African Statistical Journal - Vol. 20 by African Development Bank

African Statistical Journal Journal africain de statistiques

African Statistical Journal / Journal africain de statistiques

ENT

S ND

AF RIC

P LO AIN DE DEVE

2233 2820

Volume 20 – February / février 2018

© AfDB/BAD, 2018 – Statistics Department / Département des statistiques Complex for Economic Governance & Knowledge Management/ Complexe de la gouvernance économique et de la gestion du savoir African Development Bank Group / Groupe de la Banque africaine de développement Avenue Joseph Anoma 01 BP 1387 Abidjan 01 Côte d’Ivoire Tel: (+225) 20 26 42 43 Internet: http://www.afdb.org Email: ASJ-Statistics@afdb.org ISSN : 2233-2820

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas. Vincent Musoke Nsubuga 3. Computing Consumer Price Indices and Purchasing Power Parities: A Special case for Africa Rees Mpofu 4. Partnership, processes and possibilities: the South African experience of integrating Purchasing Power Parity (PPP) and Consumer Price Index (CPI) work Patrick Kelly and Lekau Ranoto

ELOPME

PE M

DEV

B A N QU EA F

E DE DEV EL

ENT EM PP

IN CA RI

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012 Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye

5. Counting what counts: Africa’s seminal initiative on governance, peace and security statistics Marie Laberge, Yeo Dossina and Francois Rouband

Volume 20 – February / février 2018 African Development Bank Group Groupe de la Banque africaine de développement

Designations employed in this publication do not imply the expression of any opinion on the part of the African Development Bank or the Editorial Board concerning the legal status of any country or territory, or the delimitation of its frontiers. The African Development Bank accepts no responsibility whatsoever for any consequences of its use. African Development Bank Group Avenue Joseph Anoma 01 BP 1387 Abidjan 01 Côte d’Ivoire Tel: (+225) 20 26 42 43 Email: ASJ-Statistics@afdb.org Internet: http://www.afdb.org

Les dénominations employées dans cette publication n’impliquent, de la part de la Banque africaine de développement ou du comité de rédaction, aucune prise de position quant au statut juridique ou au tracé des frontières des pays. La Banque africaine de développement se dégage de toute responsabilité de l’utilisation qui pourra être faite de ces données. Groupe de la Banque africaine de développement

Avenue Joseph Anoma 01 BP 1387 Abidjan 01 Côte d’Ivoire Tel: (+225) 20 26 42 43 Courriel: ASJ-Statistics@afdb.org Internet: http://www.afdb.org

@ADB/BAD, 2018 – Statistics Department / Département des statistiques ISSN: 2233 2820

Design/layout by Phoenix Design Aid A/S, Denmark. ISO 14001/ISO9000 certified and approved CO2 neutral company. Website: www.phoenixdesignaid.dk.

African Statistical Journal Journal africain de statistiques Volume 20 February / fĂŠvrier 2018

Contents Editorial.................................................................................................4 Acknowledgments................................................................................10 1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012 Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye..................................................................................... 14 2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas Vincent Musoke Nsubuga............................................................................ 56 3. Computing Consumer Price Indices and Purchasing Power Parities: A Special case for Africa Rees Mpofu................................................................................................ 99 4. Partnership, processes and possibilities: the South African experience of integrating Purchasing Power Parity (PPP) and Consumer Price Index (CPI) work Patrick Kelly and Lekau Ranoto................................................................ 136 5. Counting what counts: Africa’s seminal initiative on governance, peace and security statistics Marie Laberge, Yeo Dossina and Francois Rouband.................................... 146

Call for Papers...................................................................................... 183 Editorial Policy..................................................................................... 186 Guidelines for Manuscript Preparation and Submission................. 188

Journal africain de statistiques, numéro 20, février 2018

Table des matières Éditorial.................................................................................................7 Remerciements.....................................................................................12 1. Évaluation statistique et qualitative de la qualité de la conception, de la performance, de l’efficacité et de la durabilité de l’assistance de la Banque africaine de développement à la microfinance, 2000-2012 Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, et Gloria Uwingabiye..................................................................................... 14 2. Indices des prix à la consommation (IPC) pour différents groupes de population, groupes de revenu et régions géographiques Vincent Musoke Nsubuga............................................................................ 56 3. Calcul des indices des prix à la consommation et des parités de pouvoir d’achat: un cas particulier pour l’Afrique Rees Mpofu................................................................................................ 99 4. Partenariat, processus et possibilités: l’expérience sud-africaine d’intégration de la parité de pouvoir d’achat (PPA) et des travaux sur l’indice des prix à la consommation (IPC) Patrick Kelly et Lekau Ranoto.................................................................... 136 5. Compter ce qui compte: l’initiative déterminante de l’Afrique en matière des statistiques sur la gouvernance, la paix et sécurité Marie Laberge, Yeo Dossina et Francois Rouband....................................... 146

Demande de soumission d’articles..................................................... 184 Ligne éditoriale..................................................................................... 187 Instructions pour la préparation et la soumission de manuscrits.... 191

The African Statistical Journal, Volume 20, February 2018

Editorial We welcome you to Volume 20 of the African Statistical Journal (ASJ). The Journal’s mission remains unchanged since inception: to serve as a forum and common platform for sharing ideas on statistical development in Africa and to stimulate discussion and dialogue on key emerging issues. The ASJ attempts to reach out to statisticians, economists and related professionals in Africa and beyond, who are keen to engage with evolving developmental issues affecting the African citizenry. We begin this volume with an article entitled, “Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000–2012”. The article evaluates the quality, performance and effectiveness of microfinance of public, private sector and trust fund projects from 2000 to 2012. The main challenges identified revolved around the collection and use of data. Disbursement took relatively long, given the rapid changes in market conditions across client countries. The paper recommends strategic refocus, better reporting, monitoring, replication, and scale up of successful initiatives. The second article, “Consumer Price Indices (CPIs) for different population groups, incomes and geographical areas,” outlines practical challenges facing CPI practitioners when measuring inflation in different population and income groups and across geographical areas. The challenges include the phenomenon of incorporating expenditure data without corresponding prices for products such as second-hand cars. The reallocation of expenditures (notably, from regions where price collection is impossible owing to the permanent unavailability of relevant items to regions where those items are readily available) distorts price indices. The paper cites the case of second-hand motor vehicles, which are generally available in Kampala City, but not in remote rural areas like Mbarara. The third article, “Computing consumer price indices and purchasing power parities: A special case for Africa” advocates for simultaneous data collection for CPIs and the International Comparison Program (ICP). It emphasizes the essence of developing cost effective, methodologically sound and simultaneous data collection procedures for the measurement of Consumer Price Indices (CPIs) and Purchasing Power Parities (PPPs) at country level. The article highlights the necessity of mastering specific methods at basic heading levels for both statistical endeavors as a building block for methodologically sound data collection for price changes and price levels.

Journal africain de statistiques, numéro 20, février 2018

Editorial

The fourth article, “Partnership, processes and possibilities: the South African experience of integrating Purchasing Power Parity (PPP) and Consumer Price Index (CPI),” corroborates the basic premise of the third article and provides a practical case where applied work has taken place. The article proposes a case for the integration of ICP into regular pricing programs of National Statistical Offices in Africa. The European Union (EU) has a well-documented rolling benchmark dating back to the 1990s. The article advocates for a cost-effective process that avails the requisite data for the computation of annual or time series PPPs responding to policy-relevant data demand for regional integration purposes. South Africa’s applied work is in line with the adoption of the International Comparison Program (ICP) as a permanent statistical endeavor at United Nations Statistical Commission’s 46th session in 2016. To African countries that find response to ICP data needs cumbersome, the article provides indispensable practical advice: the mainstreaming of ICP activities into the regular pricing programs of African countries. The ICP data is essential for regional integration and is fit for purpose for selected Sustainable Development Goals (SDGs), Agenda 2063 indicators, the African Development Bank’s High 5s (Light up and Power Africa, Feed Africa, Industrialize Africa, and Improve the Quality of Life for the People of Africa), as well as other development agendas. The fifth article, “Counting what counts: Africa’s seminal initiative on governance, peace and security statistics”, documents the practical experience of eleven African National Statistical Offices that tested and eventually institutionalized a methodology for the harmonization of official statistics on governance, peace and security statistics between 2012 and 2017. This took place while the rest of the world was still debating the rationale for including this new domain in the next global development agenda. We hope that you will find this volume of the ASJ informative and stimulating. We encourage the African Statistical Community to continue using the ASJ as an authoritative knowledge-sharing forum, particularly at a time when the national statistical systems in Africa need to bolster their resilience and capacity to respond to growing demands for data. A decisive response to the data gaps in national development plans and in the continental and international development strategies, including the AfDB’s High 5 priorities, is even more urgent now and calls for urgent action by all stakeholders.

The African Statistical Journal, Volume 20, February 2018

Editorial

We would like to thank the contributors and reviewers, as well as all those who have made this volume a reality. Dr. Charles Leyeka Lufumpa Co-Chair, Editorial Board Statistics Department African Development Bank Abidjan, Côte d’Ivoire Email: c.lufumpa@afdb.org

Professor Ben Kiregyera Co-Chair, Editorial Board International Statistical Consultant, Kampala Uganda Email : bkiregyera@yahoo.com

Journal africain de statistiques, numéro 20, février 2018

Éditorial Nous vous souhaitons la bienvenue dans la découverte du volume 20 du Journal Statistique Africain (JSA). La mission du journal reste inchangée depuis sa création, à savoir servir de forum et de plate-forme commune de partage d'idées sur le développement statistique en Afrique et stimuler la discussion et le dialogue sur les principales questions émergentes. Le JSA entend s’adresser aux statisticiens, économistes et autres professionnels apparentés en Afrique et au-delà, désireux de s’intéresser aux questions évolutives du développement affectant les citoyens africains. Nous commençons ce volume par un article intitulé « Évaluation statistique et qualitative de la conception, de la qualité, de la performance, de l'efficacité et de la durabilité de l'assistance de la Banque africaine de développement à la microfinance de 2000 à 2012 ». L'article évalue la qualité, la performance et l'efficacité des projets de microfinance des secteurs public, privé et fiduciaire de 2000 à 2012. Les principaux défis identifiés ont porté sur la collecte et l'utilisation des données. Les procédures de décaissement ont été relativement longues, compte tenu des changements rapides des conditions du marché dans les pays bénéficiaires. L’article recommande un recentrage stratégique, un meilleur système de reporting et de suivi, de reproduction et de multiplication des initiatives réussies. Le deuxième article, « Indices des prix à la consommation (IPC) pour différents groupes de population, revenus et zones géographiques », décrit les défis pratiques auxquels sont confrontés les experts des IPC pour mesurer l'inflation dans différents groupes de population et de revenu, et dans différentes zones géographiques. Parmi ces défis figure le hiatus entre l'incorporation des données de dépenses et l’absence des prix correspondants pour des produits tels que les voitures d'occasion. La réallocation des dépenses (notamment des régions où la collecte des prix de produits est rendue impossible par l'indisponibilité permanente de ces produits à des régions où ils sont facilement disponibles) fausse les indices de prix. L’article cite le cas des véhicules à moteur d'occasion qui sont généralement disponibles dans la ville de Kampala, mais pas dans les zones rurales reculées comme Mbarara. Le troisième article « Calcul des indices des prix à la consommation et des parités de pouvoir d’achat : cas particulier pour l’Afrique » préconise la collecte simultanée de données pour les IPC et le Programme de comparaison internationale (PCI). Il met l'accent sur la nécessité essentielle de mettre en place dans les pays, des procédures efficientes, simultanées et méthodologiquement robustes de collecte de données destinées aux indices des prix à la consommation (IPC) et aux parités de pouvoir d'achat (PPP). L'article met en évidence la nécessité, pour les deux activités statistiques, de

The African Statistical Journal, Volume 20, February 2018

Éditorial

maîtriser des méthodes spécifiques relatives aux positions élémentaires, ce qui garantirait fondamentalement l’efficacité méthodologique des collectes de données servant de base à la mesure des évolutions temporelles de prix et des comparaisons spatiales de niveau des prix. Le quatrième article, « Partenariat, processus et possibilités : l'expérience sud-africaine d'intégration des parités de pouvoir d'achat (PPP) et de l'indice des prix à la consommation (IPC) » corrobore le postulat de base du troisième article dont il fournit un exemple pratique. L'article propose un plaidoyer en faveur de l'intégration du PCI dans les activités régulières de statistique des prix des instituts nationaux de statistique en Afrique. L'Union européenne (UE) dispose depuis les années 1990 d'un système à année de référence mobile bien documenté. L'article préconise un processus efficient qui génère les données requises pour le calcul des PPA annuelles ou de séries chronologiques des PPA en vue de répondre aux besoins des politiques d'intégration régionale. Le cas concret de l'Afrique du Sud est conforme à l'adoption – lors de la 46ème session de la Commission statistique des Nations Unies en 2016 – du Programme de comparaison internationale (PCI) en tant qu’initiative statistique permanente. Les pays africains qui jugent les exigences d’un PCI permanent excessives trouveront dans l’article des conseils pratiques indispensables sur l’intégration des activités du PCI dans leurs programmes de prix réguliers. Les données du PCI sont essentielles pour l'intégration régionale et sont adaptées à certains objectifs de développement durable (ODD), aux indicateurs de l'Agenda 2063, aux cinq grandes priorités de la Banque africaine de développement (Éclairer l'Afrique, nourrir l'Afrique, intégrer l'Afrique, industrialiser l'Afrique, et améliorer la qualité de vie des Africains), ainsi que d'autres programmes de développement. Le cinquième article, « Compter ce qui compte : l'initiative africaine sur les statistiques de gouvernance, de paix et de sécurité » documente l'expérience pratique de onze offices statistiques nationaux africains qui ont testé et institutionnalisé une méthodologie d'harmonisation des statistiques officielles sur la gouvernance, la paix et la sécurité entre 2012 et 2017. Cette expérience se développait alors que le reste du monde était encore en train de débattre de la pertinence d'inclure ce nouveau domaine dans le prochain agenda de développement mondial. Nous espérons que vous trouverez ce volume du JSA instructif et stimulant. Nous encourageons la Communauté statistique africaine à continuer d'utiliser le JSA comme un forum de partage des connaissances faisant autorité, en particulier à un moment où les systèmes statistiques nationaux

Journal africain de statistiques, numéro 20, février 2018

Éditorial

en Afrique doivent renforcer leur résilience et leur capacité à répondre aux demandes croissantes de données. Une action décisive s’impose de façon de plus en plus urgente face aux lacunes statistiques relevées dans le cadre des plans de développement nationaux et des stratégies de développement continentales et internationales, y compris les cinq grandes priorités de la BAD. Elle requiert la participation de toutes les parties prenantes. Nous tenons à remercier tous ceux qui ont fourni des articles ou les ont revus, ainsi que tous les autres qui ont permis que ce volume soit une réalité. Dr Charles Leyeka Lufumpa Co-président du comité de rédaction Département des statistiques Banque africaine de développement Abidjan, Côte d’Ivoire Courriel: c.lufumpa@afdb.org

The African Statistical Journal, Volume 20, February 2018

Professeur Ben Kiregyera Co-président du comité de rédaction Consultant international de statistique, Kampala Ouganda Courriel: bkiregyera@yahoo.com

Acknowledgments Co-Chairs of the Editorial Board: Dr. Charles Leyeka Lufumpa, Director, Statistics Department, African Development Bank Group, Abidjan, Côte d’Ivoire. Professor Ben Kiregyera, International Statistical Consultant, Kampala, Uganda. Editor in Chief: Fessou Emessan Lawson, Officer in Charge (O.I.C), Statistical Capacity Building Division, African Development Bank. Production Editor: Rees Mpofu, Statistics Department, African Development Bank, Abidjan, Côte d’Ivoire. English Editor: Dr. Ndaeyo Uko, Monash University, Melbourne, Australia. Expert Reviewers: Cecile Ambert, Private Sector Facility, African Development Bank. Issahaku Budali, Principal Social Protection Specialist, African Development Bank. Sanjev Bhonoo, Statistician, Statistics Mauritius, Mauritius. Soumandra Koumar Dash, Principal Risk Officer, African Development Bank. Mark Flaming, Development Finance Specialist, President of Frontier Ventures. Simon M. Gaitho, Manager, Consumer Price Index (CPI), Kenya National Bureau of Statistics (KNBS), Nairobi, Kenya. Zekebweliwai Fuh Kah Geh, Associate Microfinance Analyst at the Consultative Group to Assist the Poor (CGAP); Consultant, African Development Bank. Damien Onyema Ihedioha, Agro-Industrial Officer, African Development Bank. Kathryn Imboden, Consultants in Inclusive Finance.

Journal africain de statistiques, numéro 20, février 2018

Lilian Macharia, Principal Investment Officer, African Development Bank. Godfrey Makware, Manager, Industry, Mining and Energy Statistics, Zimbabwe National Statistics Agency (ZIMSTAT). Mohamed Manai, Former Division Manager, Independent Development Evaluation, African Development Bank. Leila Mokadem, Resident Representative for Egypt, African Development Bank. Thouraya Triki, Chief Country Economist, African Development Bank. Mayada El-Zoghbi, Senior Financial Sector Specialist; Head of Consultative Group to Assist the Poor (CGAP)â&#x20AC;&#x2122;s Strategy, Research and Development Unit.

The African Statistical Journal, Volume 20, February 2018

Remerciements Coprésidents du comité de redaction : Dr. Charles Leyeka Lufumpa, Directeur, Département de la statistique Le Groupe de la Banque africaine de développement, Abidjan, Côte d’Ivoire. Professeur Ben Kiregyera, Consultant international en statistique, Kampala, Ouganda. Rédacteur en chef : Fessou Emessan Lawson, Chargé de mission, Division du renforcement des capacités statistiques, Banque africaine de développement. Éditeur de production : Rees Mpofu, Département des statistiques, Banque africaine de développement, Abidjan, Côte d’Ivoire. Editeur anglais : Dr. Ndaeyo Uko, Université Monash, Melbourne, Australie. Examinateurs experts : Cecile Ambert, Facilité du secteur privé, Banque africaine de développement. Issahaku Budali, Spécialiste principal de la protection sociale, Banque africaine de développement. Sanjev Bhonoo, Statistician, Statistics Mauritius, Mauritius. Soumandra Koumar Dash, Agent principal de gestion des risques, Banque africaine de développement. Mark Flaming, Spécialiste du financement du développement, Président de Frontier Ventures. Simon M. Gaitho, Directeur, Indice des prix à la consommation (IPC), Kenya National Bureau of Statistics (KNBS), Nairobi, Kenya. Zekebweliwai Fuh Kah Geh, Analyste Associé en Microfinance au Groupe Consultatif d’Assistance aux Pauvres (CGAP); Consultant, Banque africaine de développemen. Damien Onyema Ihedioha, Agent agro-industriel, Banque africaine de développement.

Journal africain de statistiques, numéro 20, février 2018

Kathryn Imboden, Consultant en finance inclusive. Lilian Macharia, Agent principal des investissements, Banque africaine de développement. Godfrey Makware, Directeur des statistiques de l’industrie, mines et de l’énergie, Agence nationale de la statistique du Zimbabwe (ZIMSTAT) Mohamed Manai, Ancien chef de division, évaluation indépendante du développemen, Banque africaine de développement. Leila Mokadem, Représentant résident pour l’Égypte, Banque africaine de développement. Thouraya Triki, Chef Économiste, Banque africaine de développement. Mayada El-Zoghbi, Spécialiste principal du secteur financier; Chef du Groupe de la stratégie, de la recherche et du développement du Groupe consultatif d’assistance aux pauvres (CGAP).

The African Statistical Journal, Volume 20, February 2018

1. Statistical and Qualitative Assessment of the Design Quality, Performance, Effectiveness and Sustainability of the Assistance of the African Development Bank to Microfinance, 2000–2012 Albert-Enéas Gakusi1, Alice Negre2, Mabarakissa Diomanade3, Gloria Uwingabiye4

Abstract Until the end of 2014 when the African Development Bank (the Bank) set up a financial sector department, microfinance activities were designed and implemented by four departments. This paper presents the results of an effort to evaluate the quality, performance and effectiveness of microfinance projects approved by the Bank between 2000 and 2012. The evaluation faced several research challenges, mainly the lack of reliable data at the Bank and in the countries visited. For instance, the Bank had no comprehensive list of its microfinance projects. An evaluation of appraisal reports of relevant sectors and interviews with task managers identified 94 microfinance projects approved by the Bank between 2000 and 2012, for UA338 million, representing 0.9 percent of the Bank’s total approved amount. Overall, the project goals were consistent with the Bank’s and client countries’ development priorities, but inadequate projects design often prevented efficient implementation and achievement of results. Effectiveness was deemed unsatisfactory at best. Efficiency scored unsatisfactory, with disbursement taking place on average 11 months after signature, instead of 5 months as required by Bank procedures. These relatively long delays raised serious concerns as market conditions change rapidly and project relevance can no longer be guaranteed. Sustainability has been noted as a key concern, with only half of the 22 ongoing retail projects working with institutions likely or very likely to be sustainable, and only three out of the ten wholesale institutions deemed likely to be sustainable over time. Private sector projects were much more likely than public sector projects to produce sustainable results. The evaluation recommended different actions including: refocusing the strategy and the execution of the portfolio based on the Bank’s capabilities; improving the reporting and monitoring systems to learn from projects performance; avoiding microfinance components; replicating and scaling up successful initiatives. 1 Chief Evaluation Officer, African Development Bank, a.gakusi@afdb.org 2 Senior Consultant in Inclusive Finance, alicenegre@hotmail.com 3 Evaluation Officer, African Development Bank, m.diomande@afdb.org 4 Research fellow in applied development finance, European Investment Bank, gloryuwin@hotmail.com

Journal africain de statistiques, numéro 20, février 2018

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012

Key words: microfinance, wholesale institutions, standalone projects, components, microfinance project types. Abstrait Jusqu’à la fin de 2014, lorsque la Banque africaine de développement (la Banque) a mis en place un département du secteur financier, les activités de microfinance ont été conçues et mises en œuvre par quatre départements. Cet article présente les résultats de l’évaluation de la qualité, du performance et de l’efficacité des projets de microfinance approuvés par la Banque entre 2000 et 2012. L’évaluation a fait face à plusieurs défis de recherche, principalement le manque de données fiables à la Banque et dans les pays visités. Par exemple, la Banque n’avait pas de liste complète de ses projets de microfinance. Une évaluation des rapports d’évaluation des secteurs pertinents et des entretiens avec les chefs de projet ont identifié 94 projets de microfinance approuvés par la Banque entre 2000 et 2012, pour 338 millions d’unités de compte, UC, soit 0,9% du montant total approuvé par la Banque. Dans l’ensemble, les objectifs du projet correspondaient aux priorités de développement de la Banque et des pays clients, mais la conception inadéquate des projets a souvent empêché une mise en œuvre efficace et l’obtention de résultats. L’efficacité a été jugée insatisfaisante au mieux. L’efficience a été jugée insatisfaisante, avec un décaissement intervenant en moyenne 11 mois après la signature, au lieu de 5 mois comme requis par les procédures de la Banque. Ces retards relativement longs ont soulevé de sérieuses préoccupations, car les conditions du marché changent rapidement et la pertinence du projet ne peut plus être garantie. La pérennité a été considérée comme une préoccupation majeure, avec seulement la moitié des 22 projets de commerce de détail en cours travaillant avec des institutions susceptibles ou très susceptibles d’être durables, et seulement trois des dix institutions decommerce de gros jugées durables dans le temps. Les projets du secteur privé étaient beaucoup plus susceptibles que les projets du secteur public pour produire des résultats durables. L’évaluation a recommandé différentes actions, notamment: recentrer la stratégie et l’exécution du portefeuille en fonction des capacités de la Banque; améliorer les systèmes de rapports et de suivi pour tirer des leçons de la performance des projets; éviter les composantes de la microfinance; reproduire et intensifier les initiatives réussies. Mots clés : microfinance, institutions de gros, projets autonomes, composants, types de projets de microfinance.

