethanol fueled cars. Despite the dictatorial regime, which allowed for strong control over the economy (e.g. price fixing and control of state-owned companies), the implementation of the PROALCOOL was difficult and costly (Zapata and Nieuwenhuis, 2009). De Almeida et al. (2008) estimated that between 1979 and the mid-1990s, the national government spent US$16 billion supporting the ethanol industry. Nevertheless, the program stimulated tangible advances not only in the production of sugarcane (e.g. the introduction of operations research techniques in agricultural management and the use of satellite images for species identification in cultivated areas), but also in the production of ethanol (e.g. energy production for mills via the use of sugarcane bagasse) leading to cost reductions and productivity gains (Goldemberg, 2006). After the end of PROALCOOL program there were shortages in ethanol supply leading to price increase. Consumers reacted by switching to gasoline fueled cars or installing conversion kits (BNDES, 2008). This lead to gradual decline in ethanol demand, which was further exacerbated by a reliable supply of low price petroleum, readily available at every fueling station (Zapata and Nieuwenhuis, 2009). In 1990, the IAA program ended followed by liberalization of ethanol market. The price of anhydrous and hydrous ethanol4, previously controlled by the government, were allowed to be set by the market in 1995 and 1999, respectively (Figueira et al., 2013). Nevertheless, the ethanol industry continued to be regulated even after the price liberalization. In 1997, the Brazilian Natural Petroleum Agency (ANP) was established to encourage industrial competition and to protect consumer interests (BNDES, 2008). In the same year, a private organization of sugar and ethanol producers in the state of Sao Paulo, the union of the sugarcane industry (UNICA), was formed (UNICA, 2015). Together with the Organization of Sugarcane Growers in the Center-South of Brazil (ORPLANA), they formed the council of sugarcane, sugar and ethanol producers of the state of Sao Paulo (CONSECANA), whose role was to consolidate the relationships between the actors in the supply chain and to improve the sugarcane quality control system (Sousa et al., 2012). CONSECANA is also responsible for setting ground rules for sugarcane pricing based on total recoverable sugar (TRS), which is the amount of sugar present in the sugarcane supplied by producers minus the losses occurred during the industrial processing (CONSECANA, 2016.). After the 2000s, a series of events further promoted the expansion of the ethanol sector in the country. These events included: a setting of official blending ratio of anhydrous ethanol with gasoline between 20 and 25%; the introduction of flex-fuel cars in 2003, allowing consumers to choose freely between ethanol and gasoline at the pump; the reduction to zero of the contribution of intervention in the economic domain (CIDE)-fuel tax applied on ethanol in 2004; the reduction of the tax on industrialized products in the sales price of new flex-fuel cars in 2010; and the launch of the Sugarcane ZAE CANA in 2009 (Brasil, 2009; Manzatto, 2009) (Figure 2). Although ethanol production has increased since the year 2000, there has been a significant decrease in production of ethanol from 2011 to 2014 (Figure 2). During this period, mills suffered financial crises. From the 343 mills in operation in 2013-2014, 30 were in debt (Santos et al., 2016). Furthermore, the Brazilian government, with the intent of controlling inflation, has been fixing the prices of gasoline. This policy resulted in unintended negative consequences on the ethanol industry, because the ethanol fuel prices were dictated by market (Granco et al., 2015). This meant that ethanol prices needed to remain low, in order to remain competitive, at a time when production costs were increasing 11.5% per annum5 (Santos et al., 2016). Therefore, the price of oil has oscillated more than that of gasoline and ethanol providing stability to the consumer but not to the producer (Santos et al., 2016). Another factor for the reduction in ethanol production 4 Anhydrous ethanol is used as an additive in gasoline while hydrous ethanol is used as E100 in ethanol run cars or flex-fuel cars (BNDES, 2008, Granco et al., 2015). 5 Depending on the region production costs increased from 5.5 to 11.5% per year (Santos et al., 2016)

International Food and Agribusiness Management Review

166

Sant’Anna et al.

Volume 19, Issue 4, 2016

2009: ZAE CANA program

8,000

Millions of gallons

7,000 6,000 5,000 4,000

2004: CIDE tax exemption on ethanol

1989: PROALCOOL ends

2010: Reduction of IPI tax on flex-fuel cars

1979: launch of ethanol cars

3,000

1975: 2,000 PROALCOOL starts 1000

1997: ANP and UNICA created

19 69 19 /70 71 19 /72 73 19 /74 75 19 /76 77 19 /78 79 19 /80 81 19 /82 83 19 /84 85 19 /86 87 19 /88 89 19 /90 91 19 /92 93 19 /94 95 19 /96 97 19 /98 99 20 /00 01 /0 20 2 03 20 /4 05 20 /06 07 20 /08 09 20 /10 11 20 /12 13 /1 4

2003: Flex-fuel cars hit the market

Figure 2. Total ethanol production in Brazil from 1969-1970 to 2013-2014 harvest year and major policy events (adapted from Brasil, 2013; UNICA (http://www.unicadata.com.br)). PROALCOOL = the Brazilian National Alcohol Program; ANP = Brazilian Natural Petroleum Agency; UNICA = the union of the sugarcane industry and ZAE CANA = sugarcane agro-ecological zoning; CIDE = contribution of intervention in the economic domain; IPI = tax on industrial goods. came from the reduction in the amount of funds available for loans through the National Development Bank (BNDES). From 2010 to 2011 government disbursement of loans fell by 28% and then furthermore by 36% in the following year (Santos et al., 2016).

3. The Brazilian ethanol industry in 21st century Between 2000 and 2012 over 85% of the world’s annual supply of ethanol came from the United States and Brazil (EIA, 2015). Europe is in the third place supplying around 5% followed by Canada supplying 2% of the world’s ethanol production. Brazil was the largest ethanol producer in the world in the period from 2000 to 2005, but since 2007 the USA has surpassed Brazil, increasing its ethanol production to more than double that of Brazil (EIA, 2015). This increase in US production has been attributed to a Renewable Fuels Standard implemented in 2005 and expanded in 2007. In 2014, ethanol consumption in the USA and Brazil was 13.47 billion gallons and 6.76 billion gallons, respectively, while the ethanol production was 14.34 billion gallons and 7.56 billion gallons, respectively (EIA, 2015; Gomes, 2015). In 2008, the sugar-energy sector in Brazil accounted for approximately 2% of the country’s gross domestic product (GDP) (Neves et al., 2010), making it an important economic activity affecting job and income generation at all stages of the supply chain. Ethanol supply chain in Brazil The ethanol supply chain consists of inputs, sugarcane production, processing, and distribution (Figure 3). Major inputs for production of sugarcane include: land, fuel, labor, fertilizers, soil nutrients, machinery including harvesting machines, and labor, among others. Sugarcane grows in 5 year cycles. Yields vary according to the number of cuts (harvests). Higher yields can be achieved in the first cut after which yields decreasing with each cut. The average annual yield for Brazil in 2014 was 70 tons per hectare (IBGE, 2014). Sugarcane harvesting occurs from August to April in the Northeast Region of Brazil and from April to December in the Center-South (BNDES, 2008). Traditionally, sugarcane harvesting involves the burning of the crop followed by the manual harvest of the stalks. This practice has been gradually replaced by mechanical harvesting, which accounted for 55% of the planted area with sugarcane in 2013 (CONAB, 2013).