The African Statistical Journal, Volume 20, February 2018

Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye

1. INTRODUCTION

1.1 Objectives and scope This paper presents the results of the portfolio review, one of the four desk reviews conducted to evaluate the relevance of the Bank’s microfinance policy and strategy, as well as the performance of microfinance projects (MFPs) approved between 2000 and 2012. The purpose of this portfolio review is three-fold. First, it aims to identify the structure and trends of the Bank’s microfinance portfolio. Second, it assesses the implementation performance of a subgroup of selected microfinance projects (MFPs). Third, it presents the main findings and makes recommendations to address policy, design quality, and implementation challenges for future projects5. 1.2 Methodology and evaluation challenges The evaluation methodology consisted of three steps. First, while using the Bank’s information systems (SAP and Darms), as well as previous reviews, the evaluation team established the list of MFPs. Second, the team reviewed project appraisal reports, supervision summaries, 43 Project Completion Reports (PCRs), and 17 project completion evaluation notes. Interviews were conducted with the Bank’s microfinance task managers and managers from Agriculture and Agroindustry Department (OSAN), Human Development Department (OSHD), Private Sector Department (OPSM), Governance, Finance and Economic Management Department (OSGE), the Development Research Department (EDRE) as well as the Regional Integration and Trade Department (ONRI). Feedback received from the July 2013 workshop, which brought together staff from different operational departments, also complemented the desk review findings. Lastly, a sample of 25 approved and ongoing projects were scored to increase available information to assess the relevance, effectiveness, efficiency, and sustainability of MFPs. The evaluation team faced several challenges. The absence of a Bank-wide clear coding system for the identification of microfinance projects and the lack of an agreed definition of what constitutes microfinance were daunting obstacles to the identification of microfinance-related projects. Of 133 projects identified as microfinance by the Bank’s SAP system, the evaluation team recognized only 92 as MFPs and identified two other projects that had not previously been identified through formal channels (Table 1).6 All 5 All abbreviations and acronyms are in the annex. 6 The case studies later revealed that two of these 94 projects did not include microfinance operations: The Poverty Reduction Support project in Ghana (OSGE) and the Support to Poverty Reduction Strategy IV project in Burkina Faso (OSGE).

Journal africain de statistiques, numéro 20, février 2018

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012

tables from which the analysis is based are included in the review report of microfinance portfolio available upon request. Table 1. Definition issues: from SAP to final list of MFPs Initial screening (SAP and Task Managers): 133 projects • Less 14 projects whose microfinance operations were not funded by the Bank • Less 12 projects with no microfinance in objectives/logframe • Less 6 projects with micro-enterprise development but without microfinance • Less 7 small and medium enterprise (SME) banking rather than microfinance projects • Less 1 cancelled project • Less 1 equity top-up that was merged with first-time equity investment • Plus 2 projects, not reported in SAP, but that included microfinance  Final list: 94 projects

The definition used for this review focused on the supply of loans, savings, and other basic financial services to the poor. In the absence of clear-cut criteria for the Bank, the review had to rely on proxies used in the sector by peer institutions, such as the maximum loan amount of USD10,000 to qualify as micro-loan.7 The downside of this proxy is that it focuses on microcredit.8 While it is possible that some microfinance projects were not captured in the 94 identified projects, the process used was the most extensive thus far and is likely to cover almost all microfinance projects. The lack of a centralized performance data reporting system for microfinance projects hampered a comprehensive performance analysis. Forty-one (41) percent of projects had no microfinance indicators in their logframe, 45 percent had no set targets, and no project had clear baseline data against which performance could be measured (Table 2). There was a slight improvement from the first period (2000 to 2006) to the second period (2007 to 2012) covered by this review, with 47 percent of projects without performance indicators for the first period compared to only 31 percent for the second period. The Private Sector and Microfinance Operations Department 7 Standard proxy used by the World Bank to distinguish microfinance operations from other MSME banking operations. 8 OPSM comments to the December 2013 version of this report highlighted that two insurance companies -- Strategis Insurance (Tanzania) and Hygeia (Nigeria) -- could have been included in the portfolio because they provide insurance products to low-income families.

The African Statistical Journal, Volume 20, February 2018

Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye

(OPSM) fares much better than Agriculture and Agroindustry Department (OSAN) and the Human Development Department (OSHD) in setting and tracking performance indicators. In the same vein, of the 57 completed projects only 43 have PCRs, and only 12 of these PCRs provide information on the microfinance component. In its assessment of project performance, the evaluation team focused on 25 projects and requested task managers to individually report on a selected number of indicators. Most task managers could provide this type of performance information. However, given the data constraints, the evaluation team could not provide detailed analysis for some of the evaluation criteria, such as effectiveness and efficiency. Table 2. Availability of microfinance performance indicators and targets in logframes Number of projects 2000 2007 Availability of indicators Total OSHD OSAN OSGE OPSM –06 –12 Projects with no indicator

Projects with full set of indicators

Projects with no target

Projects with good practice targets (i.e. that include measurable indicators of output, outcome, and impact)

Total projects of the sample

Percentage Projects with no indicator

Projects with full set of indicators

Projects with no target

Projects with good practice targets

100

Total projects of the sample

Journal africain de statistiques, numéro 20, février 2018

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012

STRUCTURE AND TRENDS OF THE PORTFOLIO

2.1 Key milestones for microfinance at the Bank The composition of the Bank’s microfinance portfolio reflects the history of the Bank and the way it has been managed by its various departments. Key milestones of the Bank’s history in the microfinance sector are provided in Table 3. Table 3. Microfinance at the Bank 1970s Bank’s first experiences in microfinance through lines of credit (LOCs) as microcredit components of larger multi-sector projects. These LOCs are managed by the project implementation unit (PIU). 1990s Promotion of financial intermediaries. Bank LOCs provided to specialized institutions, such as national agricultural banks, rather than PIUs. 1999

Creation of the African Development Fund Microfinance Initiative for Africa (AMINA), a central unit reporting directly to the Vice President for Operations and focused on capacity building for Microfinance Institutions (MFIs). Public sector departments continue their credit component operations.

2000

Agricultural and Rural Development Sector Policy broadens the scope of microfinance operations from agricultural to rural credit, from credit to a broad range of financial services, and from a focus on state-owned agricultural banks to a variety of partners (including private entities).

2002

The Consultative Group to Assist the Poor (CGAP) in its peer review of the Bank’s microfinance operations identifies main institutional challenges and makes recommendations. Review of the Bank’s microfinance operations during the annual portfolio review, rating 41 percent of the 74 microfinance projects as “at risk”. The review highlights that poor performance was mostly caused by “inadequate assessment [of ] institutional and absorptive capacity.”

2003

Microfinance business plan, with conversion of AMINA into a permanent Central Microfinance Unit aimed at improving QAE and monitoring of the Bank’s microfinance operations.

2006

MPS highlights the need to reorganize the Bank’s microfinance operations and to integrate best practices in microfinance project management. The Private Sector Department (OPSD) designated as the Bank’s focal point for operations with financial intermediaries in the microfinance sector. OPSD renamed Private Sector and Microfinance Department (OPSM).

The African Statistical Journal, Volume 20, February 2018

Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye

2009

Participation in the CGAP Smart Aid Index, an evaluation that measures how well funders are set up to deliver on their financial inclusion projects. Creation of the Microfinance Capacity Building Trust Fund (MCBF) and the Migration and Transfer Trust Fund, managed by the Private Sector Department (OPSM).

2009

Joint AfDB/IFAD evaluation of agriculture and rural development concludes that the Bank should remain engaged in this sector but with a clear focus and innovative approaches, moving away from multi-component projects and partnering with other agencies.

2010

Agriculture Sector Strategy 2010–2014 confirms that the Bank addresses microfinance operations under its private sector window and through partners such as the International Fund for Agricultural Development (IFAD) and the Alliance for a Green Revolution in Africa (AGRA).

2011

Microfinance stock-taking exercise by an ad hoc task force recommends conducting an independent and comprehensive review of microfinance activities to help the Bank to address the question as to how to enhance the delivery and implementation of microfinance activities to Regional Member Countries (RMCs).

April 2012

MCBF and Migration trust funds relocated from OPSM to the Human Development Department (OSHD).

2.2 Key features of the portfolio From 2000 to 2012, the Bank approved 1,442 projects, of which 94 aimed at strengthening the microfinance sector (Table 4). From 2000 to 2012, the Bank approved UA338 million for microfinance projects, representing 0.9percent of the Bank’s total approved amount and 7.2 percent of its total approvals for the financial sector over the same period. As of May 2013, 29 projects were ongoing, 45 were closed (i.e. projects completed and PCRs available), 12 were completed without a PCR, and 8 were approved, but not yet disbursed. The average size of microfinance project is UA2.4 million, much lower than the average bank project (Figure 1). The size was more than halved from the first to the second period, mostly because of the increase of the number of private sector projects, which are smaller in size than public sector projects. For OSAN, the average microfinance project amount was UA2.6 million, compared to an average of UA9 million for the department overall. For OSHD, the average microfinance project was UA2.8 million, compared to an average of UA12 million for OSHD overall. For OPSM, the average microfinance project was UA1.3 million, compared to an average of UA53

Journal africain de statistiques, numéro 20, février 2018

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012

million for the department overall. Thirty percent of MFPs were less than UA0.5 million. Only three (3) projects were above UA15 million. The three were excluded from the calculation of the averages cited above to keep the average consistent with standard projects. These projects were the Social Fund for Development in Egypt (2006, UA57.7 million), the Rural Income and Economic Enhancement Project in Egypt (2009, UA45.7 million), and RFISP Ethiopia (2003, UA35 million). Table 4. Number and amount of microfinance projects

2000–06

2007–12

Total

94b

242

338

4.5

2.0

2.4

Number of projects approveda Amount approved (UA million) Average project size (UA million)

a The number of projects has a different meaning across departments. For OPSM, a project typically targets one recipient with one instrument. For the public sector, a project can include several recipients and sometimes several instruments. b Case studies later revealed that 2 of the 94 projects did not include microfinance operations. c The size is based on a sample of 85 projects for which data on project size is available. Data for the projects that use budget support as their key instrument were unavailable.

Figure 1. Average project size (UA million) 60 53

50 40 30 20

10 2.6

2.4 Total

OSAN

Microfinance only

Whole department

The African Statistical Journal, Volume 20, February 2018

2.8 OSHD

1.3 OPSM

Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye

Four departments have been involved in microfinance (Table 5). OSHD is the largest department in terms of volume, with 25 projects amounting to UA166 million. With 48 projects amounting to UA153 million, OSAN has the largest number of projects. Although OSAN started withdrawing from microfinance in 2003 while decreasing the number of projects with microfinance component, 9 of its projects are still active.9 OPSM represents only 5 percent of total Bank approvals to microfinance, with 16 projects amounting to UA13 million. OSGE’s microfinancing spans eight broader financial sector programs, including seven budget support programs for which it is not possible to identify the specific microfinance amount. The remaining project, the multinational African Regional Technical Assistance Centers (AFRITAC, phase II), accounts for UA3 million of the amount approved. Over the evaluation period, ninety-one (91) percent of the Bank’s microfinance funds were channeled through government. This proportion decreased from 94 percent of the funds approved during the first half of the period to 80 percent in the second half of the period. This decrease is attributable to OPSM starting microfinance operations in 2006 and funding only the private sector. Table 5. Microfinance projects by Bank department Department

Number of projects approved

Approved amount (UA million)

Number of ongoing projects as of May 2013

2000– 2007– 2000– 2007– 2000– 2007– Total Total Total 06 12 06 12 06 12

OSHD

166

OSAN

149

153

OSGE

n/a

OPSM

Total

242

338

9 According to the technical note to improve the quality of the portfolio of OSAN (2007), the “decrease in the number of projects with credit components since 2001 is closely related to the Bank’s requirement to design such projects in line with best practice standards, especially the ability of target groups to repay loans at market rates of interest. The decrease also corresponds to the requirement that financial intermediaries are identified and assessed, in order to demonstrate the feasibility of the credit components from a financial sector perspective.”

Journal africain de statistiques, numéro 20, février 2018

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012

The African Development Fund (ADF) is the main source of funds. ADF represented 61 percent of total approvals for the entire period of the evaluation, compared to 38percent for the Bank, 0.9 percent for the Fund for African Private Sector Assistance (FAPA), 0.4 percent for the Nigeria Trust Fund (NTF), and 0.2 percent for MCBF (Table 6). However, ADB represents 61 percent of approvals in the second period, being the main source of funds for a large OSHD project in Egypt and for OPSM investments. Table 6. Approved amount by source of funds (UA million) Source

2000–06 2007–12

Total

OSHD OSAN OPSM OSGE

ADB

127

103

ADF

171

205

140

NTF

FAPA

MCBF

242

338

166

153

Total

Debt is the main instrument the Bank uses to support microfinance (Table 7). Loans represent 91 percent of total approvals. This percentage decreased from 92 percent during the first period to 86 percent during the second period. According to CGAP (2012), between 2009 and 2011 loans remained the main instrument used by cross-border funders to fund microfinance. They represented 55 percent of total commitments. It should be noted that public sector loans may end up as grants for final recipient institutions, since governments might on-lend and/or extend Bank funds to recipient institutions as grants. However, there is no easy way to monitor this. The equity portfolio is particularly small and started only during 2007-2012 through OPSM approvals. As CGAP put it: “Both equity investments and guarantees are important instruments because they can enable retail providers to access local sources of funding and build sustainable markets.” Increasingly, public funders are using equity to invest in promising financial institutions serving the poor. This is evidenced by over USD300 million approved by seven public funders to equity investments in Africa in 2011 (CGAP, 2011). Those funders are AECID, AFD/Proparco, AfDB, EIB, FMO, IFC, and KfW.10 While the Bank did undertake five direct equity deals amounting 10 AECID = Spanish Agency for International Development Cooperation; AFD = Agence Française de Developpement; Proparco = Société de Promotion et de Participation pour la Coopération Economique; EIB = European Investment Bank; FMO = Dutch Development

The African Statistical Journal, Volume 20, February 2018

Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye

to UA4.7 million through OPSM, it is by far the smallest player among the public donors operating in the African equity market for microfinance. Table 7. Approved amount by instrument Instrument Loan

2000–06 2007–12

Total

OSHD OSAN

OSGE OPSM

219

299

152

140

Equity

Grant/TA

Total

237

330

165

149

Information on instruments missing for 13 projects. No amount data for OSGE.

Eighty-three (83) percent of the Bank’s approvals aim at refinancing MFIs, and 17 percent at capacity building (Table 8). This aligns with other funders (CGAP, 2012) with an average of 77 percent of total microfinance crossborder funding going to refinancing and 15 percent aimed at strengthening the capacity of various stakeholders (MFIs, MFI networks, policy makers, and supervisors) with the remaining 8 percent unspecified. There was limited variation over time. Table 8. Portfolio by purpose (% of total approved amount) Purpose

2000–06 2007–12

Total

OSHD OSAN

OSGE OPSM

Refinancing direct

Refinancing indirect

Capacity building

100

Total

100

Information missing on 12 projects. Breakdown estimated by the Operations Evaluation Department (OPEV) based on appraisal reports.

The Bank approved projects in 35 countries (Figure 2 and Table 9). Those projects were active in 21 countries as of May 2013. Five countries captured two thirds of total funding: Egypt (UA103 million), Mali (UA 37million), Bank; IFC = International Finance Corporation; KfW = Kreditanstalt für Wiederaufbau (German government-owned development bank).

Journal africain de statistiques, numéro 20, février 2018

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012

Ethiopia (UA35 million), Uganda (UA25 million), and Tanzania (UA20 million). In terms of regional distribution, North Africa, comprising ADB countries but excluding Mauritania, received the highest volume of Bank approvals (UA 108 million); followed by West Africa (UA95 million), East Africa (UA87 million), Southern Africa (UA40 million) and Central Africa (UA8 million). Figure 2. AfDB total microfinance approvals by country, 2000â&#x20AC;&#x201C;2012

No projects < UA 2 million UA 2 to 5 million UA 5 to 10 million > UA 10 million

The African Statistical Journal, Volume 20, February 2018

Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye

Table 9. Regional repartition of RMCs Central Africa

East Africa

North Africa

Southern Africa West Africa

Cameroon

Burundi

Algeria

Angola

Benin

Central African Republic

Comoros

Egypt

Botswana

Burkina Faso

Chad

Djibouti

Libya

Lesotho

Cape Verde

Congo

Eritrea

Mauritania

Madagascar

Côte d’Ivoire

Congo DR

Ethiopia

Morocco

Malawi

Gambia

Equatorial Guinea

Kenya

Tunisia

Mauritius

Ghana

Gabon

Rwanda

Mozambique

Guinea

São Tomé and Principe

Seychelles

Namibia

Guinea-Bissau

Somalia

South Africa

Liberia

South Sudan

Swaziland

Mali

Sudan

Zambia

Niger

Tanzania

Zimbabwe

Nigeria

Uganda

Senegal Sierra Leone Togo

2.3 Overall trends The approved amount followed an upward trend prior to 2006, and afterwards a decreasing trend (except for two years before and after 2006). On average, the Bank approved UA16 million (5.3 new projects) per year after the 2006 MPS was approved. It approved UA35 million (8.8 new projects) per year from 2000 to 2006 (Figure 3). This decrease was also observed when comparing the Bank to other funders. In 2007, the Bank represented 18 percent of all cross-border funds to microfinance (including bilaterals, multilaterals, and investment funds), compared to only 4 percent in 2011 (CGAP 2012). Nonetheless, the smaller size of the portfolio does not necessarily mean a weaker portfolio performance. It mainly reflects an exit from OSAN underperforming operations, and other departments not compensating for this withdrawal.

Journal africain de statistiques, numéro 20, février 2018

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012

Figure 3. Number and amount of new microfinance projects, by year of approval Number of new microfinance projects

37 27

1 2012

2003

2002

2001

2012

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

40 20

2011

2010

2009

7 5

MPS

2005

2011

2010

2004

2008

MPS

2007

2006

Amounts (UA million) 100

2000

This decrease in volume is mostly due to a deep restructuring of the Bank’s portfolio, moving away from agriculture towards more trade and services in rural, urban, and semi-urban areas. In an attempt to improve the portfolio quality, the Bank gradually withdrew from agricultural microfinance, where it had experienced poor performance (see OSAN Technical Note 2007). From 2003, OSAN gradually transferred its microfinance activities to IFAD, which specializes in rural microfinance. IFAD thus funded microfinance operations while OSAN funded other project components. OSAN represented 62 percent of the total amount approved in the first period, down to only 4 percent in the second period (Figure 4). Funding agriculture with microfinance products remains a challenge for all microfinance institutions and donor agencies in Africa and elsewhere. The decrease in approvals was also due to the change in the Bank’s strategic priorities from 2006, which gave preference to infrastructure at the expense of ‘soft’ activities.

The African Statistical Journal, Volume 20, February 2018

Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye

Figure 4. Amount approved by department (UA million) 250 200 150 100 50 0 2000/2006 OSHD

OSAN

2007/2012 OSGE

OPSM

The 2006 MPS did not impact the portfolio as much as anticipated. Although the MPS did not set specific objectives in terms of volume, number of projects, portfolio structure, and innovations, some key changes were expected. The strategy “advocate[d] private-sector leadership in the development of sound financial intermediary institutions in terms of ownership, risk-taking investment, and operational decisions.” However, from 2007 to 2012, 83 percent of the Bank’s ongoing support to private sector institutions still came from the public sector window, with little to no technical input from the private sector department. Public sector departments remain very much focused on supporting MFIs (Table 10). OPSM approvals between 2007 and 2012 totaled only UA16 million compared to more than 80 million for public sector operations during the same period. OPSM did not have staff capacity, incentives, and efficient processes to be significantly involved in microfinance.

Journal africain de statistiques, numéro 20, février 2018

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012

Table 10. Alignment with the 2006 strategy (% of projects working at each level) Pillars of the 2006 strategy

2000– 2007– 2000– 06 12 12

2000–12 only

MFI strengthening

OSHD OSAN

OSGE

OPSM

Policy and supervision

100

Partnerships

Knowledge management

Number of projects

Partnerships and knowledge management (KM) received very little attention. No project mentioned clear partnership strategies. Given the high risks and needs in Africa, partnering with a few selected funders and developing joint approaches makes a great deal of sense. On KM, only two projects mentioned surveys as outputs during the second period, but even those did not include KM as a specific objective. Analysis of the QAE conducted as part of the comprehensive evaluation of microfinance activities shows that only 15 percent of microfinance components mention lessons learned from experience and that those lessons remain vague. Some change was observed at the macro level “…creating an enabling environment that promotes building inclusive financial systems.” Six projects included support to the macro level in the first half of the period, representing 11 percent of the total number of projects, compared to nine in the second half, representing 29 percent of projects. The macro level is at the same time one of the key challenges facing financial inclusion in Africa, and one of the Bank’s key competitive advantages, given its strong relationships with African governments. However, it is unclear whether these projects have benefited from sufficient expertise as six (6) of the 15 policy projects (6 in the first period and 9 in the second) had no microfinance or financial sector specialist in the appraisal team. Because they impact the entire system, macro level projects need to be especially well-designed, and require specific skills sets. Building market infrastructure is virtually absent from the portfolio. Although the importance of the market infrastructure/meso level was highlighted in the MPS as a key constraint, only part of this infrastructure (i.e. the professional networks) was included in the four strategic pillars: “The Bank will

The African Statistical Journal, Volume 20, February 2018

Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye

additionally provide assistance to and target the increasing number of MFI professional associations and networks.” Fifteen projects, including seven from the second half of the period, are active at this level. These projects supported the networks to provide capacity building to MFIs, but without turning these networks themselves into sustainable infrastructures. The review did not identify projects that were active at other levels of the market infrastructure (payment systems reach to microfinance, benchmarking and rating initiatives, credit bureaus, etc.). Credit components are still prevalent. Seventy-eight (78) percent of projects are components of larger, non-financial sector projects; half of them are even sub-components. Microfinance represents less than 15 percent of the total amount for most projects. The situation moved toward more standalone projects after 2006, with OPSM approving only standalones and OSHD approving more standalones. Fifty-eight (58) percent of OSHD projects approved after 2006 are components, compared to 85 percent in the previous period. However, 44 percent of the projects approved between 2007 and 2012 are components (Table 11). Table 11. Components and stand-alone projects Amount (UA million) Type of project 2000–06 2007–12 Total OSHD OSAN

OSGE OPSM

Stand-alone

101

178

127

Component

141

160

117

Total

242

338

166

153

Stand-alone

100

Component

100

Total

100

Percentage

Number of projects Type of project 2000–06 2007–12 Total OSHD OSAN

OSGE OPSM

Stand-alone

Component

Total

28.0

2.1

100.0

Percentage Stand-alone

4.8

56.3

22.3

Component

95.2

43.7

77.7

72.0

97.9

100.0

Total

100.0

Amount missing for 9 projects.

Journal africain de statistiques, numéro 20, février 2018

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012

2.4 Microfinance project types Based on all 37 approved and ongoing projects, the evaluation team identified five types of project used by the Bank to intervene in microfinance (Figure 5). Figure 5. Microfinance project types Private Sector Window (OPSM)

Public Sector Window (OSHD, OSAN, OSGE)

Model A

Model B

Model C

Model D

Equity, loans, FAPA grants

Loan, equity

Loan, grant

Budget support

Financial intermediaries

Government agency or state owned apex

Government

Trust funds Model E Grant

Grant

Holdings Loan, equity, grant

Loan, equity, grant

Regulation, market infrastructure

Market infra.

MFIs

Clients

Through Model A, OPSM directly provides equity, loans, and grants from African Private Sector Assistance (FAPA) to MFIs. As shown in Table 12, 11 projects of this type are currently ongoing/approved for UA14 million. These projects provide funding to eight different MFIs, including six greenfields. They use three types of instruments: equity participation (seven projects), FAPA grants (five projects), and loans (one project). As of 2012, these eight MFIs served approximately 190,000 people, mostly urban clients.

The African Statistical Journal, Volume 20, February 2018

Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye

Table 12. Breakdown of all approved and ongoing projects by type (as of May 2013) Number of projects 2000– 2007– Total OSHD OSAN OPSM OSGE 06 12 Model A: OPSM direct

Model B: OPSM indirect

Model C: Public sector

Model D: Budget support

Model E: Trust fund

Total

Approved amount (UA million) 2000– 2007– Total OSHD OSAN OPSM OSGE 06 12 Model A: OPSM direct

Model B: OPSM indirect

Model C: Public sector

117

Model D: Budget support

Model E: Trust fund

Total

134

Through Model B, OPSM indirectly provides funding to MFIs through financial intermediaries such as investment funds or apex banks that in turn refinance MFIs. This model was promoted in the 2007 OPSM Microfinance Delivery Approach as an aggregator to refinance MFIs. Rather than working with individual MFIs, OPSM provides larger funding amounts to intermediaries that in turn fund MFIs. OPSM did not invest in any of the funds dedicated to microfinance (e.g. Regmifa, Africap, Shorecap, Rural Impulse).11 Three issues make it difficult for the Bank to work with such funds: (i) lack of formal strategy to target these funds (OPSM’s Microfinance 11 The Symbiotics MIV Survey identified 111 active microfinance funds in 2013. See details in the Symbiotics MIV Survey Report at http://www.syminvest.com/papers#/2013.

Journal africain de statistiques, numéro 20, février 2018

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012

Delivery Approach has never been formally approved); (ii) domiciliation out of Africa; and (iii) difficulty in committing to fixed percentages of the funds invested in Africa versus other regions of the world. OPSM provided funds to Mali’s apex bank — Banque Malienne de Solidarité — with the objective of influencing that bank to refinance Malian MFIs. OPSM also invested in three non-specialist funds that invested in one or several MFIs (e.g. ECP II fund invested in the MFI Blue Financial; I&P invested in ACEP; Africinvest II invested in Family Bank).12 MFIs funded by these three investment vehicles reached around 140,000 clients by 2012. However, these three investments are not accounted for in the list of MFPs for this review because it was difficult to allocate specific amounts, and consequently they could not be tagged as microfinance in the system. Model C is used by public sector operations. This model is prevalent in the approved/ongoing portfolio with 20 of 37 projects. Through governments, OSHD, OSAN, and OSGE are able to fund all levels of the financial sector: refinancing MFIs, capacity building for MFIs, capacity building for microfinance networks and other market infrastructure, and technical assistance at the policy and supervision level. Twenty projects using Model C are ongoing within OSHD and OSAN for UA115 million. They support more than 750 financial intermediaries, mostly small savings and credit associations, which served about 270,000 clients in 2012. Three implementation models co-exist within this broad Model C. Note that this model does not include OPSM direct transactions with the private sector that receive a sovereign guarantee. •

Model C.1: Supporting MFIs through government units. At least 14 projects aim at providing refinancing and/or capacity building to MFIs through a government agency. The 2006 MPS discouraged refinancing through governments and strongly advocated for private sector leadership in funding MFIs. Government units often lack the technical expertise and independence necessary for allocating and supervising funding to MFIs.

•

Model C.2: Supporting MFIs through national state-owned apexes. Two projects are managed by state-owned apex structures (SFD in Egypt and the Urban Poverty Reduction Project with the wholesaler in Ghana), which specialized in financial sector work and pool resources from various funders. This is theoretically an improvement on Model C.1, given that the apex is specialized and invests in staff capacity. Also, the

12 OPSM invested in other funds, and specifically SME funds, but these other funds were not active in microfinance as of 2013: Grofin, Aureos/Abraaj Capital, IFHA, Catalyst Fund.

The African Statistical Journal, Volume 20, February 2018

Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye

apex can manage a revolving credit fund better than a government unit. However, capacity remains an issue for most of these state-owned apexes. For instance, SFD’s achievements are highly criticized. CGAP (2012) points out that “SFD […] has not played much of a role in catalyzing development of sustainable MFIs in Egypt.” •

At least seven of the Model C.1 projects aim at transforming government units into national state-owned apexes that would manage the Bank’s revolving credit fund. Task managers envision this transformation as a solution to the current disconnect between MFIs’ repayment to government and government repayment to the Bank. However, in light of the above paragraph, creating national state-owned apexes requires a rigorous assessment of past experiences to identify key success and failure factors before replicating the model. Having multiple funders in place to provide economies of scale and transparent governance could guarantee independence from the state and lead to success.

•

Model C.3: Supporting market infrastructure and policy levels through government units. At least four public sector projects currently involve support to microfinance networks, and six OSHD projects support the policy level (training of regulators/supervisors, technical assistance to review legal framework). The 2006 MPS encouraged such projects.

Through Model D, public sector departments support the overall country budget under given conditions. These conditions include specific changes at the policy and market infrastructure levels for the microfinance sector. No specific amount is allocated to microfinance. Two projects are ongoing: one OSGE project in Morocco and one OSHD project in Tunisia. The 2006 MPS encouraged such projects. The Bank also facilitated the creation of technical assistance funds to leverage other funders’ resources. Through Model E, the Bank is managing two funds that support microfinance: MCBF, which is focused on microfinance; and the Migration and Development Trust Fund, which can fund microfinance projects that are linked to remittances. Three capacity-building projects were approved under the MCBF as of May 2013. Management of these funds was transferred from OPSM to OSHD in April 2012. The assessment of the performance by type of project could not be undertaken because of a lack of information. However, such an analysis has been performed in the summary report of the evaluation by cross-referencing

Journal africain de statistiques, numéro 20, février 2018

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012

various sources of information, including those from this portfolio review, case studies, and interviews. 3.

PERFORMANCE EVALUATION

The performance of all projects could not be assessed due to the lack of project performance data in the system. Four sources were used, each limited to a sub-set of projects (Table 13): Table 13. Sample of projects for performance evaluation Departments No. of No. of and status projects in projects portfolio in scoring sample Total, Bank

OSHD

OSAN

OSGE

No. of projects in PCR review

No. of projects No. of for implemen- projects for tation perfor- quality at enmance analysis try analysis

OPSM

Approved

Ongoing

Closed/ completed

•

Scoring exercise to collect performance data for on-going and approved projects through a questionnaire sent to task managers. The response rate from task managers was high. The exercise revealed that some projects registered as ongoing were actually cancelled, closed, or just approved with no activities yet. Of the 29 ongoing projects, 22 had performance data. A total of 25 projects were assessed and scored. Rather than comparing achievements to initial goals, this performance evaluation assessed the relevance and sustainability of the project overall. Therefore, the evaluation focused on the recipient organizations. It strived to identify the type of organizations receiving Bank funds and the added value of the Bank’s funding.

The African Statistical Journal, Volume 20, February 2018

Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye

•

Review of 43 PCRs. Although 57 projects had been closed by the time of the review, only 43 PCRs were available, and just 12 of these provided information on the microfinance components.

•

Implementation analysis. Using SAP and PCRs, the evaluation team collected key project dates (such as signature, effectiveness, first disbursement, and closing) for 71 projects in order to assess efficiency and compliance with internal norms and deadlines.

•

Review of the QAE. This portfolio review also built on the main conclusions from the QAE exercise conducted using a sample of 40 projects (half of them approved between 2000 and 2006, and the other half between 2006 and 2012; 29 were components of larger projects, while the remaining 11 were stand-alone microfinance operations).

3.1 Relevance and additionality Relevance and additionality were assessed based on (i) the 25 ongoing projects and (ii) the QAE for 40 projects. For the 25 ongoing projects, the evaluation team assessed whether the Bank assistance was relevant with regard to market/country needs and national strategies, whether added value was created that would not have been achieved without Bank support, and whether the design of the projects matched the intended objectives.13 High scores were granted to projects that targeted underserved clients, encouraged a diversity of services, built on the Bank’s competitive advantages without creating market distortion, and aligned with the Bank’s strategy and international good practices. Projects that did not start operations within three years of approval were rated low, as market conditions change rapidly and project relevance can no longer be guaranteed. Only 60 percent of the 25 ongoing projects scored “moderate and higher”. Six (6) ongoing projects were deemed to have relevant objectives and to be well designed; nine (9) projects were deemed to have relevant objectives, but unsatisfactory design; and 10 projects were scored low on relevance and project design (Table 14).

13 The evaluation team focused on the objectives of the project, context at the time of the project (post crisis/stable, microfinance market maturity, presence of other funders), intended outreach (number of end clients, rural/urban, number and type of MFIs), and total amount approved.

Journal africain de statistiques, numéro 20, février 2018

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012

Table 14. Relevance and additionality for ongoing projects Relevance/addition- Numality ber

Number of projects

% of number of projects

OSHD OSAN OPSM OSHD OSAN OPSM

Relevant and welldesigned (satisfactory)

Relevant but design/ outreach could be improved (moderate)

Questionable design and market needs assessment (unsatisfactory)

Total

100

OPSM scored high on relevance. Four ongoing projects were deemed relevant and well-designed. Two of these (MicroCred in Cote d’Ivoire and Advans Congo in Democratic Republic of Congo) were implemented in post-conflict countries. They contributed to the creation of financial intermediaries with adequate product range and good prospects of sustainability. One project, Trustco in Namibia, focused on financial innovation. The other project, KREP equity participation, also received a high score in view of its relevance at the time of the project; however, the Bank should reconsider its role in the equity and governance of this MFI given the governance crisis it is facing. OPSM’s Banque Malienne de Solidarité (BMS) project in Mali was rated as “unsatisfactory” for providing additional liquidity to struggling MFIs that need restructuring to be able to absorb more liquidity. Given that OSAN is no longer very active in microfinance, only five of its projects are included in the sample of 25 ongoing projects. One of the projects, Appui aux Communautés de Pecheurs in Madagascar, earned a “satisfactory” rating for its targeting semi-urban areas in its funding. Two projects scored “unsatisfactory” because they had not started operations for over three years after approval. Other projects scored “average” because their purpose was in line with country needs, but their project design and implementing partners did not guarantee maximum added value. OSHD scored weak on relevance, mostly because of design issues: 7 of 11 projects were in the low range:

The African Statistical Journal, Volume 20, February 2018

Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye

•

2 projects (Pauvreté urbaine Djibouti and Gender Responsiveness Ghana) approved in 2007 and 2008, respectively, were still inactive and received a very low score.

•

2 projects (Small Entrepreneurs II Tanzania and RIEEP Uganda) provide comprehensive information on outreach in terms of number of clients and on exact use of the funds. It is, however, commendable that the two projects maintained a list of all MFIs funded by the Bank (357 SACCOs for Tanzania and 400 for Uganda).

•

3 projects had limited outreach compared to the amount disbursed (Urban Poverty Reduction Project Ghana, RIEEP Egypt and Entrepreneurship Promotion Gambia).

•

Also, the Mauritanian project, Renforcement des Capacités, was scored high for relevance, given that few other funders are active in the country and that the project met the country’s needs through its effective support of various levels of the financial sector.

Table 15. QAE: projects scoring “moderate” or higher (%) Criteria

2000–12 (n = 40)

2000-06 (n = 20)

2007-12 (n = 20)

Strategic relevance

87a

Poverty, gender and social development

Risk identification

Technical and financial aspects

Project design

Project implementation arrangements

Environmental impact

Average

This decrease in strategic relevance is due to the four trust fund projects; their appraisal reports do not sufficiently assess the alignment of projects with the Bank’s microfinance strategy and the countries’ strategies. a

The major conclusions from the QAE analysis confirm the low performance. Only 52 percent of the 40 projects under review scored “moderate” or higher on QAE. This performance is particularly low compared to other Bank’s operations: scoring “moderate” or higher were 68 percent of multinational projects; 71 percent of education projects (38 projects approved between

Journal africain de statistiques, numéro 20, février 2018

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012

1975 and 2005); 76 percent of all public sector operations approved in 2005 (19 projects); and 81 percent of all public sector operations approved in 2008 (20 projects). Overall, projects scored relatively high on two of the QAE sub-criteria: “strategic relevance” and “poverty, gender and social development aspects,” while scoring low on “project design”, “project implementation arrangements”, and “environmental impact” (Table 15). Improvement was significant over the period, with a QAE score of 60 percent for “moderate and higher” from 2007 to 2012, compared to 43 percent from 2000 to 2006. However, this improvement is largely attributable to a higher number of standalone projects in the second period compared to the first period, rather than to a sound adjustment of procedures. Overall, standalone projects scored much higher than components, with a score of 74 percent as “moderate or higher” compared to only 42 percent for components. This is mostly because standalone projects are designed by financial sector experts (who also often have microfinance experience) and their project appraisal reports provide detailed analyses of the microfinance operations. Meanwhile, only three of the 29 component projects were designed or reviewed by a microfinance expert, and their project appraisal reports only include a maximum of three pages on microfinance components, which provides a limited understanding of the market and the project’s planning. The QAE for component projects has not improved significantly over the years, with a score of 45 percent as “moderate and higher” from 2007 to 2012, compared to 43 percent from 2000 to 2006. PCRs did not prove to be a good source of information for the assessment of project relevance. Of the 43 available PCRs, only six (6) provided clear assessment of the relevance of the microfinance operations. All projects scored satisfactory or higher, mostly on the strength of their alignment with the national poverty reduction, job creation and economic growth strategies. PCRs overlooked other aspects of project relevance, such as quality of project design and implementation partners, alignment with microfinance good practices and the Bank’s microfinance strategy, and targets compared to funding volume. 3.2 Effectiveness In the absence of clear targets and a centralized performance reporting system, the evaluation team used the available PCRs to assess MFP effectiveness. Only 12 of the 43 available PCRs provided information on microfinance operations, and 11 of them deemed effectiveness unsatisfactory or worse. The PCRs identified several reasons for this underperformance: unrealistic targets, implementation delays, lack of innovation, and lack of efficient

The African Statistical Journal, Volume 20, February 2018

Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye

monitoring and evaluation. Six (6) OSAN component projects were cancelled because of these deficiencies, and several were reorganized. Cancelled projects are the following: Artisanal Fisheries Development in Sierra Leone, Small Scale Irrigation in Zambia, Livestock Productivity Improvement in Uganda, Rural Development of Biltine in Chad, Family Sector Income Enhancement in Mozambique, and Livestock Development Support–Phase II in São Tomé. Only the stand-alone Rural Finance Intermediation Support Projection in Ethiopia was deemed highly satisfactory in the PCR, though only satisfactory in the project completion report note. 3.3 Efficiency On average, it takes around a year from signature to the first disbursement, i.e. not even including time spent on project preparation and design (Table 16). These relatively long delays raise questions about the validity of the market analysis performed in the early days of project design, compared to the market situation at the implementation stage. Forty-eight (48) percent of the projects did not meet the effectiveness deadline, which corresponds to six months following signature. Similarly, 48 percent of the projects failed to meet the first disbursement deadline of two months after effectiveness (Table 16). OPSM and OSGE projects performed better than OSHD and OSAN projects. Projects from the second period performed better as well. Seventy-two (72) percent of projects from the first phase closed more than 12 months after the planned date, or were cancelled. On average, projects closed 28 months after their planned closing date (Table 17). This time overrun is pronounced in the public sector projects. As of May 2013, eight (8) projects approved between 2010 and 2012 were still inactive. Six (6) OSAN components were cancelled because of very low performance. Table 16: Implementation performance: signature to closing 2000– 2007– 2000– OSHD OSAN OSGEa OPSM 06 12 12 Signature to effectiveness

(n=56)

Average number of months between signature and effectiveness

7.8

3.1

Signature to effectiveness, % projects above 6 months (Bank’s norm)

55%

20%

(n=15) (n=71)

(n=19)

(n=43)

(n=7)

(n=2)

6.8

4.7

8.8

1.5

3.1

48%

42%

06%

Journal africain de statistiques, numéro 20, février 2018

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012

Effectiveness to disbursement

(n=15) (n=71)

(n=19)

(n=43)

(n=7)

(n=2)

Average number of months between effectiveness and first disbursement

4.5

3.1

4.2

2.9

5.3

2.7

0.0

Effectiveness to 1st disbursement, % projects above 2 months (Bank’s norm)

47%

50%

48%

44%

55%

29%

(n=40)

(n=3)

(n=43)

(n=6)

(n=32)

(n=5)

(n=0)

8.2

1.8a

7.8

7.7

8.7

1.6

n/a

Signature to closing Average number of years between signature and effective closing (excluding cancelled projects) a

(n=56)

Three OSGE projects closed after 1.7; 1.1; and 2.6 years.

Table 17. Implementation performance: planned vs. effective closing date

% of projects cancelled or closed more than 12 months after planned closing date

2000– 06

2007– 12

Total

OSHD

OSAN

OSGE

OPSM

(n=50)

(n=3)

(n=53)

(n=9)

(n=39)

(n=5)

(n=0)

72%

68%

n/a

3.4 Sustainability To assess the probability of sustained long-term benefits (i.e., leading to sustainable access to finance), the evaluation team focused on the operational and financial sustainability of the financial intermediaries, the MFIs. How are these institutions performing? Are they able to sustain their operations without ongoing donor funding? To respond to these questions, the evaluation team rated a subset of ongoing projects, mostly on the financial performance of the MFIs, as reported by task managers or as found on the web. Projects for which no precise data was submitted on MFI outreach and financial performance, received a low rating. Projects whose beneficiary institutions were likely to be or were already sustainable received a high rating.

The African Statistical Journal, Volume 20, February 2018

Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye

Sustainability is a key weakness for most projects. Only half of the 22 ongoing projects worked with MFIs likely or very likely to be sustainable (Table 17). For four projects, the evaluation team could not assess MFI sustainability because of lack of data, not only in the Bank’s system but also in standard microfinance reporting places such as the MixMarket (MIX), a web-based performance reporting platform, or anywhere on the Internet. The lack of data is an indication of the weakness of the financial institutions. Financial institutions need to be transparent in the use of aid funds; most microfinance funders today require their recipients to report to the MixMarket. The situation is similar for wholesale institutions, which use Bank funding to refinance MFIs: only three out of the ten funds/apexes are likely to be sustainable over time (Table 18). If Bank support does not continue after the completion of the projects, activities may stop. Table 18. Sustainability of Bank’s funds’ recipients (number of projects) MFIs

Total

OSHD OSAN OPSM

MFIs funded by the project likely/very likely to be sustainable

MFIs funded by the project unlikely to be sustainable

Lack of data on MFI performance

Total number of projects funding MFIs

Wholesale institutions

Total

OSHD OSAN OPSM

Wholesaler funded by the project likely/very likely to be sustainable

Wholesaler funded by the project unlikely to be sustainable

Lack of data on wholesaler performance

Total number of projects funding whole salers

Questionnaires completed by task managers show that 25 projects were funding over 800 MFIs in Africa.14 Projects work with very different types of MFIs, but the size of asset and clientele of the majority of the recipient MFIs is typically small. In total, the Bank is working with approximately 6 commercial banks, 6 microfinance banks, 20 rural banks, 6 non-Bank financial institutions, 80 NGOs, and 700 SACCOs of various sizes. Less than 30 of the institutions the bank is working with report information to 14 Some projects grouped savings and credit cooperatives (SACCOs) as full-fledged MFIs, while other projects only report their apex institutions.

Journal africain de statistiques, numéro 20, février 2018

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012

the MIX Market or on a public website, and those that do report seldom update their information. OPSM is mostly working with new and greenfield institutions, while OSHD and OSAN are working with older MFIs (Table 19). Table 19. Maturity of Bank-funded MFIs (number of projects) MFIs maturity at project approval

Total

OSHD OSAN OPSM

Greenfield

< 5 years

> 5 years

Lack of data

No MFIs funded yet

Total

More information is needed on the refinancing operations of the Bank in order to assess the Bank’s impact on the sustainability of local finance markets. The African financial landscape has evolved in the past 12 years, offering MFIs increasing access to local sources of funds. Because of insufficient data, this review could not analyze the refinancing conditions to ascertain whether the Bank and its partners were offering fair rates to MFIs and not crowding out local funders. Further research is essential to determine whether the impact of the Bank’s funding approaches on local fund markets distorts local markets – For instance, could the offer of subsidized funds be distorting local markets by preventing local banks from investing local savings by refinancing MFIs at commercial rates? 3.5 Overall performance To consolidate the overall rating for the subset of 25 projects, the team weighted the rates on relevance, MFI sustainability, and wholesale sustainability, and included task managers’ feedback on other topics such as support to the enabling environment. As shown in Table 20, only 72 percent of the 25 ongoing projects were rated “moderate and higher”. The reviewer would not approve seven (7) of the 25 projects if they were submitted today. Even though task managers tend to be softer in terms of rating, they also only rated 81 percent of the projects “moderate” or higher. Stand-alone projects perform better than components (67 percent satisfactory for standalones versus 8 percent for components). Three (3) of the seven (7) projects were rated unsatisfactory because they had not disbursed to MFIs funds that had been approved more than five years previously.

The African Statistical Journal, Volume 20, February 2018

Satisfactory

Moderate

Unsatisfactory

Total

100

Very satisfactory

Satisfactory

Moderate

Unsatisfactory

Total

Percentage

Total

Very satisfactory

Assessment

Table 20. Overall project scores

100

OSHD

100

OSAN

Components

100

Number of projects

OPSM

Rated by the portfolio evaluation team

100

Stand alones

100

Total

100

OSHD

100

OSAN

Rated by task managers

100

OPSM Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye

Journal africain de statistiques, numéro 20, février 2018

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012

3.6 Key success factors By comparing satisfactory and unsatisfactory projects, the following key success indicators for project performance were identified: •

Involvement of microfinance experts in the design and implementation of projects.

•

Strong market analysis during project design, going beyond analysis of end users need and potential partners to include all levels and players of the markets. Projects are required to support the creation of a robust ecosystem rather than to support specific target sectors.

•

Solid implementing partners with microfinance expertise, whose objectives are aligned with the project’s objectives that are sustainable and capable of absorbing/managing the Bank’s funds.

•

Well-conceived financing packages that provide for the supply of capacity building funds to the MFIs to support their business plans.

•

Performance targets included in project design, based on the partner’s business plan, and tracked at all levels.

•

A well-functioning performance management system.

MAIN FINDINGS AND CONCLUSIONS

Overall, the microfinance portfolio performance was unsatisfactory. None of the 25 ongoing or approved projects attained a “very satisfactory rating”, 36 percent scored “satisfactory”, and 28 percent scored “unsatisfactory”. Sixtyseven (67) percent of standalone projects scored “satisfactory”, compared to only 8 percent for components. OPSM operations performed better than public sector operations. Relevance. Project objectives mostly responded to the needs and strategies of RMCs, but their relevance was often hindered by unsatisfactory design. Sixty (60) percent of the 25 approved/ongoing projects’ objectives were deemed relevant, i.e. appeared both to meet the country needs and to be aligned with microfinance good practices. Only 24 percent, however, were well-designed. The main conclusions from the QAE analysis confirm the low design quality. Only 52 percent of the 40 projects under review scored “moderate or higher” on QAE. The improvement over the period, with a

The African Statistical Journal, Volume 20, February 2018

Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye

QAE score of 60 percent for “moderate and higher” from 2007 to 2012, compared to only 43 percent from 2000 to 2006, is mostly due to a higher number of stand-alone projects in the second period compared to the first, rather than to a sound adjustment of project procedures. In relevance, standalone projects scored higher than components, and OPSM projects scored higher than public sector projects. Effectiveness. In the absence of clear targets and a centralized performance reporting system, the evaluation team used the available PCRs to assess MFP effectiveness. Only 12 of the 43 available PCRs provided information on microfinance operations. 11 of them with an “unsatisfactory” or worse effectiveness rating. Six of the 48 OSAN component projects were cancelled because of poor performance and several others were reorganized. Efficiency. Efficiency was deemed low because of late disbursement on average 11 months after signature instead of the Bank-stipulated five (5) months, excluding time spent on project preparation and design. These long delays raised questions about the validity of the market analysis performed in the early days of project design, compared to the market situation at the time of implementation. Seventy-two (72) percent of projects from the first phase closed more than 12 months behind schedule, or were cancelled. As of May 2013, eight (8) projects approved between 2010 and 2012 were still inactive. Sustainability. Sustainability is a key weakness for a large number of projects, with the exception of most OPSM projects. Only half of the 22 ongoing projects worked with MFIs likely or very likely to be sustainable, and less than a third of wholesale institutions funded or being created by the Bank were deemed likely to be sustainable over time. If Bank support ends on the completion of the current projects, most activities will rely on other subsidized support or come to a complete halt. Private sector projects are much more likely than public sector projects to produce sustainable results – only one of the nine (9) projects working with MFIs was deemed unlikely to be sustainable (Banque Malienne de Solidarité). 4.1

The project portfolio did not fully leverage the Bank’s competitive advantages The Bank did not use its broad suite of financing instruments. Compared to other development finance institutions, the Bank has a diversified range of financing instruments to meet varying market needs and to avoid crowding out private investors. Instead, the Bank focused on debt funding as its main instrument. It complemented this with a small but unclear grant strategy, and a small equity portfolio. The Bank did not use any guarantee. Given

Journal africain de statistiques, numéro 20, février 2018

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012

the wide market conditions in Africa, including the presence of mature institutions, it is unclear that debt is the right instrument for all conditions. The Bank is not drawing upon its extensive sources of knowledge and experience. For example, the Bank has a strong in-house research team, over 30 years of involvement in microfinance, strong relationships with member governments, and its experience through Making Finance Work for Africa. Knowledge management remains a weak link in the portfolio. The Bank is not extracting lessons from its own research or projects or building on knowledge from others working in this sector. The Bank’s strong credentials on regionalization is not evident in its support to microfinance. The Bank has a dedicated regional integration department, and yet lags behind other stakeholders in the exploration of regional policy, regulatory and financing solutions to financial sector development, and financial inclusion in Africa. 4.2 Strategic orientations were never fully implemented Although the 2006 Microfinance Policy and Strategy is built on internationally recognized good practices, the changes it promoted are not sufficiently integrated into the portfolio to ensure the relevance of projects. For instance, OPSM was supposed to “provide leadership in the area of private sector oriented investment,” yet 83 percent of the Bank’s ongoing support to private sector institutions between 2007 and 2012 still came from the public sector window. OPSM remained limited in terms of volume, with only UA16 million approved between 2007 and 2012 compared to more than UA 80 million for public sector operations during the same period. The strategic pillars of partnerships and knowledge management received very little attention. In addition, Bank projects seldom address the need to strengthen the African financial market infrastructure. The MPS also encouraged collaboration across departments of the Bank, but such collaboration remains very limited. Of the 35 countries that received assistance from the Bank towards microfinance, only five (Democratic Republic of Congo, Liberia, Mali, Tanzania and Zambia) received funding from both private and public sector windows. Although this difference in coverage makes sense in the short term, largely because public and private sector windows have different targets, it is surprising to see so little overlap over a 12-year period. One would expect some institutions supported by public sector could graduate over time and become good prospects for the private sector operations. This situation reflects shortcomings in strategic thinking, which needs to encompass both public and private sectors.