International Food and Agribusiness Management Review

167

Sant’Anna et al.

Volume 19, Issue 4, 2016 Inputs

Sugarcane production

• chemicals • machinery • labor • fuel • others • land

Processing

Output

40% produced by farmers

35% ethanol-only mills

40% ethanol

60% produced by mills

65% mixed mills (ethanol and sugar)

60% sugar

5% sugar-only mills

Figure 3. Stages of the supply chain (adapted from Neves et al., 2010 and Brasil, 2013). Around 40% of the processed sugarcane in Brazil is supplied by farmers, while the rest is produced directly by the mills on the land owned or rented by them. The average yield for farmers is 75 t/ha while that for mills is 81 t/ha (Crago, 2010). The price of sugarcane is determined by the method proposed by CONSECANA and has high variation over time (Figure 4). Once harvested, sugarcane is transported immediately to the mills in order to prevent saccharose losses. The saccharose content in the sugarcane impacts the conversion rates to sugar or ethanol and tends to decline after the harvest. Consequently, to avoid losses in saccharose content during transportation from fields to plant, mills cannot be located more than 50 km from sugarcane producers (Neves et al., 1998). The usual transportation system used are trucks with a cargo capacity ranging between 15 to 60 tons. A few companies use the waterway transporting system (BNDES, 2008). There are three types of mills in Brazil: sugar mills, ethanol mills, and mixed mills producing both sugar and ethanol. A ton of sugarcane produces about 140 kg of sugar or 86 liters of ethanol (State of Sao Paulo Government, 2014). In 2011-2012, 49% of the TRS derived from the sugarcane production, was used to produce sugar and the rest to produce ethanol (Brasil, 2013). The cost of producing a liter of ethanol in Brazil varies between US$ 0.23 and US$ 0.29 (Crago et al., 2010). Distilleries can store on average 3 billion gallons of ethanol (Zanão, 2009). Hydrous

Anhydrous

TRS 0.25

1.60 1.40

Prices (US$/l)

1.00

0.15

0.80 0.10

0.60 0.40

0.05

0.20 0.00

11 /

0.00

11 /

TRS prices (US$/kg)

0.20

1.20

Figure 4. Monthly producer prices for hydrous and anhydrous ethanol per liter and sugarcane total reduced sugars (TRS) per kilo in the state of Sao Paulo, from 2002 to 2014 (adapted from CEPEA, 2015; UDOP, 2014).

International Food and Agribusiness Management Review

168

Santâ&#x20AC;&#x2122;Anna et al.

Volume 19, Issue 4, 2016

The mills sell ethanol to distributers, a small number of firms are responsible for the retail of ethanol to consumers (Reinhardt et al., 2009). Since the 1990s, gas stations in Brazil are not required to have exclusive contracts with distributers. In 1996, Ordinance number 59/96 from the Treasury Department in Brazil deregulated the sales prices of distributers and ethanol dealers in the whole country. Later in 1999, Ordinance number 28/99 from the Treasury Department and from the Ministry of Mines and Energy deregulated the fuel prices applied to final consumers. Producer prices for anhydrous ethanol exceed those for hydrous ethanol. Hydrous ethanol prices in Sao Paulo ranged from US$ 0.19 /l in November 2003 to US$ 0.48 /l in December 2014, while, in the same period, anhydrous ethanol prices ranged from US$ 0.21 /l to US$ 0.53 /l (Figure 4). In 2014, three firms controlled around 60% of the retail ethanol market (SINDICOM, 2015). Capacity, total production, and trends The number of mills in Brazil grew by 171% between 2000 and 2013, reaching a total processing capacity of 3.6 million metric tons of sugarcane per day (Reinhard et al., 2009; Santâ&#x20AC;&#x2122;Anna et al., 2015). In 2013, there were approximately 10 million hectares planted with sugarcane producing 768 million tons of sugarcane with a production value of US$ 19 billion6 (IBGE, 2014). In the Center-South region of the country, total annual ethanol production in 2012 was 5.6 million gallons (Brasil, 2013). The sugar and ethanol production generates around 700,000 direct jobs and 200,000 indirect jobs (De Almeida et al., 2008). Nevertheless, in 2014 the share of the GDP from the sugar-energy sector has declined compared to preceding years, reaching US$ 26.7 billion from US$ 42.9 billion in 2010 (Barros, 2015)7. From the productivity perspective, there have been significant improvements over the last three decades. The productivity of ethanol per ton of sugarcane in 2012 was five times of that in 1975 (Brasil, 2013). By the end of the last decade, on average, a single mill crushed around 2 million tons of sugarcane per year, originating from approximately 30,000 hectares of land, to produce 170 to 200 million liters of ethanol (Reinhardt et al., 2009). From 1980 to 2012, the total amount of sugarcane crushed by mills in Brazil grew more than threefold, going from 170 million tons to 560 million tons (Figure 5). This growth was accompanied with a shift in the sourcing of feedstock, from the majority produced and supplied by farmers to the majority of sugarcane produced on the land operated by the mills (Neves et al., 1998).

6 Exchange 7 Exchange

rate of 2013 US$ 1.00=R$ 2.25. rate for 2014 was US$ 1.00=R$ 2.451 and for 2010 was US$ 1.00=R$ 1.838.

Own

Supplier

Total

700 600

Million tons

500 400 300 200 100

4 /1

11 /

09 20

/0 07

05 20

2 /0

03 20

0 /0

01 20

97 19

95 19

2 /9

93 19

0 /9

91 19

87 19

/8 85

83 19

79 19

Figure 5. Yearly amount of sugarcane crushed by origin (supplier or produced by the mill) (adapted from Brasil, 2013, 2015). International Food and Agribusiness Management Review

169

Santâ&#x20AC;&#x2122;Anna et al.