The African Statistical Journal, Volume 20, February 2018

Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye

4.3

Execution capabilities are not aligned with the portfolio size and complexity The current specialized-staff-to-project ratio is insufficient for adequate management of the large microfinance portfolio and the processing of new operations. In 2012, only two professional staff were dedicated to microfinance, one in OPSM and the other in OSHD. These two were charged with managing projects for over 800 MFIs spanning 21 countries, as well as with processing new operations. In no phase since 1998 were there adequate staff resources allocated to microfinance. In 1999, when the Bank launched the ADF Microfinance Initiative for Africa (AMINA), four staff members were assigned. The staff of eight the MPS needed to run the microfinance unit in 2001 was never recruited. This lack of specialized staff has a direct impact on project QAE and consequently the relevance, effectiveness, efficiency, and sustainability of the projects. Microfinance projects are often managed by committed task managers who work on multiple issues and sectors. Only three (3) of the 29 component projects included in the QAE analysis were designed or reviewed by microfinance experts. Only eight (8) of the 15 ongoing projects involved at the policy level included a microfinance or finance sector expert in the appraisal mission. 4.4 Project effectiveness is not adequately monitored The performance monitoring system does not promote accountability. The lack of a centralized performance data reporting system for microfinance projects prevents solid performance analysis and inhibits task manager accountability. The absence of a clear Bank-wide coding system to define microfinance projects makes it very difficult to identify the Bank’s microfinance-related projects or to guarantee an adequate supervision process. There is no mandatory indicator in microfinance projects, no centralized system to report on such indicators, and no regular portfolio meeting that would enable task managers compare performance across the Bank. Unlike other major funders, the Bank is slow to adapt to global developments in monitoring and evaluation (M&E) in the microfinance sector. Other funders, notably, FMO and the Gates Foundation, have developed very ambitious targets in terms of microfinance M&E. 4.5 Microfinance components cannot integrate key success factors Credit components prevail despite evidence of their limited effectiveness, and despite repeated internal and external recommendations

Journal africain de statistiques, numéro 20, février 2018

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012

for the avoidance of this type of project.15 The Bank’s 2006 strategy highlighted the limited effectiveness of credit components. Designed as inputs to larger multi-sector projects, credit components performed poorly because of five QAE flaws. First, the project team usually lacked microfinance expertise. Second, market analysis was limited, and because of length restrictions, project appraisal reports dedicated only three pages or less to the microfinance operations. Third, the project sponsor, i.e. the government implementing agency, typically lacked the requisite expertise. Fourth, microfinance funding and performance indicators were not properly tracked, and wound up lost within overall project management. Finally, the project timeline was usually disconnected from the requirement of microfinance implementation (long-term projects with government versus heavy but short-term technical assistance for microfinance). Although the Bank tried to improve credit components after 2006 (by, for example, involving a microfinance specialist in most OSHD projects), it did not demonstrate significant improvement in the design and performance tracking. Task managers faced difficulties inherent to this type of project: they can spend only a limited amount of time on designing operations that represent, on average, less than 15 percent of the total project amount. The selection of project implementing agencies was based on their relevance to the main project operations rather than their adequacy in microfinance. 4.6 Bank’s procurement processes reduce efficiency PCRs highlight the Bank’s long and tedious procurement processes, mostly due to (i) the Bank’s no-objection requirement; (ii) ratification by the country’s parliament; (iii) appointment of a qualified project manager; (iv) implementation of the financial counterparts; and (v) poor performance of local bidding companies in the case of public sector projects.

15 Many agencies recognize that credit components within larger multi-sector programs do not produce intended results. It has been recommended that components should be avoided or, at the very least, designed by financial and microfinance experts and implemented in line with good practices, while separating them from grant components or other types of support (CGAP 2006). Overall, microfinance components rarely continue beyond the project and thus do not provide permanent access to financial services.

The African Statistical Journal, Volume 20, February 2018

Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye

5. RECOMMENDATIONS

5.1

Refocus the strategy based on Bank’s strengths, capacity, and competitive advantages and develop operational guidance With the recently created department of the financial sector, the Bank should build on the results of this evaluation to conduct a strategic reorientation based on a SWOT analysis. In light of internal resources and scope of operations, the strategic focus should be limited to a few selected themes that build on existing strengths. To help translate the strategy into operations, the business plan should include baseline data and targets, and operational guidelines should include responsibility assignment. 5.2 Align execution capabilities with portfolio size and complexity To effectively design, process, monitor, evaluate, and deliver high quality and scalable projects on the ground demands strong technical skills and time. The Bank should invest in the recruitment of microfinance specialists and, at the same time, provide opportunities for task managers and non-specialists to hone skills needed to successfully process and manage microfinance projects. To promote ongoing learning and staff development in this evolving industry, the Bank could leverage its partnership with CGAP and MFW4A to learn from its peers, and encourage staff participation in external trainings such as the Boulder Microfinance Program. The Bank should not expand operations before sufficient qualified staff are hired, and a strong knowledge management program put in place. 5.3

Improve reporting and monitoring systems to learn from project implementation The Bank should set up clear criteria to identify financial inclusion projects, including components within larger projects. Currently, microfinance and SME finance are considered and managed separately, but the line is blurred and some projects reported as microfinance in SAP could be considered SME by other funders. Also, financial education programs are not included in the SAP microfinance portfolio, some projects are reported as microfinance in SAP (although the Bank is not funding their microfinance component), and indirect investments in MFIs via mainstream equity funds are not reported as microfinance. The Bank also needs to reorganize its M&E system, including developing a list of key output, outcome, and impact indicators for all-inclusive finance standalone and component projects. Building on the Additionality and Development Outcome Assessment (ADOA) system would represent a major achievement. Such a standardized performance measurement process

Journal africain de statistiques, numéro 20, février 2018

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012

should be expanded to public sector operations. Microfinance standard performance indicators should also be integrated, not only to track additionality and development outcomes, but also to keep track of the sustainability of beneficiaries Rosenberg, 2009 offers standard indicators at the MFI level, and CGAP 2010 for standard indicators for microfinance wholesalers). Additionally, a centralized database to capture and monitor portfolio data across the Bank will facilitate oversight and strengthen the Bankâ&#x20AC;&#x2122;s quality assurance system. A designated staff from Operations should be tasked with updating and verifying the accuracy of data. The Bank should require its recipient organizations to report to global microfinance platforms such as the Mix Market, and encourage them to have clear, transparent and audited financial statements. The Bank should consider participating in funder M&E initiatives such as the Social Performance Task Force social investors working group to learn from peer institutions. 5.4 Avoid microfinance components If a microfinance component cannot be avoided, it should be processed as a standalone project, specifically including: (i) an appraisal report with detailed market analysis and evaluation of implementing partners; (ii) an adequate timeline based on microfinance operations necessities rather than an overall project timeline; (iii) a technical review and monitoring performed by a microfinance specialist; and (iv) standard microfinance performance indicators. Microfinance components should not be used as inputs to broader non-financial sector projects, i.e. they should not be added to projects to help reaching agriculture and social development goals. Microfinance operations should be pursued with the central objective of building inclusive financial sectors. 5.5

Use lessons from project performance to replicate successful initiatives The lack of a standard approach for Bank projects provides limited potential for scale and replication. To promote efficiency, scalability, and learning, several funders focus on a few selected project types that they replicate. For instance, the KfW and IFC greenfielding initiative in Africa identified a need for increased competition among microfinance players in several countries, resulting in the decision to actively promote the creation of new MFIs. Most of these greenfields follow similar business models, enabling KfW and IFC to compare, learn, and better manage each project. The Bankâ&#x20AC;&#x2122;s strategy and project documents do not mention any such innovative framework.16 There is no clear reference to models that were developed, tested, and recommended 16â&#x20AC;&#x201A; While the Bank sets up finance greenfields through OPSM, the Bank is not actively promoting the creation of institutions. The OPSM 2007 Microfinance Delivery Approach

The African Statistical Journal, Volume 20, February 2018

Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye

for replication. This appears to be a missed opportunity to learn from past projects and to gain efficiency. The Bank could establish common patterns and promote the most successful project types. The five project types identified in this review could serve as a basis for such an exercise. 5.6 Improve the Bank’s processes The Bank should streamline its procedures for processing and approving with a view to accommodating the transactions of small projects. ACKNOWLEDGMENTS

This article is based on the microfinance portfolio analysis carried out by the authors in 2014. It received useful comments and support from the following Bank staff: Cecile Ambert, Soumandra Koumar Dash, Damien Onyema Ihedioha, Issahaku Budali, Lilian Macharia, Leila Mokadem, Mohamed Manai, and Thouraya Triki. Mayada El-Zoghbi, Senior Microfinance Specialist and Head of the Paris office of the Consultative Group to Assist the Poor (CGAP) and Mark Flaming of Frontier Ventures, Development Finance Specialist, served as external peer reviewers and provided invaluable inputs. Zekebweliwai Fuh Kah Geh of CGAP also reviewed an earlier version of the paper and provided useful comments. The findings, interpretations and conclusions presented in this paper are the sole responsibility of the authors. REFERENCES

AfDB (2004). “2002–2003 Annual Portfolio Performance Review,” Chapter 4: Microfinance Review. AfDB (2006). “Microfinance Policy and Strategy for the Bank Group.” (2006–2011). AfDB (2006). “Microfinance Portfolio Cleanup Report,” OPSM, July. AfDB (2011). “Preliminary Review of the Bank’s Microfinance Activities,” December. AfDB (2012). “Stock-taking Exercise of the Bank’s Microfinance Activities, “January. promoted a certain type of funding, i.e. indirect funding through intermediaries, but this approach did not fully materialize.

Journal africain de statistiques, numéro 20, février 2018

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012

CGAP (2010). “CGAP Microfinance Investment Vehicles Disclosure Guidelines.” CGAP (2012). “CGAP Microfinance Cross-border Survey.” CGAP (2011). CGAP Microfinance Cross-border Survey.” CGAP (2006). “CGAP Good Practice Guideline for Funders of Microfinance.” CGAP (2012). “CGAP Focus Note. A New Look at Microfinance Apexes.” Gakusi, A. E., M. Diomande, and A. Touré (2013). “Fostering Inclusive Growth in Africa: An Evaluation of the Bank’s Microfinance Policy, Strategy and Operations, 2000–2012, an Approach Paper”. Update, OPEV, July. OSAN (2007). Technical Note: Bank Group Rural Finance Portfolio – Improving quality and performance. Rosenberg R. (2009). Measuring Results of Microfinance Institutions: Minimum Indicators that Donors and Investors Should Track, CGAP. Scola-Gähwiler, B. and A. Nègre (2012). Portfolio Reviews Resource Guide for Funders. A Technical Guide, CGAP. Symbiotics MIV Survey Report at http://www.syminvest.com/papers#/2013.

The African Statistical Journal, Volume 20, February 2018

Albert-Enéas Gakusi, Alice Negre, Mabarakissa Diomanade, and Gloria Uwingabiye

ANNEX. ABBREVIATIONS AND ACRONYMS AfDB ADF ADOA AECID AFD AGRA AMINA BDS BMS CGAP EDRE EIB FAPA FMO IDEV IFAD IFC KfW KM LOC M&E MCBF MDF MFP

African Development Bank African Development Fund Additionality and Development Outcome Assessment Spanish Agency for International Development Cooperation Agence française de développement Alliance for a Green Revolution in Africa ADF Microfinance Initiative for Africa Business Development Services Banque Malienne de Solidarité Consultative Group to Assist the Poor Development Research Department European Investment Bank Fund for African Private Sector Assistance Dutch Development Bank Independent Development Evaluation Department International Fund for Agricultural Development International Finance Corporation Kreditanstalt für Wiederaufbau (German government-owned development bank) Knowledge management Line of credit Monitoring and evaluation Microfinance Capacity Building Trust Fund Migration and Development Fund Microfinance Project

Journal africain de statistiques, numéro 20, février 2018

1. Statistical and qualitative assessment of the design quality, performance, effectiveness and sustainability of the assistance of the African Development Bank to microfinance, 2000-2012

MFI MIX MPS NEPAD NTF OFSD OPEV ONRI OPSD OPSM OSAN OSGE

Microfinance institution Microfinance Information Exchange Microfinance Policy and Strategy New Partnership for Africa’s Development Nigeria Trust Fund Financial Sector Development Department Operations Evaluation Department Regional Integration and Trade Department Private Sector Department Private Sector and Microfinance Operations Department Agriculture and Agroindustry Department Governance, Finance and Economic Management Department OSHD Human Development Department PIU project implementation unit PCR Project Completion Report PROPARCO Société de Promotion et de Participation pour la Coopération Economique QAE Quality at entry RMC Regional Member Country SACCO Savings and Credit Co-operative SAP Systems Applications and Products SELF Small Enterprise Lending Facility SME Small and medium enterprise SWOT Strengths, weaknesses, opportunities, threats

The African Statistical Journal, Volume 20, February 2018

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas Vincent Musoke Nsubuga1

Abstract The selection of items for continuous price measurement in geographical areas/ regions depends on the item’s importance in each consumption basket derived from the proportion of household consumption expenditure for the Elementary Aggregates (EA) of items and the availability of the same items in relevant markets. Based on household consumption expenditure data for Uganda, some prominent items in the consumption basket whose prices are permanently missing in specific regional markets are imputed using prices of similar items from other regions mainly the Capital City. The approach is generally likely to give rise to some level of bias to the resultant indices. The alternative would be to exclude such items from the regional basket and include their weights in the regions where they are readily available in the markets. This would lead to a variation in consumer price index (CPI) for the regions in question although the national CPI would remain unchanged. The extent of the change in the regional CPI is the subject this paper seeks to investigate by focusing on used vehicles that households from other regions purchase from Kampala Capital City. The results from this investigation will help price statisticians in understanding the implications of data collection procedures and subsequent computation of regional CPIs. Key words: Elementary aggregate, household consumption expenditure, consumption basket Abstrait Le choix des articles pour la mesure continue des prix dans les zones géographiques / régions dépend de l’importance de l’article dans chaque panier de consommation dérivée de la proportion des dépenses de consommation des ménages pour les agrégats élémentaires (AE) et de la disponibilité des mêmes articles sur les marchés pertinents. Sur la base des données sur les dépenses de consommation des ménages pour l’Ouganda, certains articles importants du panier de consommation dont les prix sont définitivement manquants sur des marchés régionaux spécifiques sont imputés en utilisant les prix d’articles similaires provenant d’autres régions principalement la capitale. L’approche est généralement susceptible de donner lieu à un certain biais dans les indices qui en résultent. L’alternative serait d’exclure ces éléments du panier régional et d’inclure leurs poids dans les régions où ils sont facilement disponibles sur les marchés. Cela conduirait à une variation de l’indice des prix à la consommation (IPC) pour les régions 1 Nsubuga Vincent Musoke, Principal Statistician, Uganda Bureau of Statistics.

Journal africain de statistiques, numéro 20, février 2018

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas

en question, bien que l’IPC national resterait inchangé. L’ampleur de la variation de l’IPC régional est l’objet que le présent document cherche à étudier en se concentrant sur les véhicules d’occasion que les ménages d’autres régions achètent dans la capitale, Kampala. Les résultats de cette enquête aideront les statisticiens des prix à comprendre les implications des procédures de collecte de données et le calcul ultérieur des IPC régionaux. Mots clés : agrégat élémentaire, dépense de consommation des ménages, panier de consommation I.

BACKGROUND

1.1 Introduction Uganda is an East African country with a population of 34.6 million as per the 2014 Uganda Population and Housing Census (UPHC). Uganda compiles and disseminates the national consumer price index (CPI) monthly. The national CPI depends on two categories of stratification: geographical stratification and income group stratification. Individuals and organizations use each geographical area and income group’s CPI for decision-making and policy analysis. The main objective of the stratification is to improve the quality of the National CPI. Uganda is demarcated into 5 CPI Statistical regions: Kampala (the capital city), Central2, Northern, Western and Eastern regions. The Regions are further divided into 10 Statistical subregions: Kampala, East Central, Eastern, Central 1, Central 2, South-Western, Mid-Northern, West-Nile, Mid-Western, and North East. The Uganda national CPI is a weighted average of 10 sub-indices or consumption basket indices. The CPI for the 10 consumption baskets covers all the five (5) regions. Consumer prices are collected from the capital city as one of the subregions, and from other 7 urban areas that represent the remaining 9 subregions. The monthly CPI is compiled for each of the seven (7) urban areas while the capital city is divided into 3 income groups, namely, Kampala High Income, Kampala Middle Income and Kampala Low Income. Similarly, the monthly CPI is compiled for each of the income groups. The above description defines the process of computing 10 sub-indices or 10 basket indices. The current CPI uses 2009/2010 financial year as the price, weight and index reference period.

2 Kampala, Uganda’s capital city, is in the Central region. However, for the CPI regions, Central excludes Kampala, which is already a CPI region.

The African Statistical Journal, Volume 20, February 2018

Vincent Musoke Nsubuga

The major source of data used for the derivation of the CPI weights are the results from the Household Consumption Module of the Uganda National Household Survey (UNHS) IV carried out from May 2009 to April 2010. Since the UNHS was not primarily designed for the computation of the CPI weights, additional data sources were used for the compilation of reliable weights. The additional data sources included the National Accounts data (for the final household consumption) and the results of the 2012 Tourist Expenditure and Motivation Survey (TEMS). A household budget minisurvey was also conducted to collect additional data. The study shows that the weights for each of the 10 consumption baskets changed. In addition, the weights within each of the consumption baskets changed significantly within the transport division. The extreme case was for Mbarara basket whose transport division changed sharply. Results show that the all-items CPI and the all-items annual inflation rate for each of the consumption baskets for both the published and study focus areas were not much different and in some cases, the results were the same at the significance level of dissemination. The official CPI for the transport division for the Mbarara basket and the CPI computed from the study for the same division are significantly different. Consequently, the corresponding official annual inflation for the transport division for the Mbarara basket compared with the inflation computed from study are also significantly different. This is due to relatively large expenditure on second-hand vehicles by the households of Mbarara basket and the resultant reallocation of weights. This implies that, given the increasing demand for disaggregated data, especially at lower levels of local administration, decisions on the compilation leads to different results arising from wrong choices. Some countries have formed regional economic blocs like the East African Community (EAC), Common Market for Eastern and Southern Africa (COMESA), and the Southern African Development Community (SADC) among others. These regional economic blocs compile Harmonized Consumer Price Indices (HCPI) and follow the domestic concept. To obtain meaningful subregional indices within a country, it is recommended that the domestic concept be restricted to the national level and expenditures within the country be allocated to the households within the subregion they reside. This implies using the domestic concept at the national level and following the national concept at the local level. An advantage of this approach is

Journal africain de statistiques, numĂŠro 20, fĂŠvrier 2018

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas

the ability to allocate household expenditures within the economy to the subregions where the households reside, instead of allocating them to the subregions where the transactions take place, as recommended by the Domestic concept. Thus, during the compilation of CPI, both Domestic and National concepts could be used at the same time but at different levels. The results demonstrate the essence of ensuring that basic prices data used for index computation are of good quality, as price changes have a significant impact on the resultant indices. 1.2 Determination of the 10 CPI consumption baskets The 10 consumption baskets are based on the urban Household Final Momentary Consumption Expenditure (HFMCE) of the entire economy. The Kampala region is treated differently and has the unique status of a subregion. Kampala’s HFMCE was divided into three (3) income-based CPI consumption baskets: Kampala High Income, Kampala Middle Income and Kampala Low Income baskets. The remaining seven (7) baskets are spread across the remaining nine (9) subregions. The seven subregions were selected in such a way that all the remaining four (4) regions are represented. One subregion was selected from the central region, while two subregions were selected from each of the remaining three regions. The urban household expenditure for each selected subregion became a CPI consumption basket. The urban household expenditure for the two remaining subregions were proportionally distributed to the selected subregions that are close geographically. Due to resource constraints, two (2) subregions could not be included in the CPI. The first subregion excluded had a much smaller expenditure compared to the rest. However, because the second subregion’s proximity to Kampala, it is assumed that much of its price movements are parallel to changes in Kampala. However, with additional resources, geographical expansion is possible. One urban center was picked from each of the selected seven (7) subregions. The selected urban area was the one with the largest population as measured by the 2002 Uganda Population and Housing Census. 1.3 Determination of income groups Kampala’s income groups were stratified. Based on the HFMCE, a proxy for the income groups, three income consumption basket groups were categorized as: High Income, Middle Income and Low Income baskets. The Kampala High Income basket is the 10th deciles3, i.e. the expenditure of the top 10 percent households for the Kampala subregion, which comprises 3 A decile is any of the nine values that divide the sorted data into ten equal parts, so that each part represents 1/10 of the sample or population.

The African Statistical Journal, Volume 20, February 2018

Vincent Musoke Nsubuga

the entire Kampala City Council Authority (KCCA). The Kampala Middle Income basket is the 7th-9th deciles, i.e. the expenditure of the top 30 percent households for the Kampala subregion, excluding the top 10 percent households for the Kampala High Income. The Kampala Low Income basket is the 1st-6th deciles, i.e. the expenditure of lowest 60 percent of the households for the Kampala subregion. 1.4 Computation of weights After determining the seven regional baskets and the three income group baskets, household final consumption expenditures for each of the 10 baskets were classified using Classification of Individual Consumption According to Purpose (COICOP). However, illegal and black-market expenditures such as prostitution and narcotics were excluded; so was life insurance, which is regarded as an investment. Also excluded, for practical reasons, were Games of Chance (Gambling) and Financial Intermediation Services Indirectly Measured (FISIM). The weights for the Uganda CPI basket were computed in three stages. The first stage was the computation of elementary aggregate weights and higher-level aggregate weights within each of the 10 consumption baskets. The second stage involved the computation of weights for each of the 10 baskets, and the final stage involved the computation of national weights for each of the elementary aggregates and higher-level aggregate weights. 1.4.1 Elementary aggregate weights within each consumption basket After classifying all household expenditures within the COICOP headings for each of the 10 baskets, the proportion of expenditures for each Elementary Aggregate (EA) and for each basket was computed and their total sum scaled to 1,000. The cut-off point for an EA for inclusion in CPI was set at 1 part per 1,000. In cases where the expenditure proportions of an EA fell below the threshold, the expenditure for such an EA was proportionally distributed across the remaining EAs within the same COICOP sub-class or class. In some cases, the expenditure for such an EA was combined with another EA to form a new EA, provided there was homogeneity of products within the relevant EAs. 1.4.2 Higher level aggregate weights for each consumption basket Higher-level weights are weights above the Elementary Aggregate Level. These were computed as the total sum of weights for EAs classified under that Higher Level Aggregate.

Journal africain de statistiques, numĂŠro 20, fĂŠvrier 2018

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas

1.4.3 Computation of consumption basket weights The second stage was to compute consumption basket (geographical area and population income) weights. The weight for each of the geographical areas and population income groups was computed as a proportion of the geographical area or income group household expenditure to the total expenditure for all the seven (7) geographical areas and the three (3) population income groups. 1.4.4 Computation of national weights The final stage was the computation of national weights. These are the combination of weights within the consumption baskets together with those between the consumption baskets. 1.5 Selection of locations for price collection After the determination of the 10 consumption baskets, i.e. 3 population income group baskets and 7 geographical area baskets, the next step was to select locations for regular collection of price of consumer goods. Uganda disseminates the National CPI, the CPI for each of the 3 population income groups and CPI for each of the 7 geographical areas. For Kampalaâ&#x20AC;&#x2122;s 3 population income groups, consumer prices are collected from the locations where representatives of each income group normally make purchases. Owing to limited resources, only one urban center is selected for each of the 7 geographical areas, and consumer prices from each of the selected urban center are collected, processed and used to compile the CPI for the subregion. The criteria for the 2002 population and housing census results were adopted for the selection of the urban centers as price collection zones. Thus, an urban area within the subregion with the highest population were selected. 1.6 Naming of the geographical area or subregion The name of the selected urban area for price collection, as explained above, becomes the name of the subregion or geographical area. 1.7 Selection of items After the determination of the Enumeration Areas, the next step was to select items within each Enumeration Area for each of the 10 consumption baskets. Four approaches were used to select the item lists. The first approach used the list of items available in the previous CPI before the general revision exercise. All items in the previous CPI that were still popular and representative on the market were selected and kept in the revised CPI. The second approach used the 2011 ICP list of items. From the ICP list, only representative items were selected and incorporated into the CPI baskets. The third approach selected a list of items that had significant expenditure

The African Statistical Journal, Volume 20, February 2018

Vincent Musoke Nsubuga

from the 2014 Mini Household Budget survey. The fourth approach selected items from the open market. In this case, the required representative list of items was selected from various outlets in consultation with the respondents. 1.8 Selection of outlets The next step was the selection of outlets. Three approaches were used to select outlets from which items are priced regularly. The first approach used the outlets available in the previous CPI. All the Outlets in the previous CPI that were still popular and representative were retained for use in the revised CPI. The second approach used the 2011 ICP list of outlets. From the ICP list, only representative outlets were selected and incorporated into the CPI outlets. The third approach entailed field visits to identify the types of outlets stocked with representative items for regular pricing. The nonrandom (purposive) approach is still used in selecting outlets. For example, outlets selected were those most frequented by the reference population. When an outlet closes permanently, it is replaced with another outlet with similar characteristics and proximity to the closed outlet. 1.9 Price determining characteristics After the selection of items and outlets, the price-determining characteristics, particularly for the newly introduced items, were identified and documented. This process was done in consultation with the respondents. 1.10 Collection of consumer prices Consumer prices for each of the 10 consumption baskets, the 7 geographical areas and 3 population income groups are collected from the 1st to the 15th day of each month. This implies that any price change after the 15th of the month is reflected in the CPI of the following month. Consumer prices for most of the CPI items are collected once a month, some twice a month, and others are collected with more than one-month interval. The frequency of price collection of any item depends on its price volatility. For instance, prices for fresh food crops such as fresh vegetables are collected twice a month; rental charges are collected once in a quarter; education charges for pre-primary, primary and secondary schools are collected once in an academic term, at the beginning of the term. Education fees for tertiary institutions are collected once in an academic semester, at the beginning of every semester. Prices for each basket are collected from the basketâ&#x20AC;&#x2122;s geographical area. However, there are cases where item prices are permanently missing in a geographical area and households of that very area make purchases of such items from another region. An example of such items is second-hand

Journal africain de statistiques, numĂŠro 20, fĂŠvrier 2018

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas

vehicles. The prices for such items are collected from a region where they are readily available. These prices are subsequently used to compute the index of the subregion where prices for such items are unavailable. In Uganda, CPI items might be available in the selected geographical areas, yet it is extremely difficult to collect prices there. Second-hand clothes fall in this category. In such cases, prices from other geographical areas are used for the compilation of the subregionâ&#x20AC;&#x2122;s CPI. Second-hand clothes are some of the hard-to-measure products in the Ugandan CPI. In such peculiar cases, the principle of comparing like with like is generally difficult to maintain over time. For second-hand clothing, proxy wholesale prices are used. All the prices are validated and any missing prices imputed before the CPI computation. 1.11 Computation of the CPI The computation of the CPI is carried out in three stages. The first stage involves the computation of Elementary Aggregate Indices (EAIs) for each of the 10 consumption baskets. Elementary aggregate indices for the CPI are calculated using the geometric mean (GM) method (the Jevons index). The second stage is the computation of Higher Level Indices for each of the 10 consumption baskets. EAs are the lowest level at which there are weights and below which no weights exist. The arrangement is hierarchical, starting with sub-class indices, class indices, group indices, division indices and the all items index. Higher-level indices are calculated as weighted averages of their components. The third and final stage is the computation of the National Index. The National indices at all levels are calculated directly as a weighted average of the 7 geographical area indices and the 3 population income group indices (a weighted average of the 10 consumption baskets). 1.12 Dissemination of the CPI CPI and the corresponding weights are disseminated at the national level and also by geographical area and population income groups. The CPI for any particular month is disseminated to the general public at 11.00am on the last working day of the month, through a press conference, electronically specifically on the UBOS website and using emails for specific data users. In addition, major stakeholders such as Bank of Uganda (Central Bank) and the Ministry of Finance, Planning and Economic Development get the results at the same time when national dissemination of the CPI to the media is taking place. The disseminated components of the national CPI and their derived inflation rates are: Headline Index which is a combined index for all items of the 10 consumption basket, the corresponding core and non-core indices,

The African Statistical Journal, Volume 20, February 2018

Vincent Musoke Nsubuga

service indices, energy, fuel & utilities indices, food crops and related items indices and other goods indices; and indices for the 12 COICOP divisions. Other disseminated components are the All Items indices for each of the 3 population income groups; and the All Items indices for each of the 7 geographical areas. Further analysis is still on-going to categorize the CPI basket between tradable and non-tradable so that the indices of the two categories is added to the disseminated components. The weights at any level are disseminated at 4 decimal places, index numbers at 2 decimal places, and inflation rates at one decimal place. Below are the inflation rates disseminated to the public. 1.13 Computation of the derived statistics Most users are interested in statistics derived from the computed index numbers, which are the rates of price change. The most used inflation numbers are monthly, annual and annual average inflation rates. The computation of these listed rates is described below. 1.13.1 Monthly inflation rates are calculated as the percentage change between the CPIs of the current and previous month. 1.13.2 Annual inflation rates are calculated as the percentage change between CPIs of the same month in the current and previous year. 1.13.3 Annual average inflation rates are calculated as the percentage change between the Annual Average CPI for the latest 12 consecutive months and the Annual Average CPI for the previous12 consecutive months. 2. FINDINGS OF THE STUDY

2.1 Introduction The study focuses on the monthly CPI for the period July 2010 to March 2016. Sections 1.9 â&#x20AC;&#x201C; 1.13 amplified the process of collecting consumer prices, computation and dissemination of the CPI. Consumer prices for each consumption CPI baskets are collected from geographical areas and used to compute subregional CPIs for those areas. The aggregation of resultant indices produces the national CPI. However, there are special cases where households consume certain products purchased from other geographical areas. This study used second-hand vehicles as an example. In Uganda,

Journal africain de statistiques, numĂŠro 20, fĂŠvrier 2018

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas

household net final consumption for second-hand vehicles across regions is significant and on the increase. However, Kampala is the only market from which households buy such products and where price collection could be done regularly and in keeping with the principle of comparing like with like. The study examined the difference between the published CPI by using consumer prices for second-hand vehicles in Kampala to compute the CPI for other regions where such prices do not exist. The household net final expenditure on second-hand vehicles for other regions are allocated to Kampala and used to investigate the differences between the published CPI for each of the published 10 consumption baskets (i.e. three population income group indices and seven geographical area indices). The study considered the three dimensions: the weights, indices and inflation rates. 2.1

Consumption basket weights and weights for the Transport Division The consumption expenditure for second-hand vehicles for the seven geographical areas was proportionally allocated to the three Kampala population groups. This allocation had an impact on the weights in two significant ways. First, the weight for each of the 10-consumption baskets changed. Second, the weights for elementary aggregates within each of the 10 consumption baskets changed. As a result, the entire national weights system changed, more so for the transport division. 2.1.1 Consumption baskets weights The consumption baskets4 are: Kampala High Income, Kampala Middle Income, Kampala Low Income, Mbarara, Masaka, Mbale, Jinja, Gulu, Fort Portal and Arua. The consumption basket with the biggest increase in weight is Kampala High Income; the one with the biggest drop is Mbarara basket. The weight for the Kampala High Income basket increased from 31.397 parts per 100 to 31.8632 parts per 100—an increase of 1.5 percent. The weight for the Mbarara basket dropped from 9.6752 parts per 100 to 9.4189 parts per 100—a drop of 2.6 percent. Figure 1 shows changes in weights for each of the 10 consumption baskets. Table 1 provides more details on the changes in weights for each of the 10 baskets.

4 Representation as in section 1.6.

The African Statistical Journal, Volume 20, February 2018

Vincent Musoke Nsubuga

Figure 1: All Items CPI Weights for the 10 Consumption Baskets 31.86

Kampala High Income

31.40 15.73

Kampala Middle Income

15.66 10.05

Kampala Low Income

10.01 9.42

Consumption Basket

Mbarara

9.68 9.37

Masaka

9.47 5.63

Mbale

5.68 5.61

Jinja

5.65 5.00

Gulu

5.04 3.72

Fortportal

3.76 3.62

Arua

3.66 0

5 Under Study

15 Published

Weights per 100

Journal africain de statistiques, numĂŠro 20, fĂŠvrier 2018

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas

2.1.2 Weights for the transport division. The weights for the transport division changed substantially within the 8 consumption baskets where the Kampala Low Income and Kampala Middle Income baskets registered minimal changes in the transport division. The weight for the transport division for Mbarara dropped drastically by 21.6 percent, followed by that of Fort Portal which dropped by 11.56 percent, as shown in Figure 2. More details on the changes are provided in Table 2. Figure 2: Transport Division Weights for the 10 Consumption Baskets 6.68

Kampala High Income

6.25 1.79

Kampala Middle Income

1.74 1.06

Kampala Low Income

1.03 0.97

Mbarara

Consumption Basket

1.24 0.93

Masaka

1.04 0.46

Mbale

0.51 0.58

Jinja

0.63 0.47

Gulu

0.52 0.36

Fortportal

0.41 0.35

Arua

0.39 0

1 Under Study

3 Published

The African Statistical Journal, Volume 20, February 2018

Weights per 100

Vincent Musoke Nsubuga

2.2.

Annual inflation rates for the “All Items” baskets and the transport division A comparison of the 10 consumption baskets was made at the “All Items” level—between the published CPI and the CPI under study—for the 10 consumption baskets. The study included an in-depth analysis of the annual inflation rates (published and under study) of the transport divisions for each of the 10 consumption baskets. 2.2.1 Annual inflation rates at the “All Items” levels Results show that the “All Items” Annual Inflation for both the CPI (published and under study) for each of the 10 consumption baskets are about the same. More details are available in Tables 3 and 4. 2.2.2 Annual inflation rate for the transport division Results for the Mbarara transport division’s consumption basket show a significant difference between the published annual inflation rate and the one under study (see Table 5). This variation is attributable to a drop in the transport division weight for the Mbarara basket. The weight dropped from 12.41 parts per 1,000 to 9.74 parts per 1,000 representing 21.6 percent (see Table 2). This sends a strong signal to policymakers who underestimate the changes in prices in that region for the transport division by excluding such items because regular price collection is impossible. For the remaining 8 consumption baskets, the transport divisions’ annual inflation rates (published and under study) are almost the same. It was also found that, in some cases, the inflation numbers are the same at the significant level of dissemination, i.e. one decimal place. More details are available in Tables 5 and 6. 2.3 CPI for “All Items” and for the transport division Tables 7, 8, 9 and 10 show the all items CPI and indices for the transport divisions for each of the 10 consumption baskets. 3.

CONCLUSIONS AND RECOMMENDATIONS

3.1 Conclusion Although the study focused on one expenditure item, it is possible that other items fall under the same category. Given the increasing demand for data at disaggregated level, especially at lower levels of local administration, the choice taken to compile the CPI leads to different results. It is important to understand the results and implications to decision-making,

Journal africain de statistiques, numéro 20, février 2018

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas

policy formulation and design. Removing expenditures from a subregion because price collection is untenable can result in wrong policy at the local level—policy built on distorted household expenditure patterns, and hence the CPI. The best option, therefore, is to retain expenditures where they were incurred and use the relevant average prices from different regions or subregion where price collection on a like with like comparison is possible. In addition, the principle of proportional reallocation of expenditure when certain products cannot be priced within a subregion, should be discouraged unless reliable prices cannot be regularly obtained countrywide. Some countries have gotten together to form regional economic blocs. Notable examples are: the Common Market for Eastern and Southern Africa (COMESA), East African Community (EAC), South Africa Development Community (SADC), Economic Community of West African States (ECOWAS), Economic Central African States (ECCAS), Arab Maghreb Union (AMU), Intergovernmental Authority on Development (IGAD) and Community of Sahel-Saharan States (CEN-SAD). COMESA, EAC and SADC compile the Harmonized Consumer Price Indices (HCPI) and follow the domestic concept where possible. To obtain meaningful subregional indices within a country, it is recommended that the domestic concept be restricted to the national level and the expenditure within the country be allocated to the households within the subregion they reside. This implies that only consumption expenditures within the given economy should be considered and that the basis for the allocation of consumption to households should be their usual residence, and not the subregions where the expenditures occurred. Despite changes in the weights between and within the 10 consumption baskets, the annual inflation rate for the 10 consumption baskets at the “All Items” level for both the published and under study cases are almost the same, and in some cases the same at the significant level of dissemination. The differences are observed in the transport division and the Mbarara basket. These results demonstrate the need to ensure that all prices entered for the computation of the CPI are of very high quality, since price movements have a much bigger impact on the inflation to be published and disseminated, compared to the weights.

The African Statistical Journal, Volume 20, February 2018

Vincent Musoke Nsubuga

Table 1: All Items CPI Weights for the 10 Consumption Baskets Consumption Basket

Published

Under Study

% Change

Kampala High Income

31.3976

31.8632

1.5

Kampala Middle Income

15.6609

15.7268

0.4

Kampala Low Income

10.0069

10.0488

0.4

Mbarara

9.6752

9.4189

-2.6

Masaka

9.4693

9.3690

-1.1

Mbale

5.6799

5.6298

-0.9

Jinja

5.6522

5.6094

-0.8

Gulu

5.0414

4.9966

-0.9

Fortportal

3.7608

3.7177

-1.1

Arua

3.6557

3.6197

-1.0

Source: UBOS and author

Table 2: Transport Division Weights for 10 Consumption Baskets Consumption Basket