Volume 19, Issue 4, 2016

From 2008 to 2014 more than 40 mills closed due to financial problems, however the number of new mills continued to grow (Barros, 2014). Most of the newer mills are designed to produce both sugar and ethanol from sugarcane. In early 2000s the majority of the TRS was directed towards sugar production, but since 2006, more has been directed toward ethanol production (Figure 6). This change has been driven by public policies and market drivers discussed in more detail in the following sections. During the 21st century, Brazil increased its ethanol production capacity by expanding into the Cerrado region, located in the center of the country. Although most of the production still comes primarily from the state of Sao Paulo, the states of Goias and Mato Grosso do Sul have increased their contribution to the Brazilian supply of sugarcane fivefold between 2000 and 2013 (IBGE, 2014)8. This expansion resulted from cheaper and flatter land in these two states, which allowed for easy expansion and greater mechanization (Granco et al., 2015). It has been argued that the expansion of sugarcane in Brazil was driven by international demand for sugar and ethanol and national policies promoting ethanol production and commercialization. Additionally, other factors such as technological changes in production (e.g. development of mills that produce both ethanol and sugar) and the vertical integration in the industry played a role in the sugarcane expansion (GĂźnther et al., 2008).

4. Market drivers shaping the Brazilian ethanol industry Government programs The ethanol production in Brazil has been stimulated and supported by a number of government programs and policies including the PROALCOOL (Goldemberg, 2006). In addition to eliminating the CIDE-fuels tax, applied on ethanol and the lower tax rates for purchasing flex-fuel cars, in 2013, the government exempted ethanol producers, distributers and importers from paying their contribution to the Social Integration Program for financing social security (Barros, 2014). The advantages brought on by the elimination of the CIDE-fuel tax, however, began to decline in mid-2008 when the same tax applied on gasoline was reduced gradually. The CIDE-fuel tax on gasoline was reduced to zero in mid-2012 returning only in May 2015 (Ramos, 2016). On the regional and local level in center west, the state governments of Goias and Mato Grosso do Sul have provided fiscal incentives: the PRODUZIR in Goias allowes a grace period for paying 73% of the tax on the 8 Although the state of Minas Gerais (MG) is also an agricultural frontier for sugarcane it was not added to this study. MG has a smaller area with Cerrado and a smaller increase in sugarcane production than the states of Goias (GO) and Mato Grosso do Sul (MS). Sugarcane production in GO and MS has increased 474 and 546% from 2000 to 2012, respectively, whereas in MG it has increased 276% (IBGE, 2014). Also 61% of MS and 97% of GO are covered with Cerrado while 57% of MG is covered with Cerrado (Sano et al., 2008).

Sugar

Ethanol

60% 50% 40% 30% 20% 10%

4 /1 13

3 20

/1 12

11 /

/1 1

9 20

08 20

06 20

05 20

/0 04

4 20

/0 03

3 20

/0 02

2 20

01 20

Figure 6. Yearly destination of total reduced sugars of sugarcane towards sugar and ethanol between 2000 and 2012 (Brazil, 2013). International Food and Agribusiness Management Review

170

Santâ&#x20AC;&#x2122;Anna et al.

Volume 19, Issue 4, 2016

circulation of goods and services (ICMS) until 2020 while in Mato Grosso do Sul, the MS Empreendedor program exempts the payment of 67% of the ICMS tax for the industry (Granco et al., 2015). Historically, the Brazilian government has controlled the price of gasoline in order to help curb inflation. This policy has hurt the ethanol industry, as gasoline prices have artificially been maintained at a rate lower than its international prices, while ethanol prices were determined by the market (Freitas and Dezem, 2015). Consequently, for consumers gasoline was more price efficient than ethanol9. The policy of freezing gasoline prices along with that of reducing the CIDE-fuel tax on gasoline has been blamed as responsible for the crises to the ethanol sector. Since June 2014, with the rapid decline in the price of oil and global gasoline prices, the controlled price of gasoline in Brazil became higher than the international price, benefitting the ethanol industry (Freitas and Dezem, 2015; OECD/IEA, 2015). In 2006, the government launched the Brazilian Agro-energy Plan with the goal of increasing the competitiveness of agro-energy supply chains (MAPA, 2006). The Plan instituted the creation of a new research unit in agroenergy (Embrapa Agroenergy) and was followed by the Sugarcane Agro-ecological Zoning program. In 2007, the government announced a plan to construct two ethanol pipelines in conjunction with the Program for Economic Acceleration. The first is a 1,150 km extension of one of the ethanol pipelines connecting Goias to the port of Santos in Sao Paulo and passing through other major ethanol producing counties. The second pipeline, 900 km in length, will connect Mato Grosso do Sul to the port of Paranagua in the state of Parana (Transpetro, 2007). Currently 206 km of pipeline is in operation and it is estimated that logistic costs will be reduced from US$ 18.7 /m3 for road transportation to US$ 7.65 /m3 for transport using the pipeline (Folha de S. Paulo, 2013). In addition to lenient tax policies and large infrastructure projects, the Brazilian government also provides subsidized loans to the industry through the BNDES. The amount of loans conceded ranged from US$ 0.51 billion in 2000 to US$ 3.39 billion in 201110, 65% of which were provided to industries in the South-East and Center-West (Milanez and Nyko, 2012). Among the BNDES programs for subsidized loans, is the PRORENOVA, directed towards the renewal or expansion of sugarcane fields, and the Paiss, a joint plan of BNDES and the Research and Projects Financier to fund industrial technological innovations aimed at the sugar-energy sector (Barros, 2014). To complement these programs, in 2015, the federal government increased the mandated amount of ethanol to be mixed into gasoline from 25 to 27% (Amato and Matoso, 2015). Technology In the 2000s, the number of patents obtained by the ethanol industry increased significantly (Freitas and Kaneko, 2012). Innovations in sugarcane production included new crop varieties and mechanized sugarcane harvesting, while those in ethanol production included the development of mixed mill technologies enabling concurrent production of sugar and ethanol, as well as energy generation with bagasse, a processing byproduct. Further byproducts of sugarcane are yeasts and additives, carbon credits, bioplastics, and vinasse. These have different uses and markets. For instance, 10% of yeasts recovered from ethanol production can be sold to the livestock feed industry to be mixed into its products (Neves et al., 2010). Mills also sell additives based on the sugarcane yeast. In addition, mills in Brazil sell carbon credits in the market through the clean development mechanism (Neves et al., 2010). Ethanol can also be used in the production of bioplastics and the left over bagasse can be used to generate electricity (Arruda, 2011; Neves et al., 2010). This energy is used in the ethanol and sugar production process and the excess is sold to local counties. The vinasse, a fertilizer substitute high in nitrogen content, can be combined with water to irrigate sugarcane planted land (fertigation), thus reducing fertilizer costs for mills producing their own sugarcane (Neves et al., 2010).

9 Ethanol 10 Values

is less energy efficient than gasoline, so it should be chosen when its price is 70% of that of gasoline. reflect 2011 prices and use the following exchange rate: US$ 1.00=R$ 1.742.

International Food and Agribusiness Management Review

171

Santâ&#x20AC;&#x2122;Anna et al.

Volume 19, Issue 4, 2016

The plant breeding research in Brazil has been driven by both private and public institutions: the Brazilian Agricultural Research Corporation (EMBRAPA), the Agronomic Institute of Campinas (IAC) and Interuniversity network for the development of the sugar and ethanol sector (RIDESA) in the public sector; Center of Sugarcane Technology funded by the sugarcane mill industry (CTC) in the private sector (Torres et al., 2011). Genetic modifications to sugarcane are regulated by the Department of Science and Technology (BNDES, 2008). The CTC expects that by 2018 farmers will be able to grow genetically modified sugarcane. New varieties are expected to be pest resistant, drought tolerant, higher yielding, and with higher sugar content (CTC, 2013). These technological enhancements eased the introduction of mechanized agriculture into the states of Goias and Mato Grosso do Sul which have more conducive geographic characteristics. Nevertheless, there has been a decline in the productivity of sugarcane due to the crises in the ethanol sector, as well as, climactic problems. On the production side, sugarcane producers were facing financial difficulties and were unable to invest in maintaining their sugarcane fields (Ramos, 2016). Additionally, productivity in the Center-South has declined due to technical issues such as the lack of adaptation to harvest mechanization, weather events (such as droughts), aging of sugarcane fields and related technology (Santos, 2016). Another productivity inhibiting factor is the time it takes to approve newly developed higher yield sugarcane varieties, which can take up to years and can delay the time from the technology development to adoption in the field (Santos, 2016). An alternative, which would be the cellulosic ethanol, or second generation ethanol made from the bagasse and straw of the sugarcane, is still inviable at the industrial scale due to enzyme costs (EMBRAPA, 2014). On the consumer side, an important innovation which increased ethanol demand, was the introduction of flex-fuel cars, allowing consumers to switch more freely between gasoline and ethanol (E100). Flex-fuel cars have been so well accepted in the market that its production has grown from 858 thousand cars in 2005 to 2.6 million in 2010, currently representing 90% of all new cars being sold (ANFAVEA, 2015). Global trade Historically, Brazil has been exporting more sugar than ethanol: on average 70% of the sugar produced went to export compared to 10% of the ethanol (UNICA, 2015). In 2014, Brazil exported 369 million gallons at an average value of US$ 2.42 /gallon free on board (UNICA, 2015; SECEX, 2015). From 2000 to 2012 ethanol was exported mainly to five countries: United States, Jamaica, the Netherlands, South Korea and Japan. The amount imported by the United States increased since 2008, replacing exports to Jamaica (Figure 7). United States

Jamaica

The Netherlands

South Korea and Japan

70% 60% 50% 40% 30% 20% 10% 0%

2014

2012

2010

2008

2005

2002

2000

Figure 7. Major importers of Brazilian ethanol in terms of their participation in the total amount exported (adapted from SECEX, 2015).

International Food and Agribusiness Management Review

172

Sant’Anna et al.

Volume 19, Issue 4, 2016

The removal of US tariffs on Brazilian ethanol imports in 2012 helped to increase exports to the USA. Prior to 2012, hydrous ethanol was shipped to the Caribbean where it was transformed into anhydrous and then shipped to the US (Elobeid and Tokgoz, 2008)11. The sugar and ethanol industries in Brazil have seen major investments by foreign companies since 2008-2009, when the global economic crisis affected Brazilian mills (Oliveira, 2013). By 2013, about half of the sugarcane produced (654 million tons) was supplied by foreign owned mills. This accounted for 33% of the total production of sugar and ethanol in Brazil (Oliveira, 2013). Foreign investors included US firms (e.g. Bunge and Cargil), French firms (e.g. Louis Dreyfus Commodities); Chinese firms (e.g. Noble), and others, including Shell (Oliveira, 2013). It is estimated that foreign direct investment amounted to US$ 22 billion through the purchasing and establishment of Brazilian mills (Oliveira, 2013). However foreign land ownership and rental is limited. The Brazilian law 5709/71 regulates the acquisition of rural properties by foreigners. A judgement by the Attorney General’s Office in Brazil (Judgement AGU/LA-01/2010) allows foreign companies, permitted to function in the country, to acquire up to 100 modules of land12 subject to approval from Congress. This judgement defines as foreign companies those with a foreign director or with the majority of the shares belonging to foreigners. Oil Markets Changes in the oil markets affect ethanol production, not only because ethanol and gasoline are substitute goods, but also because fuels represent 10% or more of the energy input costs (Valdes et al., 2016). Cheaper oil decreases demand for biofuels which, in turn, influences farmers to reallocate resources from producing sugarcane to producing other crops, such as soybeans and corn, or to livestock grazing (Valdes et al., 2016). The increase of land devoted to crops and grazing may lead to a decline in their world prices due to the increase in exports. When oil prices are high, demand for ethanol increases and consequently more land is devoted to sugarcane and less to other crops. If less land is devoted to other crops, such as soybeans, this impacts Brazilian exports of these crops, reducing the world’s supply in, say soybeans, and, thus increasing its world prices (Valdes et al., 2016). Supply and demand for oil plays an important role in determining crude oil prices (Figure 8). Increases in oil demand can occur due to increases in global economic growth, an example being between 2000 and 2008 (Levine et al., 2014). In the second half of 2008 oil prices fell. The causes were the 2008 financial crisis bringing a decline in oil demand, aligned with an increase in oil production by OPEC countries (Levine et al., 2014). OPEC countries cut down on production from 2009 onwards bringing prices slowly up. From 2010 to 2012 there was an increase in oil demand coming mainly from Asia Pacific markets, but there was also a decrease in oil supply due to political instabilities in North Africa, sanctions imposed on Iran and declines in production in the United Kingdom and Norway (Levine et al., 2014). More recently, since June 2014, oil prices have collapsed. This decrease is brought by the imbalance between oil supply and demand. There has been an increase in crude oil supply from non-OPEC countries (OECD/IEA, 2015). On the demand side, a reduction in crude oil demand has been brought by: emerging countries entering a stage of development that demands less oil; environmental concerns, and; increased availability of renewable fuels, among others (OECD/IEA, 2015).

11 Caribbean countries profited from this transaction due to the Caribbean Basin Initiative, which exempts these regions of paying US import tariffs

(Elobeid and Tokgoz, 2008). 12 The size of a module varies by county, considering the smallest and the largest modules, 100 modules of land can vary from 500 to 11,000 hectares (Landau et al., 2012).

International Food and Agribusiness Management Review

173

Santâ&#x20AC;&#x2122;Anna et al.