Published

Under Study

% Change

Kampala High Income

6.2519

6.6844

6.9

Kampala Middle Income

1.7423

1.7900

2.7

Kampala Low Income

1.0298

1.0598

2.9

Mbarara

1.2410

0.9735

-21.6

Masaka

1.0389

0.9275

-10.7

Mbale

0.5145

0.4576

-11.1

Jinja

0.6317

0.5822

-7.8

Gulu

0.5214

0.4706

-9.7

Fortportal

0.4131

0.3656

-11.5

Arua

0.3945

0.3542

-10.2

Journal africain de statistiques, numéro 20, février 2018

The African Statistical Journal, Volume 20, February 2018

19.7

21.4

20.8

20.5

19.7

21.4

19.7

16.6

15.6

12.1

6.9

9.4

6.1

6.6

5.8

Sep-11

Oct-11

Nov-11

Dec-11

Jan-12

Feb-12

Mar-12

Apr-12

May-12

Jun-12

Jul-12

Aug-12

Sep-12

Oct-12

Nov-12

5.3

14.8

Aug-11

Dec-12

18.1

Published

Jul-11

Month Ending

5.2

5.8

6.6

6.0

9.4

6.9

12.1

15.4

16.5

19.6

21.3

19.5

20.4

20.7

21.3

19.7

14.8

18.0

Under Study

Kampala High Income

3.5

3.9

3.3

4.1

9.6

8.3

16.0

19.7

20.9

26.0

26.9

25.9

25.6

26.2

26.4

24.4

17.8

20.3

Published

3.4

3.9

3.3

4.1

9.6

8.3

16.0

19.7

20.9

26.0

26.9

25.9

25.6

26.2

26.3

24.4

17.8

20.3

Under Study

Kampala Middle Income

-0.3

-0.2

-1.2

0.2

9.3

10.3

16.1

22.7

25.1

25.9

27.7

25.3

29.1

29.2

29.4

27.7

20.5

17.6

Published

-0.4

-0.2

-1.2

0.2

9.3

10.3

16.1

22.7

25.1

25.9

27.6

25.3

29.1

29.3

27.7

20.5

17.6

Under Study

Kampala Low Income

7.5

6.0

6.8

6.4

12.6

15.5

16.0

18.8

17.5

21.3

20.3

21.2

22.6

20.0

19.0

12.9

14.1

7.1

5.6

6.4

5.9

12.2

15.0

15.5

18.5

19.3

17.8

21.8

20.7

21.6

23.2

20.5

19.2

12.9

14.2

Under Study

Mbarara Published

Table 3: Annual Inflation Rates – All Items – Published and Under Study

6.2

8.2

7.4

6.7

9.7

12.4

15.2

18.1

19.7

19.9

25.8

23.5

22.6

23.3

24.3

21.7

19.9

15.7

Published

6.3

8.3

7.5

6.7

9.7

12.4

15.2

18.3

19.8

20.0

26.0

23.7

22.7

23.4

24.4

21.8

20.0

15.7

Under Study

Masaka

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas

6.4

6.6

6.7

5.8

5.9

4.0

2.9

3.0

Nov-13

Dec-13

Jan-14

Feb-14

Mar-14

Apr-14

May-14

Jun-14

Jul-14

7.2

Jul-13

5.4

7.8

Jun-13

Oct-13

6.0

May-13

6.9

4.5

Apr-13

Sep-13

3.8

Mar-13

6.9

2.2

Feb-13

Aug-13

3.6

Published

Jan-13

Month Ending

2.9

4.0

3.9

5.9

5.8

6.7

6.5

6.4

5.3

6.8

6.7

7.0

7.7

5.9

4.4

3.8

2.2

3.5

Under Study

Kampala High Income

1.8

1.9

3.6

4.3

4.8

5.1

5.5

6.5

7.6

8.6

9.4

7.7

6.1

4.0

3.2

1.3

1.2

2.6

Published

1.8

1.9

3.6

4.3

4.7

5.1

5.4

6.5

7.6

8.6

9.4

9.3

7.6

6.1

4.0

3.2

1.3

1.2

2.6

Under Study

Kampala Middle Income

0.6

3.1

4.4

6.3

6.8

7.3

8.1

8.5

9.3

11.4

11.1

8.5

9.5

4.3

2.9

1.1

0.7

-0.1

0.5

Published

0.6

3.1

4.4

6.3

6.7

7.2

8.1

8.5

9.3

11.3

11.1

8.5

9.5

4.3

2.9

1.1

0.7

-0.1

0.4

Under Study

Kampala Low Income

1.3

2.1

3.3

3.1

0.4

1.6

3.2

4.7

7.1

8.4

8.0

5.2

5.4

3.4

4.7

6.8

7.3

8.2

Published

1.3

2.2

2.1

3.5

3.2

0.4

1.6

3.3

4.9

7.4

8.7

8.3

5.4

5.6

3.5

4.3

6.5

7.0

8.0

Under Study

Mbarara

0.9

1.5

1.3

1.5

2.9

3.1

3.5

3.6

5.3

8.7

8.5

8.1

6.7

5.0

4.7

6.6

4.0

5.8

Published

0.9

1.6

1.3

1.6

2.9

3.1

3.5

3.6

5.4

8.8

8.7

8.2

6.8

5.1

4.7

6.7

4.1

5.9

Under Study

Masaka

Vincent Musoke Nsubuga

Journal africain de statistiques, numéro 20, février 2018

The African Statistical Journal, Volume 20, February 2018

1.5

3.4

3.0

3.9

5.8

4.8

5.2

6.7

6.3

7.0

7.4

8.3

6.5

Dec-14

Jan-15

Feb-15

Mar-15

Apr-15

May-15

Jun-15

Jul-15

Aug-15

Sep-15

Oct-15

Nov-15

Dec-15

Jan-16

6.2

2.9

Nov-14

Mar-16

3.3

Oct-14

6.9

3.6

Sep-14

Feb-16

4.2

Published

Aug-14

Month Ending

6.4

7.0

6.5

8.5

7.4

7.1

6.4

6.8

6.7

5.2

4.8

5.8

3.9

3.0

3.4

1.5

3.0

3.4

3.6

4.2

Under Study

Kampala High Income

4.9

5.9

6.5

8.8

8.2

7.6

7.2

6.8

6.9

6.8

7.0

6.4

5.7

5.2

4.9

2.4

2.3

2.1

0.9

1.3

Published

4.9

6.0

6.5

8.8

8.2

7.6

7.2

6.8

6.9

6.8

7.0

6.4

5.7

5.2

4.9

2.4

2.3

2.1

0.9

1.3

Under Study

Kampala Middle Income

8.1

7.1

9.1

9.8

9.4

8.8

6.7

4.7

6.0

5.3

5.0

3.6

2.5

2.2

2.1

1.3

0.8

0.1

0.4

0.0

Published

8.1

7.1

9.1

9.8

9.4

8.8

6.7

4.8

6.0

5.3

5.0

3.6

2.5

2.2

2.1

1.3

0.8

0.1

0.4

0.0

Under Study

Kampala Low Income

5.9

7.8

8.7

8.0

7.7

7.8

5.8

3.3

2.6

2.8

2.9

3.5

1.7

2.2

1.7

1.3

0.3

-0.7

-0.1

-1.1

Published

5.7

7.8

8.8

7.9

7.7

7.8

5.8

3.3

2.6

2.8

2.9

3.5

1.7

2.2

1.6

1.4

0.2

-0.8

-0.1

-1.2

Under Study

Mbarara

9.2

10.6

10.2

10.6

12.1

13.0

11.0

7.3

6.5

6.6

6.3

4.9

3.2

2.8

3.6

4.2

0.1

-1.9

-2.2

-0.2

Published

9.1

10.5

10.2

10.5

12.1

13.0

11.0

7.3

6.4

6.5

6.3

4.9

3.2

2.8

3.6

4.3

0.0

-2.0

-2.3

-0.2

Under Study

Masaka

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas

Published

16.7

18.6

22.1

22.4

23.2

25.4

27.1

18.4

18.2

17.3

16.7

15.0

9.1

4.0

4.9

3.7

Month Ending

Jul-11

Aug-11

Sep-11

Oct-11

Nov-11

Dec-11

Jan-12

Feb-12

Mar-12

Apr-12

May-12

Jun-12

Jul-12

Aug-12

Sep-12

Oct-12

Nov-12

Dec-12

3.8

4.9

4.0

9.1

15.1

16.8

17.4

18.3

18.5

27.2

25.5

23.3

22.5

22.2

18.6

16.7

Under Study

Mbale

2.5

2.8

4.2

4.1

8.4

11.7

13.1

14.6

16.0

17.2

24.8

24.9

25.4

26.0

24.4

25.1

22.0

20.6

Published

Jinja

2.6

2.9

4.2

4.1

8.4

11.7

13.1

14.6

16.0

17.3

24.9

25.1

25.5

26.1

24.5

25.1

22.1

20.6

Under Study

6.0

3.7

2.3

8.5

9.6

12.8

18.7

19.3

20.9

22.8

28.3

25.4

25.8

29.5

30.1

26.3

24.7

19.4

Published

Gulu

6.1

3.7

2.4

8.5

9.6

12.8

18.8

19.5

21.0

23.0

28.5

25.5

25.9

29.7

30.3

26.4

24.8

19.4

Under Study

Arua

3.8

3.3

4.0

5.4

11.3

15.1

23.6

21.8

22.4

24.9

28.4

26.3

32.9

35.7

29.9

29.6

23.1

20.3

Published

Table 4: Annual Inflation Rates – All Items – Published and Under Study

3.9

3.3

4.0

5.4

11.3

15.1

23.8

22.0

22.6

25.0

28.5

26.5

33.1

35.9

30.0

29.6

23.1

20.2

Under Study

5.3

6.3

4.7

5.1

13.9

15.5

16.4

20.3

21.2

20.4

22.9

25.1

27.1

28.1

30.1

28.4

19.6

16.4

Published

5.4

6.3

4.7

5.1

14.0

15.6

16.5

20.5

21.3

20.6

23.0

25.3

27.3

28.3

30.3

28.5

19.6

16.4

Under Study

Fort Portal

Vincent Musoke Nsubuga

Journal africain de statistiques, numéro 20, février 2018

Published

2.0

-1.4

0.0

-0.2

0.5

2.6

3.0

4.7

5.7

5.6

3.9

3.5

3.2

3.9

5.1

4.2

2.9

3.1

Jan-13

Feb-13

Mar-13

Apr-13

May-13

Jun-13

Jul-13

Aug-13

Sep-13

Oct-13

Nov-13

The African Statistical Journal, Volume 20, February 2018

Dec-13

Jan-14

Feb-14

Mar-14

Apr-14

May-14

Jun-14

Jul-14

Aug-14

3.1

2.9

4.3

5.1

3.9

3.2

3.5

3.9

5.6

5.8

4.8

3.0

2.7

0.5

-0.2

0.0

-1.4

2.0

Under Study

Mbale

Month Ending

1.8

3.6

2.3

2.7

3.3

3.9

2.6

4.8

4.5

5.2

6.9

6.0

6.2

2.9

2.8

3.5

2.6

3.7

2.7

1.1

Published

Jinja

1.8

3.6

2.3

2.8

3.3

3.9

2.7

4.9

4.5

5.3

6.9

6.1

6.4

3.0

2.9

3.5

2.6

3.8

2.7

1.1

Under Study

1.6

3.9

2.1

3.2

7.1

9.5

10.1

8.5

5.9

7.4

5.6

5.7

4.2

3.8

3.5

1.3

0.4

-0.9

4.8

Published

Gulu

1.6

3.9

2.2

3.2

7.2

9.5

10.2

8.6

5.9

7.5

5.7

5.8

4.3

3.9

3.6

1.3

0.5

-0.9

4.9

Under Study

4.9

7.0

4.1

2.9

4.9

4.3

2.4

2.3

2.1

2.2

2.7

1.8

0.6

-2.5

-2.8

0.1

0.8

3.0

4.4

Published

Arua

4.9

7.0

4.1

2.9

5.0

4.3

2.4

2.2

2.3

2.8

1.9

0.7

-2.4

-2.7

0.1

0.9

3.1

4.5

Under Study

3.3

5.0

5.4

7.7

8.1

6.7

5.4

4.8

3.9

2.5

5.3

3.9

1.9

3.0

0.3

0.8

1.9

2.1

4.3

Published

3.3

5.1

5.5

7.8

8.2

6.7

5.5

4.9

3.9

2.6

5.4

4.0

2.0

3.1

0.3

0.8

1.9

2.1

4.4

Under Study

Fort Portal

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas

Published

2.0

2.2

2.6

2.4

2.7

3.8

3.0

3.8

3.7

1.9

0.7

2.3

3.4

5.0

5.5

6.1

5.3

3.8

4.0

Sep-14

Oct-14

Nov-14

Dec-14

Jan-15

Feb-15

Mar-15

Apr-15

May-15

Jun-15

Jul-15

Aug-15

Sep-15

Oct-15

Nov-15

Dec-15

Jan-16

Feb-16

Mar-16

3.8

3.7

5.3

5.9

5.4

5.0

3.4

2.2

0.6

1.8

3.7

3.8

3.0

3.8

2.6

2.5

2.6

2.2

2.0

Under Study

Mbale

Month Ending

7.0

9.1

9.3

7.8

7.9

6.8

5.2

5.0

5.5

4.6

5.8

2.3

4.0

2.7

1.8

2.5

1.1

2.1

Published

Jinja

6.8

7.0

9.1

9.2

7.8

6.7

5.1

5.0

5.5

4.6

5.8

2.2

4.0

2.7

1.9

2.5

1.0

2.1

Under Study

4.1

6.2

5.3

7.2

7.3

5.4

5.0

4.8

3.3

4.5

2.1

0.7

1.1

-1.3

0.7

1.8

2.7

2.1

1.7

Published

Gulu

3.9

6.2

5.3

7.1

7.2

5.3

5.0

4.8

3.2

4.5

2.1

0.7

1.1

-1.4

0.6

1.8

2.6

2.1

1.7

Under Study

8.0

6.5

7.8

7.4

8.6

6.3

6.7

4.4

3.9

4.7

5.5

2.9

3.2

2.6

3.3

4.7

3.6

3.0

2.8

Published

Arua

8.0

6.5

7.9

7.4

8.6

6.2

6.6

4.3

3.8

4.6

5.4

2.8

3.1

2.5

3.2

4.7

3.4

2.9

2.8

Under Study

4.5

6.6

6.5

7.0

6.3

4.2

4.7

5.7

7.0

2.8

3.2

2.9

2.7

2.3

3.7

3.6

2.3

3.6

Published

4.3

6.6

6.4

6.9

6.3

4.1

4.6

5.7

7.0

2.7

3.2

2.9

2.6

2.3

3.8

3.5

2.3

3.6

Under Study

Fort Portal

Vincent Musoke Nsubuga

Journal africain de statistiques, numéro 20, février 2018

The African Statistical Journal, Volume 20, February 2018

14.5

19.0

14.9

12.6

10.4

4.8

4.2

3.8

3.9

3.5

-0.6

1.1

4.4

Nov-11

Dec-11

Jan-12

Feb-12

Mar-12

Apr-12

May-12

Jun-12

Jul-12

Aug-12

Sep-12

Oct-12

Nov-12

0.3

15.6

Oct-11

Jan-13

18.1

Sep-11

1.5

14.6

Aug-11

Dec-12

14.3

Published

Jul-11

Month Ending

0.1

1.2

4.2

1.1

-0.4

3.8

4.2

4.3

5.0

10.4

12.5

14.5

18.7

14.4

15.6

17.9

14.7

14.4

Under study

Kampala High Income

1.6

3.0

5.6

4.1

1.6

6.5

6.3

5.9

6.1

5.8

13.5

15.0

14.9

18.7

10.7

10.4

14.3

10.2

9.2

Published

1.4

2.8

5.4

4.0

1.6

6.6

6.4

6.0

6.1

5.9

13.4

14.9

14.7

18.6

10.8

10.5

14.3

10.3

9.4

Under study

Kampala Middle Income

-0.9

0.2

5.2

2.3

-0.2

5.9

7.9

8.8

9.0

7.6

15.6

17.8

20.1

25.4

14.6

16.3

19.7

13.8

11.3

Published

-0.9

0.1

5.0

2.3

-0.1

5.9

8.0

8.9

7.7

15.5

17.6

19.8

25.1

14.6

16.3

19.6

13.8

11.5

Under study

Kampala Low Income

Table 5: Transport Annual Inflation Rates â&#x20AC;&#x201C; Published and Under Study

5.7

5.4

4.8

6.1

5.1

6.0

16.2

12.6

8.7

11.7

10.1

12.9

13.7

14.4

14.3

12.3

14.6

17.1

12.1

Published

3.1

1.4

0.2

1.6

0.1

0.0

11.9

6.8

3.3

13.7

11.1

14.4

15.4

17.0

17.3

14.3

15.4

18.9

12.7

Under study

Mbarara

-0.7

1.0

1.4

3.2

2.9

6.6

6.8

8.0

7.9

4.5

6.4

18.7

24.9

21.8

18.8

12.4

20.2

18.6

17.2

Published

-0.5

1.5

3.4

2.9

6.4

6.5

7.7

8.3

4.0

5.9

19.8

27.1

22.8

19.6

12.1

20.9

19.0

17.4

Under study

Masaka

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas

2.0

3.6

4.0

2.3

1.8

1.4

-2.6

-2.1

Dec-13

Jan-14

Feb-14

Mar-14

Apr-14

May-14

Jun-14

Jul-14

5.5

Aug-13

2.7

7.2

Jul-13

Nov-13

6.9

Jun-13

4.9

4.8

May-13

Oct-13

6.8

Apr-13

5.9

3.4

Mar-13

Sep-13

1.1

Published

Feb-13

Month Ending

-2.0

-2.5

1.3

1.7

2.2

3.9

3.6

2.0

2.6

4.5

5.4

4.8

6.5

6.1

4.4

6.4

3.1

1.0

Under study

Kampala High Income

-2.8

-2.9

2.9

3.2

4.2

6.7

6.1

3.7

8.0

8.1

9.1

8.7

10.9

10.8

6.1

7.9

2.1

-0.5

Published

-2.7

-2.8

2.8

3.2

4.1

6.6

6.0

3.7

7.8

8.9

8.3

10.5

10.4

5.9

7.7

2.0

-0.5

Under study

Kampala Middle Income

-3.2

-4.0

2.2

2.1

3.2

7.0

4.6

2.8

6.9

7.9

8.6

8.1

10.3

10.4

5.5

8.5

2.4

-1.2

Published

-3.1

-3.9

2.2

2.0

3.1

6.9

4.6

2.8

6.7

7.6

8.3

7.7

9.9

10.0

5.3

8.3

2.3

-1.2

Under study

Kampala Low Income

1.1

0.9

0.7

0.8

0.6

-0.1

-1.4

-0.9

-0.1

-0.2

1.4

0.5

-4.0

-0.4

-0.6

2.5

2.8

4.5

Published

1.3

1.6

1.2

1.5

1.4

0.3

-2.1

-0.9

-0.1

-0.2

2.1

1.6

-4.2

0.5

-0.5

-1.9

-1.0

1.2

Under study

Mbarara

0.2

-1.5

-0.4

-0.8

-0.2

-0.3

0.2

-0.8

-0.7

-0.4

0.8

1.4

1.1

2.5

-1.2

3.5

3.4

0.9

Published

0.1

-1.7

-0.5

-0.9

-0.3

-0.5

-0.1

-1.1

-0.8

-0.3

1.0

2.2

1.7

3.4

-1.3

3.8

3.9

1.1

Under study

Masaka

Vincent Musoke Nsubuga

Journal africain de statistiques, numéro 20, février 2018

The African Statistical Journal, Volume 20, February 2018

-3.5

0.6

-0.2

-1.0

-1.3

-0.2

1.8

3.6

4.8

5.6

5.8

5.2

8.0

5.7

Dec-14

Jan-15

Feb-15

Mar-15

Apr-15

May-15

Jun-15

Jul-15

Aug-15

Sep-15

Oct-15

Nov-15

Dec-15

Jan-16

8.0

0.0

Nov-14

Mar-16

-0.6

Oct-14

6.7

-1.6

Sep-14

Feb-16

-0.9

Published

Aug-14

Month Ending

9.1

7.1

6.2

9.0

5.7

6.4

6.1

5.3

3.9

2.2

0.1

-1.0

-0.7

0.2

1.0

-3.6

0.4

-0.2

-1.4

-0.6

Under study

Kampala High Income

5.1

7.4

1.3

2.5

1.7

1.6

1.8

1.6

0.5

0.6

0.1

-1.3

-1.0

-4.4

-3.4

-2.8

-3.2

-2.9

Published

5.6

7.6

1.6

3.1

2.1

2.0

2.1

1.9

0.7

0.8

0.2

-1.2

-1.1

-0.9

-0.8

-4.4

-3.2

-2.6

-3.1

-2.8

Under study

Kampala Middle Income

5.8

4.3

2.7

4.7

3.6

3.0

2.3

1.4

0.8

-0.2

-1.7

-1.3

-0.9

0.7

-3.6

-2.5

-2.4

-1.9

Published

6.2

4.6

2.9

5.2

3.9

3.3

2.6

1.6

0.9

-0.1

-1.6

-1.2

-0.8

0.8

-3.6

-2.3

-1.8

Under study

Kampala Low Income

3.7

3.5

5.4

4.4

3.3

3.7

2.7

1.8

0.7

0.4

-0.7

-0.8

0.8

0.9

0.3

1.7

1.0

0.5

1.0

Published

0.6

2.4

5.1

2.0

2.3

1.4

0.6

-0.2

0.1

-1.7

-1.8

0.1

0.3

0.8

0.9

0.5

0.3

0.9

Under study

Mbarara

9.4

10.1

13.0

15.7

15.2

11.6

11.5

10.3

10.0

9.5

6.5

7.7

6.9

5.9

5.0

3.4

3.3

2.2

0.4

-0.2

Published

7.9

9.8

13.2

14.9

15.5

11.3

11.4

10.2

9.7

6.8

8.1

7.2

5.9

4.8

4.2

3.0

1.8

0.3

-0.5

Under study

Masaka

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas

Published

14.9

15.4

14.6

12.6

13.5

15.5

16.3

15.7

15.1

6.6

7.7

8.6

5.1

4.0

7.0

6.1

4.6

5.2

Month Ending

Jul-11

Aug-11

Sep-11

Oct-11

Nov-11

Dec-11

Jan-12

Feb-12

Mar-12

Apr-12

May-12

Jun-12

Jul-12

Aug-12

Sep-12

Oct-12

Nov-12

Dec-12

6.2

5.1

6.6

7.4

3.5

4.6

8.5

8.2

6.4

15.8

16.4

17.3

15.7

13.7

12.4

14.5

15.5

14.8

Under study

Mbale

6.2

2.0

3.0

2.8

3.6

5.7

7.2

9.0

13.2

17.4

18.7

18.9

24.9

22.7

19.6

25.3

25.8

22.3

Published

Jinja

7.0

2.2

3.2

2.8

3.3

5.5

6.9

9.3

13.5

17.9

19.4

19.8

25.7

23.5

19.9

26.0

26.6

22.7

Under study

4.4

5.4

9.9

12.7

9.6

11.5

11.8

12.0

7.5

8.9

15.0

15.8

19.0

20.9

15.1

19.9

20.2

17.8

Published

Gulu

5.2

5.8

10.8

13.7

9.8

11.7

12.0

12.8

7.4

8.8

15.6

16.7

19.6

21.8

15.2

20.4

20.8

18.0

Under study

4.8

1.6

2.9

2.4

6.8

8.2

7.3

8.1

9.8

14.1

19.5

18.0

20.7

21.4

20.8

22.7

21.4

19.4

Published

Arua

Table 6: Transport Division – Annual Inflation Rates – Published and Under Study

6.0

1.7

3.3

2.3

6.5

8.0

6.8

8.4

9.8

14.4

20.0

18.9

20.9

22.0

20.8

22.7

20.9

18.5

Under study

1.7

4.0

2.6

0.1

3.8

12.2

14.0

19.4

18.9

22.1

21.6

25.3

25.9

22.4

26.0

28.4

23.5

13.9

Published

2.3

4.4

2.7

-0.2

3.2

12.5

14.4

21.3

20.2

23.7

23.0

27.6

27.5

23.8

27.5

30.1

24.6

13.7

Under study

Fort Portal

Vincent Musoke Nsubuga

Journal africain de statistiques, numéro 20, février 2018

Published

4.4

4.7

6.4

5.9

5.6

6.3

3.5

2.5

0.5

-0.7

-0.4

-1.6

-1.2

-2.6

-2.7

-2.4

-2.1

0.4

1.6

Jan-13

Feb-13

Mar-13

Apr-13

May-13

Jun-13

Jul-13

Aug-13

Sep-13

Oct-13

Nov-13

The African Statistical Journal, Volume 20, February 2018

Dec-13

Jan-14

Feb-14

Mar-14

Apr-14

May-14

Jun-14

Jul-14

Aug-14

1.5

0.3

-2.3

-2.8

-3.0

-2.9

-1.5

-2.1

-0.7

-0.5

-0.7

0.7

3.6

4.4

7.6

6.4

6.5

7.2

5.3

5.2

Under study

Mbale

Month Ending

1.0

0.6

2.5

2.3

3.8

1.1

2.3

1.5

-2.3

3.6

5.7

5.3

4.5

6.0

6.1

5.7

4.0

7.3

6.4

6.0

Published

Jinja

0.9

0.6

2.7

2.4

4.0

1.1

2.3

1.4

-2.6

3.9

6.2

5.8

5.3

6.8

7.0

6.3

4.4

8.0

7.1

6.8

Under study

3.3

3.0

0.9

-1.1

-1.7

0.2

-2.8

4.1

0.5

-0.2

0.9

1.5

2.2

1.6

1.9

2.4

7.1

6.2

8.2

2.4

Published

Gulu

3.3

3.2

1.0

-1.2

-1.9

0.2

-3.2

4.1

0.3

-0.2

1.1

1.8

3.0

2.2

2.7

2.8

7.7

6.9

9.1

3.0

Under study

0.6

-0.5

0.4

0.3

1.8

3.1

3.6

2.7

4.1

4.8

5.0

3.5

4.1

5.9

3.4

2.7

4.2

Published

Arua

0.1

-0.7

0.4

0.3

1.9

3.4

3.9

3.0

3.3

5.0

5.7

6.1

5.0

5.5

7.7

7.0

7.4

4.6

4.0

5.5

Under study

2.2

0.6

0.0

2.7

2.3

2.4

0.2

1.2

2.8

2.5

0.3

1.6

0.6

0.1

1.8

0.9

1.4

2.8

1.6

5.7

Published

2.2

0.6

-0.1

3.0

2.5

2.6

0.0

1.0

2.9

2.8

0.4

2.0

1.3

0.6

2.7

1.1

1.5

3.2

1.9

6.6

Under study

Fort Portal

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas

Published

1.5

2.1

1.9

-0.1

1.7

0.4

0.0

-0.7

0.2

0.9

1.7

4.5

5.1

5.7

6.4

4.3

5.8

6.5

Sep-14

Oct-14

Nov-14

Dec-14

Jan-15

Feb-15

Mar-15

Apr-15

May-15

Jun-15

Jul-15

Aug-15

Sep-15

Oct-15

Nov-15

Dec-15

Jan-16

Feb-16

Mar-16

4.5

5.1

3.6

4.6

4.8

4.7

4.3

3.6

1.0

0.2

-0.3

-1.2

-0.5

-0.2

1.1

0.3

1.5

1.6

1.4

Under study

Mbale

Month Ending

4.9

4.5

5.7

7.0

2.8

3.0

2.4

3.0

2.3

-0.5

-1.0

-1.4

-2.0

-0.9

1.3

-1.0

1.6

1.2

0.9

Published

Jinja

3.3

3.8

5.0

5.6

1.9

2.1

1.6

2.3

1.8

-1.1

-1.4

-1.8

-2.5

-1.3

0.9

-0.8

1.2

0.8

Under study

3.6

2.6

2.3

2.2

1.6

2.1

1.2

0.3

-1.1

-1.5

0.0

0.4

-0.1

2.8

2.4

3.0

3.8

3.4

1.5

Published

Gulu

1.6

1.7

1.5

0.4

0.6

0.9

0.1

-0.8

-1.9

-2.4

-0.4

0.1

-0.6

2.5

2.0

3.6

3.5

3.2

1.5

Under study

7.7

7.0

6.1

4.3

3.1

4.6

3.9

3.8

3.2

0.6

-2.2

-4.5

-4.0

-2.4

0.1

-1.6

1.5

1.3

0.7

Published

Arua

7.5

7.1

6.5

3.2

3.1

4.0

2.6

2.4

2.0

-0.5

-3.4

-6.3

-5.9

-4.3

-1.6

-2.4

-0.4

0.1

0.2

Under study

8.6

6.9

11.4

5.9

5.0

2.6

2.7

1.9

1.3

1.1

-2.1

-2.5

-4.6

-0.9

-5.5

-1.9

-1.1

0.2

0.5

Published

6.8

6.3

11.5

4.1

1.3

1.6

0.9

0.5

0.4

-2.8

-3.1

-5.6

-1.7

-6.8

-1.7

-1.9

-0.4

0.3

Under study

Fort Portal

Vincent Musoke Nsubuga

Journal africain de statistiques, numéro 20, février 2018

The African Statistical Journal, Volume 20, February 2018

102.24

101.74

102.15

103.46

105.26

106.21

106.91

110.13

110.41

112.67

118.88

116.31

122.39

Sep-10

Oct-10

Nov-10

Dec-10

Jan-11

Feb-11

Mar-11

Apr-11

May-11

Jun-11

Jul-11

Aug-11

Sep-11

123.40

101.32

Aug-10

Nov-11

100.70

Jul-10

123.48

Month

Oct-11

313.9763

Published

Weight

Particulars

123.28

123.36

122.28

116.26

118.80

112.68

110.44

110.10

106.90

106.20

105.28

103.46

102.15

101.70

102.19

101.27

100.65

318.6325

Under study

Kampala High Income

128.82

128.89

127.63

120.20

121.64

114.64

112.14

111.93

108.67

106.15

104.47

103.68

102.09

102.00

102.56

102.02

101.12

156.6087

Published

128.79

128.85

127.59

120.18

121.62

114.64

112.15

111.92

108.67

106.15

104.48

103.68

102.09

101.99

102.55

102.01

101.10

157.2684

Under study

Kampala Middle Income

132.76

133.39

131.72

122.87

118.82

115.90

110.46

109.07

107.81

106.20

105.33

103.24

102.79

103.11

103.15

101.95

101.07

100.0688

Published

132.71

133.34

131.67

122.84

118.81

115.90

110.47

109.07

107.80

106.20

105.33

103.24

102.79

103.10

103.13

101.93

101.06

100.4877

Under study

Kampala Low Income

128.44

126.63

125.81

120.01

116.72

114.38

114.16

112.52

111.20

107.23

105.63

104.97

104.73

105.49

105.71

106.34

102.33

96.7521

Published

128.76

126.88

126.04

120.13

116.75

114.37

114.12

112.35

111.01

106.96

105.34

104.72

104.50

105.31

105.73

106.39

102.25

94.1887

Under study

Mbarara

Table 7: All Items CPI – Published and Under Study (July 2009 – June 2010) = 100

127.93

128.31

125.33

121.36

117.17

113.89

112.98

112.77

109.29

105.61

104.83

104.02

103.75

103.20

102.99

101.20

101.25

94.6932

Published

128.10

128.48

125.47

121.47

117.22

113.91

112.99

112.83

109.33

105.61

104.82

104.02

103.78

103.25

103.05

101.24

101.30

93.6900

Under study

Masaka

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas

127.28

129.81

131.66

130.58

131.35

130.52

131.78

132.87

134.18

Aug-12

Sep-12

Oct-12

Nov-12

Dec-12

Jan-13

Feb-13

Mar-13

Apr-13

128.40

Apr-12

127.10

127.96

Mar-12

Jul-12

128.93

Feb-12

126.35

125.96

Jan-12

Jun-12

124.70

Dec-11

127.59

Month

May-12

313.9763

Published

Weight

Particulars

133.95

132.64

131.55

130.27

131.12

130.37

131.45

129.65

127.22

127.05

126.32

127.47

128.25

127.82

128.77

125.83

124.62

318.6325

Under study

Kampala High Income

139.60

138.68

136.27

134.96

134.76

133.87

133.16

132.88

131.69

131.75

133.00

134.24

135.32

136.94

134.71

131.56

130.25

156.6087

Published

139.53

138.61

136.21

134.90

134.70

133.81

133.11

132.84

131.66

131.73

132.98

134.20

135.27

136.89

134.66

131.51

130.22

157.2684

Under study

Kampala Middle Income

137.94

136.67

135.45

132.58

132.83

132.47

131.78

131.96

134.32

131.10

134.57

135.54

136.48

135.75

135.57

131.99

133.28

100.0688

Published

137.89

136.61

135.40

132.53

132.77

132.42

131.73

131.92

134.29

131.07

134.54

135.50

136.43

135.70

135.52

131.94

133.24

100.4877

Under study

Kampala Low Income

140.01

139.46

139.55

137.55

136.71

136.17

135.30

133.83

135.15

134.79

132.63

135.63

133.73

130.63

130.10

127.10

127.18

96.7521

Published

139.78

139.25

139.38

137.37

136.48

135.97

135.03

133.49

134.75

134.28

132.06

135.27

133.98

130.76

130.23

127.15

127.38

94.1887

Under study

Mbarara

141.34

139.68

138.21

137.02

135.44

138.44

137.84

133.69

133.13

131.70

131.16

133.48

135.00

131.01

132.90

129.52

127.49

94.6932

Published

141.63

139.97

138.49

137.32

135.69

138.72

138.11

133.90

133.27

131.81

131.25

133.68

135.23

131.19

133.10

129.69

127.61

93.6900

Under study

Masaka

Vincent Musoke Nsubuga

Journal africain de statistiques, numéro 20, février 2018

The African Statistical Journal, Volume 20, February 2018

136.21

136.03

138.73

138.79

138.98

139.97

139.26

139.44

140.73

139.50

140.69

140.24

141.77

Jul-13

Aug-13

Sep-13

Oct-13

Nov-13

Dec-13

Jan-14

Feb-14

Mar-14

Apr-14

May-14

Jun-14

Jul-14

Aug-14

143.43

136.24

Jun-13

Oct-14

135.26

May-13

143.72

Month

Sep-14

313.9763

Published

Weight

Particulars

143.14

143.38

141.46

139.96

139.95

140.37

139.21

140.41

139.14

138.95

139.65

138.65

138.46

138.41

135.75

135.98

136.00

135.01

318.6325

Under study

Kampala High Income

147.59

146.76

145.93

144.48

143.80

144.67

145.66

145.27

143.25

142.32

143.52

144.09

144.62

145.42

144.00

141.87

141.11

139.63

156.6087

Published

147.51

146.67

145.85

144.41

143.73

144.60

145.58

145.19

143.17

142.25

143.44

144.01

144.53

145.33

143.92

141.80

141.04

139.57

157.2684

Under study

Kampala Middle Income

146.83

147.19

145.83

144.41

144.78

145.58

146.67

145.92

145.28

143.38

144.10

144.83

146.74

146.62

145.77

143.59

140.37

139.49

100.0688

Published

146.76

147.10

145.75

144.34

144.71

145.50

146.59

145.83

145.20

143.30

144.03

144.74

146.65

146.53

145.68

143.52

140.30

139.42

100.4877

Under study

Kampala Low Income

143.97

144.95

144.29

143.60

142.64

143.08

144.65

143.74

140.06

139.70

141.10

142.63

144.94

145.05

145.94

141.75

139.74

140.20

96.7521

Published

143.86

144.92

144.22

143.47

142.54

143.00

144.61

143.73

139.94

139.54

141.02

142.62

144.96

145.06

145.96

141.59

139.52

140.01

94.1887

Under study

Mbarara

142.42

142.07

144.17

143.64

142.05

142.02

143.51

143.71

142.53

141.25

140.17

143.42

145.17

145.28

144.47

142.33

139.90

140.16

94.6932

Published

142.68

142.36

144.48

143.94

142.34

142.32

143.83

144.04

142.84

141.56

140.46

143.76

145.54

145.65

144.83

142.62

140.17

140.45

93.6900

Under study

Masaka

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas

149.60

151.33

152.82

153.49

153.59

153.91

153.29

153.62

155.27

Jul-15

Aug-15

Sep-15

Oct-15

Nov-15

Dec-15

Jan-16

Feb-16

Mar-16

146.27

Mar-15

147.53

143.64

Feb-15

Jun-15

143.98

Jan-15

147.45

142.10

Dec-14

May-15

143.06

Nov-14

147.63

Month

Apr-15

313.9763

Published

Weight

Particulars

155.24

153.44

153.11

153.72

153.40

153.30

152.58

151.09

149.33

147.26

147.11

147.29

145.94

143.38

143.71

141.69

142.79

318.6325

Under study

Kampala High Income

161.01

159.57

159.01

159.92

159.43

158.83

157.33

155.81

154.40

153.54

154.76

155.05

153.52

150.63

149.36

146.99

147.38

156.6087

Published

161.00

159.52

158.95

159.87

159.37

158.78

157.27

155.75

154.33

153.46

154.67

154.95

153.43

150.55

149.29

146.89

147.31

157.2684

Under study

Kampala Middle Income

161.73

158.97

159.73

160.25

159.66

159.75

157.01

152.74

153.01

152.39

152.87

151.92

149.63

148.45

146.36

145.97

100.0688

Published

161.69

158.92

159.68

160.20

159.61

159.70

156.95

152.68

152.94

152.32

152.79

151.84

149.55

148.38

146.30

145.87

145.90

100.4877

Under study

Kampala Low Income

154.80

154.26

154.43

154.39

154.13

155.22

153.40

149.10

147.40

146.61

147.21

149.67

146.15

143.15

142.03

142.93

143.12

96.7521

Published

154.45

154.13

154.30

154.25

153.99

155.10

153.34

148.93

147.24

146.48

147.20

149.69

146.11

142.99

141.81

142.95

142.96

94.1887

Under study

Mbarara

162.02

161.17

161.59

160.94

160.96

157.68

154.70

152.93

151.36

150.99

150.59

148.32

146.55

146.31

146.11

143.52

94.6932

Published

162.16

162.30

161.47

161.88

161.22

161.24

157.98

154.98

153.22

151.67

151.35

150.93

148.63

146.83

146.58

146.52

143.78

93.6900

Under study

Masaka

Vincent Musoke Nsubuga

Journal africain de statistiques, numéro 20, février 2018

Published

56.7989

100.21

101.33

103.17

103.61

103.82

104.15

104.83

106.65

113.19

114.58

116.08

115.33

116.93

120.16

125.98

126.52

127.10

Particulars

Weight

Month

Jul-10

Aug-10

Sep-10

Oct-10

Nov-10

Dec-10

Jan-11

Feb-11

The African Statistical Journal, Volume 20, February 2018

Mar-11

Apr-11

May-11

Jun-11

Jul-11

Aug-11

Sep-11

Oct-11

Nov-11

127.23

126.66

126.11

120.24

116.98

115.36

116.12

114.66

113.26

106.66

104.81

104.16

103.84

103.66

103.22

101.36

100.24

56.2978

Under study

Mbale

130.15

128.01

127.34

122.69

120.59

119.53

117.44

115.57

114.08

107.87

106.47

105.07

103.32

102.91

101.82

100.55

99.98

56.5224

Published

Jinja

130.29

128.13

127.45

122.77

120.65

119.58

117.48

115.64

114.15

107.88

106.47

105.07

103.33

102.95

101.85

100.58

100.00

56.0942

Under study

129.42

129.58

125.78

123.74

119.73

114.57

114.85

114.06

111.91

107.77

103.27

101.31

99.92

99.57

99.61

99.21

100.28

50.4143

Published

Under study

129.58

129.75

125.91

123.86

119.81

114.59

114.88

114.13

111.97

107.79

103.24

101.28

99.91

99.57

99.62

99.23

100.31

49.9660

Gulu

132.62

132.06

131.66

125.10

122.24

114.68

114.15

112.37

110.60

106.62

105.62

98.84

97.76

101.66

101.61

101.59

36.5573

Published

Under study

132.76

132.19

131.79

125.17

122.26

114.63

114.09

112.34

110.57

106.56

105.52

98.74

97.69

101.68

101.66

101.68

36.1974

Arua

Table 8: All Items CPI – Published and Under Study (July 2009 – June 2010) = 100

140.21

139.91

137.01

126.51

123.46

122.44

122.79

120.13

118.49

114.85

112.32

109.81

109.41

107.55

106.73

105.82

106.11

37.6079

Published

140.55

140.25

137.32

126.69

123.60

122.57

122.93

120.30

118.65

114.97

112.40

109.89

109.51

107.66

106.83

105.92

106.23

37.1773

Under study

Fort Portal

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas

Published

56.7989

128.33

131.41

135.51

134.06

135.45

136.14

134.65

134.50

131.06

131.04

131.53

133.29

133.14

134.04

133.66

134.05

135.22

Weight

Month

Dec-11

Jan-12

Feb-12

Mar-12

Apr-12

May-12

Jun-12

Jul-12

Aug-12

Sep-12

Oct-12

Nov-12

Dec-12

Jan-13

Feb-13

Mar-13

Apr-13

135.40

134.23

133.85

134.26

133.34

133.48

131.70

131.18

131.15

134.62

134.77

136.34

135.65

134.24

135.71

131.58

128.44

56.2978

Under study

Mbale

Particulars

137.52

138.70

138.24

134.41

135.02

133.84

133.37

132.56

132.99

134.73

135.18

134.54

134.05

133.72

134.64

133.01

131.71

56.5224

Published

Jinja

137.70

138.90

138.45

134.60

135.21

134.01

133.53

132.70

133.07

134.82

135.27

134.67

134.20

133.86

134.79

133.15

131.82

56.0942

Under study

139.70

138.08

137.01

135.69

135.08

134.16

132.62

136.47

135.65

135.07

135.99

137.05

137.86

137.47

138.28

129.46

127.40

50.4143

Published

Under study

139.94

138.31

137.24

135.93

135.30

134.36

132.80

136.67

135.79

135.20

136.11

137.25

138.09

137.68

138.50

129.61

127.50

49.9660

Gulu

138.69

139.23

140.98

139.20

136.36

137.00

137.31

138.72

139.23

140.68

141.78

139.04

137.56

138.09

136.87

133.39

131.39

36.5573

Published

Under study

138.93

139.48

141.24

139.45

136.53

137.18

137.51

138.89

139.33

140.78

141.87

139.19

137.68

138.22

136.96

133.46

131.42

36.1974

Arua

146.65

145.41

144.06

146.61

146.94

149.01

146.48

143.95

144.12

142.63

142.55

147.73

145.55

142.71

141.09

140.56

139.59

37.6079

Published

147.04

145.79

144.43

147.05

147.37

149.45

146.88

144.29

144.40

142.89

142.79

148.13

145.93

143.05

141.41

140.88

139.87

37.1773

Under study

Fort Portal

Vincent Musoke Nsubuga

Journal africain de statistiques, numéro 20, février 2018

Published

56.7989

136.78

138.20

138.48

137.21

138.51

138.85

138.46

137.78

138.34

138.87

140.88

142.10

142.55

142.17

142.71

141.45

141.33

141.96

Weight

Month

May-13

Jun-13

Jul-13

Aug-13

Sep-13

Oct-13

Nov-13

Dec-13

Jan-14

The African Statistical Journal, Volume 20, February 2018

Feb-14

Mar-14

Apr-14

May-14

Jun-14

Jul-14

Aug-14

Sep-14

Oct-14

142.17

141.58

141.68

142.95

142.42

142.82

142.35

141.13

139.10

138.57

138.00

138.71

139.11

138.75

137.43

138.69

138.41

136.99

56.2978

Under study

Mbale

Particulars

144.08

143.48

143.81

143.69

142.14

143.01

142.07

144.12

141.90

140.92

141.08

140.83

142.55

140.54

141.29

138.69

138.97

139.18

56.5224

Published

Jinja

144.28

143.71

144.03

143.91

142.35

143.24

142.28

144.36

142.12

141.14

141.30

141.07

142.81

140.76

141.53

138.87

139.16

139.39

56.0942

Under study

145.38

146.61

145.74

146.22

144.16

146.43

149.62

151.13

150.85

147.25

143.01

142.04

142.41

144.09

143.37

140.74

141.13

141.91

50.4143

Published

Under study

145.63

146.91

146.02

146.50

144.43

146.74

149.95

151.50

151.21

147.58

143.29

142.32

142.71

144.40

143.67

140.98

141.37

142.18

49.9660

Gulu

145.23

145.24

146.90

146.74

143.46

143.11

145.49

145.17

144.30

142.45

139.28

140.01

141.03

141.24

140.04

137.17

137.79

139.14

36.5573

Published

Under study

145.38

145.47

147.17

147.01

143.70

143.35

145.81

145.49

144.60

142.73

139.52

140.29

141.32

141.51

140.29

137.35

137.98

139.33

36.1974

Arua

157.90

157.04

154.76

152.63

154.81

159.64

158.56

155.13

151.91

153.72

152.66

152.79

154.31

151.62

149.76

145.33

146.83

148.19

37.6079

Published

158.37

157.55

155.23

153.06

155.29

160.19

159.09

155.63

152.36

154.20

153.13

153.28

154.83

152.09

150.21

145.68

147.21

148.61

37.1773

Under study

Fort Portal

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas

Published

56.7989

142.12

141.08

142.01

144.11

145.13

147.57

147.89

144.83

143.72

144.73

146.19

149.11

149.95

149.64

149.56

149.61

150.93

Weight

Month

Nov-14

Dec-14

Jan-15

Feb-15

Mar-15

Apr-15

May-15

Jun-15

Jul-15

Aug-15

Sep-15

Oct-15

Nov-15

Dec-15

Jan-16

Feb-16

Mar-16

150.94

149.72

149.69

149.76

150.08

149.23

146.33

144.86

143.87

145.02

148.15

147.83

145.37

144.32

142.20

141.38

142.34

56.2978

Under study

Mbale

Particulars

157.67

157.90

158.00

157.06

155.61

155.43

153.22

151.24

150.89

149.96

149.63

150.30

147.36

147.55

144.79

143.69

144.33

56.5224

Published

Jinja

157.69

158.04

158.14

157.19

155.74

155.56

153.37

151.38

151.07

150.16

149.87

150.55

147.59

147.76

144.98

143.98

144.52

56.0942

Under study

159.05

158.11

156.09

156.07

156.40

153.25

154.00

152.74

150.98

150.67

149.52

150.73

152.85

148.82

148.22

145.56

145.82

50.4143

Published

Under study

159.14

158.31

156.29

156.26

156.59

153.42

154.22

152.96

151.21

150.92

149.81

151.03

153.17

149.08

148.47

145.91

146.06

49.9660

Gulu

161.73

157.79

158.68

156.63

157.41

154.33

154.96

153.41

152.49

150.22

150.96

149.71

149.78

148.11

147.15

145.79

145.01

36.5573

Published

Under study

161.88

157.91

158.86

156.76

157.55

154.44

155.07

153.50

152.62

150.36

151.15

149.88

149.93

148.22

147.25

146.01

145.05

36.1974

Arua

166.83

166.30

167.45

169.45

168.27

164.51

164.37

163.60

163.34

159.13

164.76

163.17

159.64

155.96

157.25

158.36

158.25

37.6079

Published

167.04

166.66

167.85

169.86

168.67

164.86

164.78

164.01

163.79

159.56

165.32

163.70

160.12

156.38

157.68

158.96

158.72

37.1773

Under study

Fort Portal

Vincent Musoke Nsubuga

Journal africain de statistiques, numéro 20, février 2018

The African Statistical Journal, Volume 20, February 2018

102.74

104.00

103.87

105.03

106.68

107.65

109.64

112.62

114.13

116.38

116.03

116.90

121.32

Sep-10

Oct-10

Nov-10

Dec-10

Jan-11

Feb-11

Mar-11

Apr-11

May-11

Jun-11

Jul-11

Aug-11

Sep-11

118.93

102.00

Aug-10

Nov-11

101.56

Jul-10

120.20

Month

Oct-11

62.5185

Published

Weight

Particulars

118.65

119.82

120.86

116.65

115.84

116.16

114.04

112.30

109.42

107.53

106.68

104.97

103.73

103.68

102.47

101.73

101.27

66.8400

Under study

Kampala High Income

115.16

116.11

117.95

112.99

111.56

111.74

110.10

109.46

106.79

105.43

104.17

103.84

104.01

105.18

103.20

102.57

102.20

17.4231

Published

115.16

116.08

117.87

113.00

111.61

111.78

110.18

109.42

106.78

105.45

104.25

103.85

103.95

105.02

103.08

102.44

102.06

17.9000

Under study

Kampala Middle Income

118.98

121.23

123.26

116.73

113.66

112.71

110.81

110.83

108.01

105.81

104.63

103.67

103.83

104.21

102.96

102.60

102.11

10.3000

Published

118.88

121.06

123.03

116.65

113.66

112.73

110.88

110.75

107.97

105.81

104.69

103.69

103.78

104.07

102.83

102.47

101.97

10.6000

Under study

Kampala Low Income

121.49

120.47

119.79

119.97

114.90

113.97

117.27

112.03

113.65

108.38

107.59

106.63

106.32

107.29

104.51

102.42

102.46

12.4097

Published

122.73

121.17

120.37

121.10

114.61

113.79

117.76

110.18

112.47

106.15

105.36

104.63

104.60

105.99

104.35

101.81

101.73

9.7400

Under study

Mbarara

Table 9: CPI Transport – Published and Under Study (July 2009 – June 2010) = 100

123.99

121.98

121.07

116.61

116.85

114.54

118.43

116.64

113.87

104.65

100.79

102.59

104.33

108.55

100.69

98.34

99.67

10.3894

Published

125.22

123.00

122.00

117.13

117.39

114.80

119.19

117.78

114.85

104.55

100.10

102.43

104.67

109.71

100.95

98.41

99.98

9.2800

Under study

Masaka

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas

121.04

120.60

121.48

124.18

126.86

122.94

122.55

125.20

126.07

Aug-12

Sep-12

Oct-12

Nov-12

Dec-12

Jan-13

Feb-13

Mar-13

Apr-13

117.99

Apr-12

120.56

121.08

Mar-12

Jul-12

121.20

Feb-12

120.80

122.53

Jan-12

Jun-12

125.00

Dec-11

118.94

Month

May-12

62.5185

Published

Weight

Particulars

125.53

124.62

122.09

122.27

126.05

123.59

121.10

120.38

121.14

120.74

121.05

118.90

117.93

120.83

120.93

122.12

124.61

66.8400

Under study

Kampala High Income

124.96

123.68

120.68

121.56

126.92

121.57

120.91

119.82

120.34

118.58

118.34

116.83

115.83

121.17

121.23

119.67

123.23

17.4231

Published

124.77

123.50

120.54

121.32

126.59

121.40

120.78

119.76

120.41

118.71

118.52

116.89

115.88

121.09

121.14

119.60

123.14

17.9000

Under study

Kampala Middle Income

129.44

127.84

123.14

124.54

130.27

125.11

124.05

123.05

123.58

122.68

122.66

120.81

119.28

124.86

124.60

125.63

130.01

10.3000

Published

129.12

127.53

122.93

124.21

129.84

124.84

123.83

122.90

123.57

122.72

122.73

120.76

119.24

124.67

124.41

125.39

129.73

10.6000

Under study

Kampala Low Income

128.27

128.65

127.85

129.33

128.56

127.36

127.83

125.94

127.14

133.46

128.28

127.45

125.17

125.10

122.39

122.36

121.97

12.4097

Published

122.81

123.65

122.90

125.30

124.09

122.97

123.15

120.45

121.06

128.22

121.56

121.65

125.25

124.90

121.47

121.57

122.42

9.7400

Under study

Mbarara

126.06

125.25

125.42

125.06

126.22

125.74

125.86

124.58

124.33

124.79

123.74

127.79

121.84

121.11

124.28

125.90

124.96

10.3894

Published

127.14

126.39

126.68

126.60

127.70

127.07

127.15

125.50

124.59

125.02

123.69

129.08

122.53

121.68

125.26

127.18

125.81

9.2800

Under study

Masaka

Vincent Musoke Nsubuga

Journal africain de statistiques, numéro 20, février 2018

The African Statistical Journal, Volume 20, February 2018

129.25

127.69

127.68

127.40

127.56

129.43

127.39

127.49

128.06

128.29

126.37

125.75

126.50

126.59

Jul-13

Aug-13

Sep-13

Oct-13

Nov-13

Dec-13

Jan-14

Feb-14

Mar-14

Apr-14

May-14

Jun-14

Jul-14

Aug-14

126.62

129.15

Jun-13

Oct-14

124.63

May-13

125.61

Month

Sep-14

62.5185

Published

Weight

Particulars

126.28

125.12

126.10

126.05

125.25

125.78

127.61

127.34

126.83

126.68

128.60

126.75

126.56

126.90

128.56

128.43

124.11

66.8400

Under study

Kampala High Income

126.95

126.63

126.98

127.84

127.34

127.51

129.00

128.90

128.77

128.94

131.60

131.31

130.66

130.78

131.46

131.16

123.95

17.4231

Published

126.81

126.41

126.78

127.63

127.11

127.25

128.71

128.60

128.47

128.62

131.22

130.89

130.24

130.40

131.13

130.83

123.77

17.9000

Under study

Kampala Middle Income

130.43

130.42

131.08

130.95

130.03

130.35

132.13

131.88

131.81

130.28

133.89

133.72

133.81

133.65

133.63

135.30

135.42

127.50

10.3000

Published

130.19

130.09

130.76

130.64

129.71

130.00

131.75

131.47

131.41

129.91

133.42

133.21

133.28

133.16

133.14

134.85

134.95

127.21

10.6000

Under study

Kampala Low Income

128.80

128.27

128.98

129.51

128.91

127.52

129.28

129.43

127.77

127.55

127.47

127.21

127.51

127.70

127.73

128.15

127.74

126.62

12.4097

Published

123.55

123.40

124.07

124.37

124.20

122.48

124.65

125.37

123.21

122.64

122.93

122.86

122.95

122.98

122.96

122.80

122.22

121.02

9.7400

Under study

Mbarara

128.17

126.08

125.82

126.32

124.89

125.71

125.04

124.99

125.07

125.34

125.22

124.82

125.35

125.53

126.01

126.12

126.78

126.24

10.3894

Published

129.07

127.11

126.70

127.20

125.77

126.78

125.97

126.03

126.08

126.47

126.33

126.07

126.73

126.79

127.33

127.10

127.90

127.46

9.2800

Under study

Masaka

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas

131.09

132.69

132.64

134.01

134.16

134.85

135.55

135.82

136.91

Jul-15

Aug-15

Sep-15

Oct-15

Nov-15

Dec-15

Jan-16

Feb-16

Mar-16

126.75

Mar-15

128.02

127.26

Feb-15

Jun-15

128.18

Jan-15

126.16

124.85

Dec-14

May-15

127.56

Nov-14

126.64

Month

Apr-15

62.5185

Published

Weight

Particulars

137.90

136.11

135.84

135.18

134.48

134.34

132.74

132.73

131.00

127.95

125.91

126.34

126.45

127.04

127.91

124.00

127.22

66.8400

Under study

Kampala High Income

133.74

136.92

129.31

128.97

129.01

129.02

128.91

129.00

128.48

128.13

127.68

127.27

127.43

127.71

125.78

126.80

17.4231

Published

134.28

137.05

129.59

129.28

129.32

129.33

129.09

129.15

128.53

128.13

127.55

127.14

127.34

127.61

125.41

126.69

17.9000

Under study

Kampala Middle Income

137.70

136.16

134.68

135.10

135.09

135.14

134.33

134.12

132.76

131.01

130.05

129.87

130.11

130.57

131.18

129.07

130.43

10.3000

Published

138.01

136.33

134.83

135.26

135.25

135.31

134.38

134.14

132.70

130.93

129.85

129.66

129.90

130.39

130.99

128.58

130.21

10.6000

Under study

Kampala Low Income

133.10

133.38

135.65

133.49

133.60

133.55

131.71

131.34

130.37

129.83

128.06

128.33

128.40

128.85

128.71

127.90

129.35

12.4097

Published

123.78

126.35

129.25

126.37

126.52

126.44

125.13

124.80

124.08

123.99

122.66

122.58

123.07

123.36

122.95

123.86

124.02

9.7400

Under study

Mbarara

146.31

145.75

148.59

149.77

148.57

143.00

140.61

138.77

138.90

136.72

133.88

134.64

133.68

132.43

131.55

129.48

128.99

10.3894

Published

145.84

146.62

150.04

151.28

149.93

143.68

141.57

139.63

140.16

138.01

135.36

136.19

135.12

133.51

132.50

131.62

129.82

9.2800

Under study

Masaka

Vincent Musoke Nsubuga

Journal africain de statistiques, numéro 20, février 2018

Published

5.1454

101.98

102.59

102.62

106.44

106.73

105.55

106.72

106.21

106.69

115.95

114.85

112.85

117.15

118.38

117.56

119.86

121.18

Particulars

Weight

Month

Jul-10

Aug-10

Sep-10

Oct-10

Nov-10

Dec-10

Jan-11

Feb-11

The African Statistical Journal, Volume 20, February 2018

Mar-11

Apr-11

May-11

Jun-11

Jul-11

Aug-11

Sep-11

Oct-11

Nov-11

122.10

120.65

118.08

119.15

117.75

112.92

115.18

117.05

106.79

106.30

106.75

105.77

107.37

103.12

103.19

102.58

4.5800

Under study

Mbale

134.29

131.09

130.99

130.42

126.74

123.89

122.29

118.05

113.46

110.70

110.83

111.12

109.43

109.62

104.52

103.66

103.60

6.3167

Published

Jinja

135.93

132.49

132.38

131.88

127.87

124.78

123.05

118.92

114.06

111.10

111.15

111.71

110.10

110.54

105.05

104.18

5.8200

Under study

126.14

119.83

118.13

118.34

117.44

116.74

116.58

117.44

115.66

109.51

107.66

106.36

104.33

104.08

98.53

98.45

99.71

5.2142

Published

Gulu

127.49

120.51

118.64

118.99

117.99

117.22

117.05

118.56

116.73

109.95

107.79

106.64

104.63

104.64

98.52

99.99

4.7100

Under study

130.20

127.87

128.17

124.63

122.69

121.94

121.30

119.02

115.82

110.71

110.98

107.95

107.24

105.84

104.49

102.62

102.75

3.9451

Published

Arua

131.31

128.72

129.14

125.35

123.02

122.18

121.56

119.49

116.08

110.54

110.60

107.94

107.62

106.53

105.27

103.65

103.80

3.5400

Under study

Table 10: CPI Transport – Published and Under Study (July 2009 – June 2010) = 100