Volume 19, Issue 4, 2016

120 100

US$/barrel

80 60 40 20

19 97 19 98 19 99 20 00 20 01 20 02 20 03 20 04 20 05 20 06 20 07 20 08 20 09 20 10 20 11 20 12 20 13 20 14 20 15

Figure 8. Evolution of Brent oil prices (US$/barrel) (adapted from ANP, 2007, 2016). Social and environmental issues Production of ethanol from sugarcane generates less greenhouse gas emissions compared to production of ethanol from corn. The life cycle of greenhouse gas emissions from ethanol production in the Brazilian state of Sao Paulo is lower than that in the United States, 1.42-1.5 and 3.6-6.02 kg CO2-eq per gallon respectively (Crago et al., 2010). In Brazil, most of the CO2 is released during the harvest as a result of burning sugarcane fields is prior to manual harvest (BNDES, 2008). The state of Sao Paulo passed the law in 2002 limiting the practice of sugarcane burning with plans to completely eliminate it by 2021. Additionally, the burning of the sugarcane is lessened by the replacement of manual harvesting with mechanized harvesting. This is the case in Cerrado, where an extensive plain surface allows for the use of mechanized harvest. It is estimated that 80% of the harvested sugarcane in Goias is done mechanically and 90% in Mato Grosso do Sul (Conab, 2013). While the increasing use of mechanical harvesting has positive environmental impact, it creates a social problem by reducing labor demand. One mechanical harvester replaces 100 workers (Neves et al., 2010). Land use change After 1933, the sugarcane plantations in the North-Northeast Brazil declined and, sugarcane expanded to the Southeast Brazil (BNDES, 2008). More recently, with the increase in land prices in Sao Paulo, sugarcane has expanded more intensely towards the center-west of Brazil into the states of Goias and Mato Grosso do Sul, where land is relatively cheaper and flatter (Figure 1) (Granco et al., 2015). These shifts are also associated with the development of sugarcane varieties more adaptable to different climates allowing sugarcane to be grown in areas previously seen as less hospitable for this crop. Since 2000, the agricultural enterprise mix in the Central-West of Brazil has changed. This is particularly noticeable in Goias and Mato Grosso do Sul where the percentage of area planted with sugarcane has increased (Figure 9). Although Sao Paulo continues to produce more than half of the Brazilian sugarcane, Goias and Mato Grosso do Sul have increased their share in the nationâ&#x20AC;&#x2122;s sugarcane production from 2% in 2000 to 10% in 2013 (IBGE, 2014). During the same period the number of mills in Goias and Mato Grosso do Sul have doubled (Granco et al., 2015). It is important to highlight that cattle ranching is by far the dominant land use in Cerrado but is share has decreased in recent decades by the expansion of large mechanized agriculture. This crop expansion in Cerrado has already called the attention of scholars due to its effects on land cover change in the region but also for its indirect effect in the deforestation of the Amazon. Evidences exist that increased soybean expansion in Cerrado may had led to the movement of cattle ranching to the Amazon International Food and Agribusiness Management Review

174

Sant’Anna et al.

Volume 19, Issue 4, 2016 Soybeans

Sugarcane

Others

Corn

Sao Paulo (2012) Sao Paulo (2000)

Mato Grosso do Sul (2012) Mato Grosso do Sul (2000)

Goias (2012) Goias (2000)

Brasil (2012) Brasil (2000) 0%

20%

40%

60%

80%

100%

Figure 9. Percentage planted with soybeans, corn, sugarcane and others in 2000 and 2012 in Brazil and selected states (adapted from IBGE, 2014). (Walker et al., 2011; Morton et al., 2006; Mc Manus et al., 2016). In addition, recent study (Adami et al., 2012) demonstrates the fact that sugarcane expansion in Cerrado is also occurring on degraded pasture, thus suggesting that sugarcane could be driving cattle ranching into the Amazon. Although these studies are not conclusive, they have heated the debate in both regions.

5. Ethanol expansion into the Cerrado: strategic implications for supply chain actors Decisions facing farmers and landowners The shift in sugarcane production to the Central part of Brazil as a reaction to market drivers has created unprecedented competitive dynamics in all stages of the Brazilian ethanol supply chain. The arrival of mills in new areas increased the demand for land and labor. One of the farmers in Mato Grosso do Sul complains that ‘the arrival of the mill increased the prices of land’13, while that of Goias explains that ‘with the establishment of mills producers are forced to fight over the leftover land and labor’13. In referring to the fact that the price to rent the land increased from 12 bags of soybeans per hectare to 16 bags of soybeans per hectare. Andre Rocha, the president of the Syndicate of the Industries of Ethanol in the State of Goias in 2011 says ‘Obviously [the land price increase] is not in the interest of some southerners that prefer to rent the land, preferably cheap, to produce grains’ (O Popular, 2011). Statistics show that the salaries received by workers in the agricultural sector have increased from US$ 347 in 2000 to US$ 487 in 2014 in Mato Grosso do Sul and from US$ 326 to US$ 502 in Goias in the same period in 2010 values (RAIS, 2016)14. On that note, cattle farmers in Mato Grosso do Sul commented: ‘I have had difficulties in hiring people to work with cattle ranching. I have had to contract workers from Paraguay.’13 and ‘The arrival of the mill created shortage of workers in cattle ranching. Workers migrated to work at the mills due to the higher salaries they offer’ 13. As a result farmers are contemplating a decision to enter into the sugarcane sector by reallocating land from their current economic activity. In order to make his decision the farmer has to take into account the revenue, costs and profits from each activity (Table 1). The farmer may also choose to have a variety of activities, for example rent out a portion of their land to mills and produce cattle on the rest.

13 Responses

given during the survey applied in Mato Grosso do Sul and Goias in 2014 as part of the NSF grant Collaborative Research: direct and indirect drivers of land cover change in the Brazilian Cerrado: the role of public policy, market forces, and sugarcane expansion (Sant’Anna et al., 2015). 14 Exchange rate used for 2000 was US $1=R$ 1.83. Wages were deflated using CPI index from Eurostat data page.

International Food and Agribusiness Management Review

175

Sant’Anna et al.

Volume 19, Issue 4, 2016

Table 1. Profits from various activities in Goias and Mato Grosso do Sul (FNP, 2014a, 2014b; Lima Filho et al., 2015).1 Activity

Goias (GO)

Mato Grosso do Sul (MS)

soybeans (US$/ha)2 corn rotation (US$/ha)3 cattle (500 UA)4 extensive breeding and fattening extensive breeding extensive – complete cycle sugarcane production5 land rental (US$/ha)6

$ 555.58 $ 22.24

$ 405.69 $ 12.54

$ 96.80 $ 324.52 - $ 463.61

$ 66.18 $ 26.75 $ 43.11 $ 96.80 $ 208.62 - $ 324.52

1 All

values in the table are US$. Exchange rate used is US$1 = R$ 2.157. Profits from soybeans consider a production of 9,000 kg/ha for GO (1,200 ha module) and 2,880 kg/ha for MS (850 ha module). 3 Profits from corn are based on the production of 5,400 kg/ha for MS and 6,000 kg/ha for GO, for the corn planted in between the soybean rotation. 4 UA = animal units. Information on cattle was only available for MS. 5 Information for sugarcane production were based on rentability and production costs in the state of Sao Paulo. 6 Land Rental reflects the average land rental paid per hectare in each state. 2

If the farmer decides to go into the sugarcane sector, there are two main options. Option one would entail producing sugarcane and supplying it to the mill. Second option would entail entering into land lease arrangement with the mill. The option to rent part of the land may be preferred by those who wish to diversify their farm activities without having to invest in adopting a new crop system. As cattle ranchers in Mato Grosso do Sul explain: ‘I rent out degraded land to the mills which helps me maintain the cattle ranching’13; ‘with the increase in the cost of livestock production, I have decided to rent out my land to the mill’13, while another states that ‘A reason to rent out land to sugarcane is to diversify economic activities’13. One of the limiting factors to new crop adoption is distance, sugarcane loses its sugar content rapidly after being harvested, limiting sugarcane supply to farms within a 50 km from the mill. Another limiting factor for adoption is perceived risk and uncertainty, the global crises that occurred in 2008 forced many mills into bankruptcy and closure. As a cattle rancher in Mato Grosso do Sul explains: ‘There is a lack of security and stability in the sugarcane sector’13. As one farmer explained: ‘Farmers are worried that the mills will not pay’13. A farmer that decides to supply sugarcane must also understand how the price is set by CONSECANA. The price of sugarcane received by the supplier is calculated based on: (1) the amount of TRS in the sugarcane supplied; (2) the average prices of the final products; (3) the production mix of the mill; (4) how much the sugarcane represents in the production costs of each final product (Belik et al., 2012). Although this system functions, there are complaints from suppliers. Specifically, suppliers complain about: (1) the fact that the price of sugarcane does not consider the prices of sub-products, such as energy and vinasse; (2) the lack of access to the mills’ decisions in the marketing of the final products; (3) the lack of transparency in the process of quality inspection of the sugarcane; (4) the size of participation of sugarcane on the production costs of the final goods (Belik et al., 2012). For example, one supplier in Mato Grosso do Sul complained about the lack of transparency on the product mix produced by the mill and how the TRS was not calculated for each individual supplier. The transactions between farmers and mills are governed using three types of contracts: land rental, agricultural partnership, and supply contract. With the first two arrangements, the mill is responsible for the production of sugarcane by paying fixed rent or through a sharecropping arrangement respectively. In the case of the supply contract, the producer agrees to supply a predetermined amount of sugarcane to the mill at a certain International Food and Agribusiness Management Review

176

Sant’Anna et al.

Volume 19, Issue 4, 2016

price and schedule. A common feature in this type of contract is the provision of harvesting, hauling and delivery services by the mill at a cost of 30% of the price of a ton of sugarcane (Neves et al., 1998). Decisions facing the ethanol plants Given the distance limitation for procuring sugarcane, mills must decide whether to install a facility in an area with established sugarcane production and, possibly, compete with existing mills for inputs or to locate in a new area and have to invest in establishing new procurement base. In order to secure sufficient supply of raw materials mills have two options: incentivize farmers into growing sugarcane, or backward integrate and organize their own sugarcane production. Both options have their costs and benefits. Backward vertical integration would provide a mill with full control over the supply of feedstock but would require significant capital investments and would expose the business to risks inherent in agricultural production (Neves et al., 1998). Relying on farmers for the supply of feedstock does not require large capital investments in production but increases transaction costs associated with coordination and contract enforcement. Some mills use a combination of own production and contracted supply. One example is the ETH, Odebrecht, in the ‘Polo Araguaia’ in the Central-West, their sugarcane comes partially from land they own or rent and partially from farmers: ‘For ETH, it does not make sense to immobilize its capital in lands. That is why we do an intense work to promote (sugarcane) planting in this area’ said Luiz Pereira de Araujo, ETH’s director of people and sustainability (Freitas, 2010: §6). Irrespective of the type of arrangement, farmers expect mills to invest in building relationships and earning their trust. As one farmer stated ‘If the mill did more for the community maybe it would be easier for it to rent land and to produce sugarcane’15. Mills generally argue that their arrival brings benefits to the community: ‘The business (SJC Bioenergy) brings development and work to the region,’ explains Ingo Kalder, director of SJC Bioenergy in 2013 (Siqueira, 2013: §16). The government also shares this vision, after all, the arrival of the mill has been associated with an increase in social welfare (Sant’Anna et al., 2015). Strategic implications for industry stakeholders Multitude of market drivers are reshaping competitive forces in the Brazilian sugarcane farming and processing sectors with important strategic implications for farmers and ethanol plant managers. The geographic expansion of sugarcane production into states of Goias and Mato Grosso do Sul, historically known for livestock and soybean production, have forced actors at all levels of the ethanol supply chain to reevaluate their strategies. This shift is accompanied and driven by changes in government policies and regulations, technological innovations, as well as, domestic and global demand. The situation warrants careful industry analysis and evaluation of strategic implications for both, the processors with regards to expanding capacity and securing a procurement base, and by farmers, with regards to their production systems and marketing strategies. The industry analysis should provide answers to following questions: ■■ How the changes in industry’s external environment would affect competitive forces in the farming industry? ■■ What are the strategic implications for sugarcane growers in Sao Paulo? ■■ What are the strategic implications for farmers in Goias and Mato Grosso do Sul? ■■ How the changes in industry’s external environment would affect competitive forces in the sugarcane processing industry? ■■ What are the implications for growth and vertical coordination strategy in the processing industry? ■■ What are the strategic implications for new and established ethanol mills in Sao Paulo, Goias and Mato Grosso do Sul?

15 Responses

given during the survey applied in Mato Grosso do Sul and Goias in 2014 as part of the NSF grant Collaborative research: direct and indirect drivers of land cover change in the Brazilian Cerrado: the role of public policy, market forces, and sugarcane expansion (Sant’Anna et al., 2015).

International Food and Agribusiness Management Review

177

Sant’Anna et al.

Volume 19, Issue 4, 2016

Acknowledgements This research was supported by NSF grant, ‘Collaborative Research: Land Change in the Cerrado: Ethanol and Sugar Cane Expansion at the Farm and Industry Scale’ – NSF BCS-1227451.

Supplementary material Supplementary material can be found online at https://doi.org/10.22434/IFAMR2015.0195. Teaching note for ‘Ethanol and sugarcane expansion in Brazil: what is fueling the ethanol industry?’

References Adami, M., B.F.T. Rudorff, R.M. Freitas, D.A. Aguiar, L.M. Sugawara and M.P. Mello. 2012. Remote sensing time series to evaluate direct land use change of recent expanded sugarcane crop in Brazil. Sustainability 4: 574-585. Amato, F. and F. Matoso. 2015. Mistura de etanol na gasolina sobe hoje. G1-Economia. Available at: http:// tinyurl.com/zyqalql. ANFAVEA. 2015. Tabelas Estatísticas. National Association of Automatic Vehicle Producers. Available at: http://tinyurl.com/p9652tf. Annual Relation of Social Information (RAIS). Brazilian Ministry of Work and Employment. Dataset available at: http://tinyurl.com/nfsyoze. Arruda, P. 2011. Perspective of the sugarcane industry in Brazil. Tropical Plant Biology 4: 3-8. Barros, G.S.A.C. 2015. Agronegócio sucroalcooleiro: o que o país está perdendo. Revista Opiniões 12: 8-10 Barros, S. 2011. Brazil: biofuels annual. Gain report number: BR14004. USDA-FAS. Available at: http:// tinyurl.com/oybtpyx. Barros, S. 2014. Brazil: biofuels annual. Gain report number: BR14004. USDA-FAS. Available at: http:// tinyurl.com/hzyydre. Belik, W., L.F. Paulillo and C.E. de F. Vian. 2012. A emergência dos conselhos setoriais na agroindústria brasileira: gênese de uma governança mais ampla? Revista de Economia e Sociologia Rural 50: 9-32. Brasil. 2009. DECRETO No 6.961, de 17 de Stembro de 2009. Aprova o zoneamento agroecológico da cana-de-açúcar e determina ao Conselho Monetário Nacional o estabelecimento de normas para as operações de financiamento ao setor sucroalcooleiro, nos termos do zoneamento. Available at: http:// tinyurl.com/pp4t2cq. Brasil. 2015. Anuário estatístico de agroenergia 2014 statistical yearbook of agrienergy. Brasilia: Ministério da Agricultura, Pecuária e Abastecimento – MAPA/ACS, Recife, PE, Brasil. Brasil. 2013. Anuário estatístico de agroenergia 2012: statistical yearbook of agrienergy. Ministério da Agricultura, Pecuária e Abastecimento – MAPA/ACS, Recife, PE, Brasil. Brazilian Institute of Geography and Statistics (IBGE). 2014. Produção Agrícola Municipal. Dataset available at: http://tinyurl.com/o7mbvvt. Brazilian National Agency for Petroleum, Natural Gas and Biofuel (ANP). 2007. Anuario Estatistico 2007. ANP, Brasilia, Brazil. Brazilian National Agency for Petroleum, Natural Gas and Biofuel (ANP). 2016. Anuario Estatistico 2016. ANP, Brasilia, Brazil. Brazilian Agricultural Research Corporation (EMBRAPA). 2014. Etanol 2G depende do aprimoramento de enzimas. Available at: http://tinyurl.com/hv8rfeb. Brazilian National Development Bank (BNDES). 2008. Sugarcane-based bioethanol: energy for sustainable development. Coordination BNDES and CGEE – BNDES, Rio de Janeiro, Brasil. Crago, C. L., M. Khanna, J. Barton, E. Giuliani and W. Amaral. 2010.‘Competitiveness of Brazilian sugarcane ethanol compared to US corn ethanol. Energy Policy 38: 7404-7415. Center for Advanced Studies on Applied Economics (CEPEA). 2015. Ethanol Index Sao Paulo state. Available at: http://tinyurl.com/hrkedq6. International Food and Agribusiness Management Review

178

Sant’Anna et al.

Volume 19, Issue 4, 2016

Centro de Tecnologia Canavieira (CTC). 2013. Inovação para o future da energia. Available at: http://tinyurl. com/hchsd7t. Companhia Nacional de Abastecimento (CONAB). 2013. Perfil do Setor do Açucar e do Etanol no Brasil. Vol. 5. Available at: http://tinyurl.com/z332xgh. CONSECANA. 2016. Formação e implantação do CONSECANA-SP. Available at: http://tinyurl.com/p6bu577. De Almeida, E. F., J. V. Bomtempo and C. M. De Souza e Silva. 2008. The performance of Brazilian biofuels: an economic, environmental and social analysis. In: Biofuels – linking support to performance, edited by OECD/ITF. OECD, Paris, France, pp. 151-188. De Freitas, L.C. and S. Kaneko. 2012. Is there a causal relationship between ethanol innovation and the market characteristics of fuels in Brazil? Ecological Economics 74: 161-168. De Melo, H.F. 1981. Proalcool, energia e transportes. FIPE, Sao Paulo, Brazil. Dos Santos, G.R. 2016. Produtividade na agroindústria canavieira: um olhar a partir da etapa agrícola. In: Quarenta anos de etanol em larga escala no Brasil: desafios, crises e perspectivas, edited by G.R. dos Santos. IPEA, Brasilia, Brazil, pp.165-186. Dos Santos, G.R., E.A. Garcia, P.F.A. Shikida and D.J. Rissardi. 2016. A agroindústria canavieira e a produção de etanol no Brasil: características, potenciais e perfil da crise atual. In: quarenta anos de etanol em larga escala no Brasil: desafios, crises e perspectivas, edited by G.R. dos Santos. IPEA, Brasilia, Brazil, pp. 17-46. Elobeid, A. and S. Tokgoz. 2008. Removing distortions in the U.S. ethanol market: what does it imply for the United States and Brazil? American Journal of Agricultural Economics 90: 918-932. Figueira, S.R.F., B.B. Perosa and W. Belik. 2013. Impacto da desregulamentação e da concorrência. AgroAnalisys 33: 23-25. FNP. 2014a. Agrianual 2014. AGRA FNP Pesquisas Ltda, Sao Paulo, Brazil. FNP. 2014b. Anualpec 2014. AGRA FNP Pesquisas Ltda, Sao Paulo, Brazil. Folha de S. Paulo. 2013. Alcoolduto ainda não reduz no de caminhões na rodovia. Available at: http://tinyurl. com/qa382ll. Freitas, G. 2010. Um mar de etanol. Carta Capital. Available at: http://tinyurl.com/jzhhjzf. Freitas, G. and V. Dezem. 2015. Ethanol mills benefit from Brazil gas-price control: commodities. Bloomberg, 21 January 2015. Available at: http://tinyurl.com/j3g2yjv. Goldemberg, J. 2006. The ethanol program in Brazil. Environmental Research Letters 1: 1-5. Gomes, J.R. 2015. Produção de etanol no brasil atinge recorde de 28,6 bi de litros em 2014. O Estadão de S. Paulo. Available at:http://tinyurl.com/ntery4p. Granco, G., M.M. Caldas, J.S. Bergtold and A.C. Sant’Anna. 2015. Exploring the policy and social factors fueling the expansion and shift of sugarcane production in the Brazilian Cerrado. GeoJournal. DOI: https://doi.org/10.1007/s10708-015-9666-y. Günther, F., E. Teixeira, E.T. Hizsnyik and H. Van Velthuizen. 2008. Land use dynamics and sugarcane production. In: Sugarcane ethanol: contributions to climate change mitigation and the environment, edited by P. Zuurbier and J. Van de Vooren. Wageningen Academic Publishers, Wageningen, the Netherlands, pp. 29-62. Landau, E.C., R. Kelly da Cruz, A. Hirsch, F.M. Pimenta and D.P. Guimaraes. 2012. Variação geográfica do tamanho dos módulos fiscais no Brasil. Embrapa Milho e Sorgo, Sete Lagos, Brazil. Levine, S., G. Taylor, D. Arthur and M. Tolleth. 2014. Understanding crude oil and product markets. American Petroleum Institute, Washington, DC, WA, USA. Lima Filho, R.R., G.A.M. Aguiar and A.M. Torres. 2014. Arrendar ou produzir? AgroAnalysis 29: 26-27. Manzatto, C. V. 2009. Zoneamento agroecológico da cana-de-açúcar. Embrapa Solos, Rio de Janeiro, Brazil. McManus, C., J.O.J. Barcellos, B.K. Formenton, P.M. Hermuche, O. A. de Carvalho, R. Guimarães, M. Gianezine, E.A. Dias, N. Van de Lampert, D. Zago and J.B. Neto. 2016. Dynamics of cattle production in Brazil. PLoS ONE 11: e0147138. Milanez, A.Y. and D. Nyko. 2012. O futuro do setor sucroenergético e o papel do BNDES. In BNDES 60 Anos: perspectivas setorias, edited by F.L. de Sousa, 1st ed., 2:62-87. BNDES, Rio de Janeiro, Brazil. Ministry of Agriculture, Live Stock and Supply (MAPA). 2006. Brazilian agroenergy plan 2006-2011. Embrapa, Brazil. International Food and Agribusiness Management Review

179

Sant’Anna et al.

Volume 19, Issue 4, 2016

Morton, D.C., R.S. DeFries, Y.E. Shimabukuro, L.O. Anderson, E. Arai, F. del Bon Espirito-Santo, R. Freitas and J. Morisette. 2006. Cropland expansion changes deforestation dynamics in the southern Brazilian Amazon. Proceedings of the National Academy of Sciences of the USA 103: 14637-14641. Neves, M.F., V.G. Trombin and M. Consoli. 2010. The sugar-energy map of Brazil. In: Ethanol and bioelectricity: sugarcane in the future of the energy matrix, edited by E.L. Leao de Sousa and I. de C. Macedo. UNICA, Sao Paulo, Brazil, 14-43. Neves, M.F., R.S. Waack and M.K. Marino. 1998. Sistema agroindustrial da cana-de-açúcar: caracterização das transações entre empresas de insumos, produtores de cana e usinas. Available at: http://tinyurl. com/jrpohjo. O Estadão de S. Paulo. 2011. Restrição é ameaça à Cosan. Available at: http://tinyurl.com/zcvvetk. O Popular. 2011. Jataí aprova legislação restritiva ao cultivo da cana-de-açúcar. Available at: http://tinyurl. com/zs8jtg5. OECD/IEA. 2015. Oil medium-term market report 2015. IEA Publications, Paris, France. Oliveira, G. 2013. Estrangeiros são a nova geração de usineiros. O Globo. Available at: http://tinyurl.com/ jk63tb3. Ramos, P. 2016. Trajetória e situação atual da agroindústria canavieira do Brasil e do mercado de álcool carburante. In: Quarenta anos de etanol em larga escala no Brasil: desafios, crises e perspectivas, edited by G.R. dos Santos. IPEA, Brasilia, Brazil, pp. 47-82. Reinhardt, F., M. Noel and R.R. de Pinho. 2009. Cosan: thinking outside the barrel. Harvard Business Review Case Study: 710017. Sano, E.E., R. Rosa, J.L.S. Brito and L.G. Ferreira. 2008. Mapeamento semidetalhado do uso da terra do Bioma Cerrado. Pesquisa Agropecuaria Brasileira 43: 153-156. Sant’Anna, A.C., G. Granco, J. Bergtold and M.M. Caldas. 2015. Os desafios da expansão da cana-de-açúcar: como pensam e agem arrendatários e produtores? Radar 39: 39-47. SECEX. 2015. Alice Web. Trade database. Ministry of Development, Industry and Foreign Trade of Brazil. Dataset available at: http://tinyurl.com/q26tup5. SINDICOM. 2015. Statistics from the sector. National Syndicate of the Fuel and Lubricant distributers. Available at: http://tinyurl.com/z58l2kh. Siqueira, C. 2013. Usinas de cana se instalam na região. O Estadão de São Paulo. Available at: http://tinyurl. com/znvvu6g. Sousa, P.N., G. Scur and R. de Castro Souza. 2012. Panorama da cadeia produtiva do etanol no Brasil: gargalos e proposições para seu desenvolvimento. Gestão da Produção, Operações e Sistemas 7: 45. State of Sao Paulo Government. 2014. Fabricação de açúcar e etanol a partir da cana-de-açúcar. Portal Do Biogás. Available at: http://tinyurl.com/h3x72ff. Torres, D., C. Pray, G. Martha, Y. Jun, S. Msangi, S. Rozelle and F. Avila. 2011. The Brazilian ethanol industry: an overview of its production, technology, location, land use, regulations, and future prospects. Available at: http://tinyurl.com/h4z93pf. Transpetro. 2007. Transpetro Relatorio Annual. Transpetro, Rio de Janeiro, Brazil. US Energy Information Administration (EIA). 2015. International Energy Statistics. Available at: http:// tinyurl.com/h97oksz. UNICA. 2015. Histórico e Missão. Available at: http://tinyurl.com/j4qe4rt. Union of the Producers of Bioenergy (UDOP). Values of ATR and prices for the ton of sugarcane. Available at: http://tinyurl.com/h4vp2mm. Valdes, C., K. Hjort and R. Seeley. 2016. Brazil’s agricultural land use and trade: effects of changes in oil prices and ethanol demand. ERR-210. USDA Economic Research Service, Washington D.C., WA, USA. Vieira, M. C., J.F. Lima and N.M. Braga. 2007. Setor sucroalcooleiro Brasileiro: evolução e perspectivas. Available at: http://tinyurl.com/4h2a5y5. Walker, R., N.J. Moore, E. Arima, S. Perz, C. Simmons, M. Caldas, D. Vergara and C. Bohrer. 2009. Protecting the Amazon with protected areas. Proceedings of the National Academy of Sciences of the USA 106: 10582-10586.

International Food and Agribusiness Management Review

180

Sant’Anna et al.

Volume 19, Issue 4, 2016

Zanão, A.G. 2009. Caracterização da infra-estrutura de armazenagem de álcool no Brasil e análise da sua concentração na região Centro-Sul. Page 72 Master thesis. Escola Superior de Agricultura ‘Luiz de Queiroz’, Universidade de São Paulo, Piracicaba, Brazil. Zapata, C. and P. Nieuwenhuis. 2009. Driving on liquid sunshine – the Brazilian biofuel experience: a policy driven analysis. Business Strategy and the Environment 18: 528-541.

International Food and Agribusiness Management Review

181