127.38

130.78

132.51

128.94

119.51

116.75

112.19

112.12

108.57

109.49

106.25

103.79

104.04

103.76

103.20

104.38

104.91

4.1309

Published

129.15

133.05

135.02

131.12

120.44

117.32

112.18

112.77

108.93

110.02

106.22

103.78

104.36

104.38

103.80

105.24

105.93

3.6600

Under study

Fort Portal

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas

Published

5.1454

121.90

124.11

122.92

122.81

123.66

123.74

122.59

123.13

123.15

125.74

127.16

126.78

128.19

129.58

128.67

130.61

130.90

Weight

Month

Dec-11

Jan-12

Feb-12

Mar-12

Apr-12

May-12

Jun-12

Jul-12

Aug-12

Sep-12

Oct-12

Nov-12

Dec-12

Jan-13

Feb-13

Mar-13

Apr-13

132.62

132.45

130.38

131.75

129.97

128.29

128.65

126.84

123.27

123.17

122.56

124.57

123.61

123.77

125.21

122.40

4.5800

Under study

Mbale

Particulars

139.05

142.86

139.88

139.78

147.43

137.02

135.05

134.64

135.14

133.98

132.77

133.32

133.66

133.15

131.44

131.81

138.77

6.3167

Published

Jinja

140.89

145.15

142.00

142.14

150.29

138.91

136.73

136.12

136.17

134.84

133.40

134.54

135.01

134.45

132.62

133.11

140.41

5.8200

Under study

135.20

133.74

136.29

127.77

132.12

132.92

131.65

133.13

129.75

130.91

130.52

126.25

125.97

125.99

124.71

126.55

5.2142

Published

Gulu

137.21

135.73

138.66

129.49

134.13

134.94

133.48

134.93

130.62

131.83

131.25

132.01

127.38

127.02

127.07

125.74

127.49

4.7100

Under study

138.47

136.64

135.84

136.53

136.55

132.23

131.54

131.19

133.12

132.74

130.87

131.19

130.73

132.15

132.27

131.00

130.29

3.9451

Published

Arua

140.89

138.84

137.96

138.72

138.27

133.60

132.93

132.06

133.50

132.84

130.52

131.74

131.23

132.78

132.68

131.50

130.48

3.5400

Under study

135.21

136.31

135.28

140.77

132.94

132.49

134.20

132.66

133.80

134.08

133.08

133.90

133.30

132.58

133.12

133.17

130.67

4.1309

Published

137.57

138.99

137.94

144.49

135.42

134.81

136.68

134.72

135.32

135.55

134.25

136.10

135.56

134.70

135.35

135.51

132.34

3.6600

Under study

Fort Portal

Vincent Musoke Nsubuga

Journal africain de statistiques, numéro 20, février 2018

Published

5.1454

130.69

130.29

127.45

126.27

126.37

126.21

126.23

127.63

127.50

127.10

127.24

127.35

127.49

127.56

127.93

128.30

128.25

128.80

Weight

Month

May-13

Jun-13

Jul-13

Aug-13

Sep-13

Oct-13

Nov-13

Dec-13

Jan-14

The African Statistical Journal, Volume 20, February 2018

Feb-14

Mar-14

Apr-14

May-14

Jun-14

Jul-14

Aug-14

Sep-14

Oct-14

129.81

129.59

129.52

129.05

128.81

128.83

128.60

128.40

128.93

129.08

127.70

127.75

127.78

127.66

128.63

131.89

132.51

4.5800

Under study

Mbale

Particulars

144.45

143.06

142.62

142.87

144.33

144.15

144.31

144.38

143.06

141.93

144.03

142.02

142.76

141.75

141.17

141.96

140.81

140.97

6.3167

Published

Jinja

146.41

145.21

144.63

144.86

146.58

146.45

146.59

146.75

145.29

144.12

146.41

144.37

145.22

144.01

143.38

143.98

142.77

143.07

5.8200

Under study

137.37

137.13

136.96

137.05

134.26

132.26

132.88

134.01

132.51

132.97

132.72

132.68

132.80

135.08

132.63

133.02

133.03

133.70

5.2142

Published

Gulu

139.23

139.30

139.01

139.06

136.12

133.98

134.62

135.97

134.27

134.86

134.58

134.71

134.90

137.30

134.57

134.71

134.76

135.65

4.7100

Under study

139.70

138.76

138.57

137.49

139.23

139.33

140.90

140.87

140.70

140.21

137.68

137.88

137.76

137.81

138.19

138.65

138.94

3.9451

Published

Arua

140.62

140.36

140.30

139.18

141.12

141.38

143.60

143.57

143.38

142.83

140.25

140.47

140.10

140.15

140.10

140.61

140.94

3.5400

Under study

134.81

135.42

137.61

135.05

135.46

138.77

138.28

139.53

135.53

142.43

136.60

135.75

134.55

134.80

134.58

134.25

135.51

135.09

4.1309

Published

136.65

137.76

140.10

137.16

137.80

141.65

141.02

142.56

137.99

145.89

139.29

138.54

137.25

137.37

137.12

136.38

137.87

137.57

3.6600

Under study

Fort Portal

2. Consumer Price Indices (CPIs) for different population groups, income groups and geographic areas

Published

5.1454

128.67

127.52

129.65

127.64

127.29

126.44

127.77

128.76

130.09

134.02

134.85

136.08

135.97

135.69

135.25

135.08

135.55

Weight

Month

Nov-14

Dec-14

Jan-15

Feb-15

Mar-15

Apr-15

May-15

Jun-15

Jul-15

Aug-15

Sep-15

Oct-15

Nov-15

Dec-15

Jan-16

Feb-16

Mar-16

133.71

134.64

135.07

135.48

135.80

135.90

135.13

134.23

130.28

129.04

128.44

127.02

127.97

128.15

130.39

129.50

129.59

4.5800

Under study

Mbale

Particulars

148.43

148.25

151.95

152.53

148.25

148.77

146.55

146.93

146.12

143.68

142.72

142.22

141.43

141.81

143.78

142.54

144.23

6.3167

Published

Jinja

147.84

148.76

152.80

153.43

148.87

149.42

147.47

147.97

147.45

145.03

144.45

143.96

143.10

143.35

145.47

145.25

146.12

5.8200

Under study

138.66

139.78

139.30

139.73

140.02

140.23

138.76

137.34

135.61

132.24

133.43

133.87

136.20

136.18

136.69

137.76

5.2142

Published

Gulu

137.41

139.91

139.58

139.99

140.31

140.53

139.43

137.95

136.35

132.89

133.43

134.72

135.20

137.60

137.56

139.42

139.49

4.7100

Under study

145.66

146.94

148.88

143.96

144.05

146.15

144.14

143.80

141.85

140.12

136.30

134.62

135.26

137.37

140.33

137.99

139.77

3.9451

Published

Arua

145.26

146.94

149.59

143.82

143.92

146.26

144.02

143.64

141.95

140.42

136.57

134.62

135.10

137.22

140.51

139.36

139.65

3.5400

Under study

144.55

143.52

149.87

141.97

141.01

138.28

139.07

140.25

136.77

136.92

135.83

134.88

133.15

134.28

134.58

134.02

134.27

4.1309

Published

143.82

144.18

151.60

142.57

141.48

138.38

139.92

141.37

137.84

138.34

137.69

136.59

134.63

135.69

136.00

136.94

135.86

3.6600

Under study

Fort Portal

Vincent Musoke Nsubuga

Journal africain de statistiques, numéro 20, février 2018

3. Computing Consumer Price Indices and Purchasing Power Parities: A Special Case for Africa Rees Mpofu1

Abstract One of the main challenges facing some practising statisticians and economists, in Africa is mastering the properties of methods used to compute basic heading level PPPs, interrelating the same with CPI methods and properties at elementary aggregate levels, and putting in place a methodologically sound simultaneous data collection process. The methods and their properties at basic and elementary aggregate levels are the building blocks and heartbeat of both the PPPs and CPIs. The all-round mastery of PPP and CPI methods at basic and elementary aggregate levels requires consistency in product and item specifications. An understanding of the methodology at detailed levels, not only facilitates efficient data collection, but also enhances data dissemination and continuous engagement with users of PPPs— a process that naturally grows the value of the ICP data over time. The level of statistical development in Africa is uneven, so much so that a clear distinction between the CPIs and PPPs remains a challenge among some practitioners and users. The paper highlights the properties of methods using limited data and the similarities and differences between PPPs and CPIs. It demonstrates that, what is good for the CPIs in terms of item specifications may not necessarily be good for the PPPs. It emphasises the mastery of key methods and their properties as a catalyst to the successful integration of ICP data collection activities into regular pricing programs of national Statistical Offices in Africa. In so doing, the paper posits that, as with other social sciences, there is no measurement without theory in economics. Keywords: Purchasing Power Parities (PPPs), International Comparison Program (ICP), Consumer Price Index, Elementary Aggregates, Basic Headings. Abstrait L’un des principaux défis auxquels sont confrontés certains statisticiens et économistes en Afrique est de maîtriser les propriétés des méthodes utilisées pour calculer les PPA des rubriques de base, en les associant aux méthodes et propriétés de l’IPC aux niveaux agrégés élémentaires, et de mettre en place des données simultanées de processus de collecte des données. Les méthodes et leurs propriétés aux niveaux agrégés de base et élémentaire sont les briques et les pulsations des PPA et des IPC. La maîtrise globale des méthodes PPA et IPC aux niveaux agrégés de base et élémentaire nécessite des spécifications de produit et d’article. Une compréhension de la méthodologie à des niveaux détaillés, non seulement 1 Principal Statistician, Statistics Department, Statistical Capacity Building Division, African Development Bank. r.mpofu@afdb.org.

The African Statistical Journal, Volume 20, February 2018

Rees Mpofu

facilite la collecte de données efficace, mais améliore également la diffusion des données et l’engagement continu avec les utilisateurs de PPA - un processus qui augmente naturellement la valeur des données ICP au fil du temps. Le niveau de développement statistique en Afrique est inégal, si bien qu’une distinction claire entre les IPC et les PPA reste un défi pour certains praticiens et utilisateurs. Le document met en évidence les propriétés des méthodes utilisant des données limitées et les similitudes et les différences entre les PPA et les IPC. Cela démontre que ce qui est bon pour les IPC en termes de spécifications des articles n’est pas forcément bon pour les PPA. Il met l’accent sur la maîtrise des méthodes clés et de leurs propriétés en tant que catalyseur de l’intégration réussie des activités de collecte de données du PCI dans les programmes de tarification régulière des instituts nationaux de statistique en Afrique. Ce faisant, l’article postule que, comme pour les autres sciences sociales, il n’y a pas de mesure sans théorie en économie. Mots clés : Parités de pouvoir d’achat (PPA), Programme de comparaison internationale (PCI), Indice des prix à la consommation, Agrégats élémentaires, Rubriques de base. 1. INTRODUCTION

One of the biggest challenges some practicing African statisticians face is operationalizing the integration of ICP2 data requirements for the main survey into regular pricing programs of national statistical offices, as done through the national consumer price surveys. The general tendency is to provide basic data without ensuring that the product specifications meet the requested ICP specifications. This is attributable to inadequate understanding of the attributes of ICP basic data, which should pass the units of measurement test3 within and across countries. The tight product specification regime that comes with the ICP product list provides practical guidance on the collection of similar products across countries. The use of the arithmetic mean in averaging the same prices confirms the desire for homogeneity within and across countries. 2 The paper uses the terms ICP and PPPs interchangeably. The ICP is a worldwide statistical program that collects data on prices of goods and services with a view to computing PPPs that are an alternative to official market exchange rates. 3 The fact that the ICP average prices are arithmetic means is an attempt to produce national average prices that are comparable across countries in the quest for international comparison using homogeneous products.

100

Journal africain de statistiques, numéro 20, février 2018

3. Computing Consumer Price Indices and Purchasing Power Parities: A Special case for Africa

However, there is a tendency by some practitioners to pursue a “one size fit all” approach to item specifications for the ICP and CPI. Such practice is at variance with the measurement principles and detrimental to the ICP that endeavors to measure price levels across countries as opposed to the CPI that measures average price changes over time within the same countries. The possibility and existence of common products between the ICP and CPI is always there at country level, and varies from time to time. The coincidence in terms of product specifications between the ICP and CPIs is not permanent but evolves with market realities. As part of best practice, the CPI replaces item varieties from time to time as they evolve. It follows that common products require periodic qualified identification and verification. Data collectors need thorough training to clearly understand the distinction between the measurement of price levels in the context of the ICP, and average price changes in the context of the CPIs. Regular training is essential to keep data collectors abreast of market realities and improve the measurement process. It is important to understand implications of dumping products with wrong specifications into the ICP measurement process Using theory and practice to demonstrate the merit of complying with the product specifications for the ICP, this paper seeks to amplify the need to adhere to prescribed specifications leading to methodologically sound measurement of price levels. For completion and a balanced account, the paper also highlights the negative implications of mixing units of measurement using a standard example. The paper spells out the objectives and sets the tone and stage for the core business of demonstrating practical measurement steps. The paper uses limited and manageable data for the computation of CPIs and PPPs. The quantitative determination of basic heading PPPs and CPI4 price relatives at elementary aggregate levels provides useful guidance. The paper demonstrates the properties of methods, transitivity and country reversal tests.

4 The paper generally refers to the CPIs; however, we take note of the existence of harmonized consumer price indices (HCPIs) in some countries. The existence of the HCPIs does not however change this paper’s line of argument. HCPIs are special type of CPIs measured using a harmonized methodology across a set of countries.

The African Statistical Journal, Volume 20, February 2018

101

Rees Mpofu

In the conclusion, the paper stresses that the integration of ICP into regular pricing programs of the national statistical offices is knowledge driven. The paper provides an additional practical example in the Appendix for further practice as part of fostering a continuous learning culture. 1.1.1 Specific objectives The specific objectives of the paper are to: •

• •

demonstrate the computation of basic PPPs using limited data showing the equality of the Jevons – GEKS, and the Country Product Dummy (CPD) when the price tableau is full and no weights are used. It replaces space with time and compute CPI price relatives that are equal to the PPPs. use the computed basic heading PPPs to demonstrate key properties of transitivity, and base country invariance: draw parallels with transitivity and time reversal tests for the CPIs. guide African statisticians in the integration of ICP data requirements into regular pricing programs, as done through the monthly consumer price surveys, making a clear distinction between price levels and price changes.

2. BACKGROUND

The recommendation by the United Nations Statistical Commission (UNSC) to integrate ICP data into regular pricing programs of the national statistical systems is not new. The recommendation featured prominently in the evaluation of the ICP led by Jacob Ryten in the late 1990s. The Jacob Ryten report galvanized the ICP following challenges encountered by the global statistical initiative in 1993. The report stated in part: “The commission expressed support for the plan (that) there would be a departure from the costly practice of launching benchmark year comparisons every five years and move towards integration with work on national accounts and consumer price indexes”. The UNSC, at its 47th session, endorsed the ICP as a permanent statistical endeavor. This resonates with the Jacob Ryten report of the late 1990s. Integrating ICP data requirements to regular pricing programs in the Regional Member Countries (RMCs) national statistical offices through monthly consumer price surveys is fundamental to the sustainable production of PPPs.

102

Journal africain de statistiques, numéro 20, février 2018

3. Computing Consumer Price Indices and Purchasing Power Parities: A Special case for Africa

Once done, it should buttress the process and create a conveyor belt for nationally consistent data poised to avail regular or high frequency PPPs in a predictable manner. 3.

PPPS ARE AN EXTENSION OF CPIS WHEN METHODS THAT INVOKE INDEX NUMBER THEORY ARE USED AT BASIC HEADING LEVEL

There is a very close relationship between the methods used to compute the CPI calculated within a country and the PPPs across countries in practice. However, differences in terminology often give rise to confusion. Table 1 below provides the basic data used to demonstrate the equality of the methods used to compute the CPIs and PPPs in practice and in particular when all products have prices in all countries. The paper explores Table 1 data to unravel the computation of PPPs from basic heading levels using the Jevons-GEKS method and the famous Country Product Dummy (CDP). Table 1. Basic data for illustration purposes, the price data tableau is full. Description

Quantity Unit Country A

Country B Country C Country D

Paint, indoor use

33.88

34.90

753.36

89.45

Paint, outdoor use

49.19

71.34

1317.93

149.05

Silicone

300

4.54

5.29

84.74

7.54

Cement

4.57

6.30

60.07

5.55

Geometric mean 13.64 16.97 266.63 27.33 Source: World Bank (2013), Measuring the Real Size of the World Economy: The Framework, Methodology, and Results of the International Comparison Program, World Bank, Washington, DC.

We uncover the basic PPPs properties and demonstrate clearly the computation process using limited data. The practical utility of limited data enables focus and minimises confusion. Similarities between CPIs and PPPs are brought to the fore as part of statistical capacity building yet finer details distinguishing multilateral indices from temporal indices are amplified at the same time. There are four countries (A, B, C and D) involved in the international comparison of prices or the computation of PPPs. All the countries priced the

The African Statistical Journal, Volume 20, February 2018

103

Rees Mpofu

requested ICP products. Pricing all products across all ICP basic headings is not necessary but possible for some basic headings. The computations using the Jevons-GEKS and CPD follow each other. The choice of methods used in the computation of PPPs affects the results. Some methods are more efficient than others, hence understanding the strengths and weaknesses of various methods is useful as a guide. As Hill and Hill (2009) put it:

Perhaps the most pressing concern in the international comparisons literature is the problem of obtaining unbiased price indices at the basic heading level (the lowest level of aggregation at which the weights are available). The basic heading price indices provide building blocks from which the overall comparison is constructed. If these building blocks are biased or otherwise flawed, then everything that builds on them will likewise be tainted (192-93).

Efficient methods that give the best estimates are preferred and used. Based on Table 1 the general elements: annual average prices for four products are as follows;  Ρ1 Α Ρ Α  2 Ρ3 Α  Ρ4 Α

Ρ1Β Ρ2 Β Ρ3 Β Ρ4 Β

Ρ1C Ρ1 D  Ρ2 C Ρ2 D  Ρ3 C Ρ3 D   Ρ4 C Ρ4 D 

Where: Ρ1 Α = the annual average price of product 1 in country A; Ρ2 Α = the annual average price of product 2 in country A; Ρ3 Α = the annual average price of product 3 in country A; Ρ4 Α = the annual average price of product 4 in country A The specific elements: annual average prices.

104

Journal africain de statistiques, numéro 20, février 2018

R2 A = the annual average price of product 2 in country A; R3 A = the annual average price of product 3 in country A; 3. Computing Consumer Price Indices and Purchasing Power Parities: A Special case for Africa R4 A = the annual average price of product 4 in country A

The specific elements: annual average prices.

é33.88 34.90 753.36 89.45 ù ê49.19 71.34 1317.93 149.05ú ê ú ê 4.54 5.29 84.74 7.54 ú ê ú 60.07 5.55 û ë 4.57 6.30 The computations together with thethe correThe computations below belowuse usethe thenotation notationabove above together with corresponding sponding average prices. The average prices are arithmetic means of prices average prices. The average prices are arithmetic means of prices provided by the 3.1. computationcountries. of the Jevons-GEKS (J-GEKS) when the provided byThe the participating participating countries. price data tableau is full. 3.1 The computation of the Jevons-GEKS (J-GEKS) when the price We delve into thetableau similarities data is full.and differences between PPPs and CPIs, amplifying We delve into differences As between andprocess, CPIs, it is measurement factsthe ofsimilarities paramountand importance. part PPPs of the amplifying measurement facts of paramount importance. As part of theeither to important to amplify the following: index number formulas are designed process, it ischanges important to amplify the following: indexPrice number formulas measure price over time (e.g., a Consumer Index, CPI) or to are designed either to measure price changes over time (e.g., a Consumer measure Prices Levels between countries (i.e., PPPs), but they are not designed to Price both Index,aspects CPI) orsimultaneously. to measure Prices Levels countries (i.e., PPPs), measure This isbetween a critical measurement point that but they are not designed to measure both aspects simultaneously. This is that a practising statisticians need to grasp. Some statisticians believe the critical measurement point that practising statisticians need to grasp. Some Harmonised Consumer Price Indices (HCPIs) based on a harmonised 6 statisticians believe that the Harmonised Consumer Price Indices (HCPIs) methodology should price similar items and follow the ICP approach of tight based on a harmonised methodology should price similar items and follow product The reasoning no basis in theory and practice. the specifications. ICP approach of tight producthas specifications. The reasoning has no basis in theory and practice.

Given the the data above, J-GEKSmethod methodproceeds proceeds as follows. The process Given data above,the the J-GEKS as follows. The process begins with the computation of geometric mean of prices from arithmetic begins with the computation of geometric mean of prices from arithmetic means of country means data. of country data. The The simple geometric mean specification is as follows: simple geometric mean specification is as follows: 1

æ n ö n nn .....aann ç P ai ÷ == aa11aa22..... è i =1 ø

(1) (1)

When country A isAbase, thethe computation ofofPPPs The direct When country is base, computation PPPsproceeds proceedsas as follows. follows. The method successively compares the comparator country with the base direct method successively compares the comparator country with the basecountry. Eachcountry. country is country used asis used a base whilst thethesame country and andothers others are Each as a base whilst same country systematically used asused comparator countries. are systematically as comparator countries. measurement of PPPs usingthis thisapproach approachentails entails computing computing interThe The measurement of PPPs using international price relatives; PPPs are means geometric means of international price As a pricenational relatives; PPPs are geometric of international price relatives. starting point, the base country’s average prices are divided by themselves to obtain an average PPP of 1.000. For the subsequent PPPs,Volume the numerator varies whilst the base country remains the The African Statistical Journal, 20, February 2018 105 same until the possible cases are exhausted. We start with country A.

Rees Mpofu

relatives. As a starting point, the base country’s average prices are divided by themselves to obtain an average PPP of 1.000. For the subsequent PPPs, the numerator varies whilst the base country remains the same until the possible cases are exhausted. We start with country A.

R ΡA Α R ΡA Α R ΡA Α R ΡA Α RRR ΡΡΡAΑA = 44 11 ´ × 22 ´ × 33 ´ × 44 ;; R Ρ11A Α R Ρ22A Α R Ρ33A Α R Ρ44A Α A Α A

R1A R2 A R3 A R4 A ; ´ 4.54 ´ 4.´ RRR33 = 4 49 .19 57 A .88 4 ´R1A R´2 A R´3 A R4==A1.000 RRR ΡΡΡAΑA = 4 A 1.000 33.88 49.19 4.54 4.57 A Α A 33.88 49.19 4.54 4.57 = 1.000 ´ ´ ´ RRRA = 4 33 . 88 49 . 19 4 . 54 4 . 57 geometric mean of price relatives country AA is is both the Where RRRAA is the geometric mean of price relativeswhen when country the A both the comparator and the base, as shown below: L L comparator and the base, as shown below: Ρ1 Α = .the annual average price of product 1 in country A; Where RRRA is the geometric mean of price relatives when country A is both price of product 1 in country A; R11A = .the annual average L Ρ2 Α = the annual average price of product 2 in country A; and the price base, of as product shown below: the annual average 2 in country A; R22A = comparator the annual annualaverage averageprice priceofofproduct product3 3inincountry country ΡR333Α A;A; A == the R1 A = .the annual average price of product 1 in country A; of 4product 2 in A country A; Ρ Α= R the annual average price product = the annual 2 Aannual averageaverage priceofofprice product 4inincountry country A R444A = the R A = the annual average price of product 3 in country A; The same3logic applies to the calculation of annual average prices for other The sameinlogic applies to the calculation of annual average prices for other countries the comparison. the annual average price of product 4 in country A countries in the comparison. R4 A = The PPP above is 1.00, given that country A is the base. We subsequently The PPP given thattocountry A asis the thecomputation base. We subsequently change change theabove numerator or comparator countries process The sameis 1.00, logic applies the calculation of annual average prices for o the numerator orincomparator countries as theofcomputation process unfolds. We provide thecomparison. standard interpretation results in each case. unfolds. We countries the provide the standard interpretation of results in each case. The PPP above is 1.00, given that country A is the base. We subsequently cha the RΡ numerator comparator countries as the computation process unfolds. Β RΡ444B Β RΡ222B Β orRΡ333B Β 111B ; ΡΡΡBΒB =provide 444 ´× ´×interpretation ´×the standard RRR of results in each case. Α RΡ222A Α RΡ333A Α RΡ444A Α RΡ111A Α A A R1B R2 B R3 B R4 B 4 71.37 ´ 5.´29 6´.30 ; RRR35 90 B .= = 1.245 1.245 ΡΡΡBΒB = 44 A ´R1 A R´ RRR R3´A R4 A = 2A 33 . 88 49 . 19 4 . 54 4 . 57 Α A A 35.90 71.37 5.29 6.30 = 1.245 ´ ´ ´ RRRB = 4 .88R C49.19 4Journal .54 africain 4.57de statistiques, numéro 20, février 2018 106 RA11C R33 C R C ´ 22 ´ 33 ´ 44 ; RRRCC = 44 R A R22A R33A R44A 11 A A

3. Computing Consumer Price Indices and Purchasing Power Parities: A Special case for Africa

ΡΡΡC =

Ρ1C Ρ2C Ρ3C Ρ4C ; × × × Ρ1Α Ρ2 Α Ρ3 Α Ρ4 Α

RRRCC = 44 RRR ΡΡΡ RRRACCA = = 44 A Α A

735.36 1317.93 84.74 60.07 19.553 735 ..36 ..93 ..74 ..07 735 36 ´´ 1317 1317 93 ´´ 84 84 74 ´´ 60 60 07 = 19.553 33.88 49.19 4.54 4.57 == 19.553 ´ ´ ´ 33 49 44..54 33..88 88 49..19 19 54 44..57 57

RRRDD = 44 ΡΡΡ RRR RRRADDA = = 44

R11D R22 D R33 D R44 D Ρ D Ρ D Ρ D Ρ D R R R R R33 A D ´× R11A D ´× R22 A D ´× R44 A D ´ ´ ´ R R R R ´ ´ ´ 11 22 33 44 Ρ Α Ρ Α Ρ Α Ρ Α R A R A R A R R11 A R22 A R33 A R44 A A

RRRDD = 44 ΡΡΡ RRR RRRADDA = = 44

89.45 149.05 7.54 5.55 2.004 89 45 ..05 54 55 89...88 45 ´´ 149 149 05 ´´ 477...54 54 ´´ 455...54 55 = == 2.004 33 49.19 2.004 ´ ´ ´ 33 33..88 88 49 49..19 19 44..54 54 44..54 54

Α A A

Interpretation: For every local currency unit spent in country A, 1.245; 19.553; Interpretation: ForFor every local local currency unit spent in countryin A, 1.245; 19.553; Interpretation: every currency unit A, 19.553; Interpretation: For everyunits localare currency unittospent spent in country country A, 1.245; 1.245; 19.553; and 2.004local localcurrency currency required purchase the same quantity and and 2.004 units are required to purchase the same quantity and and 2.004 local currency units are required to purchase the same quantity and 2.004 local currency unitsB,are required to purchase the same quantity and and quality of products in countries C, and D respectively. quality of products in countries B, C, and D respectively. quality quality of of products products in in countries countries B, B, C, C, and and D D respectively. respectively. When country BB isis base, base,country countryBBbecomes becomesthe thedenominator. denominator. When country When When country country B B is is base, base, country country B B becomes becomes the the denominator. denominator.

RRR ΡΡΡAΑA = 44 RRR RRRBΒAAB = = 44

R Ρ33 A Α R Ρ11A Α R Ρ22 A Α R Ρ44 A Α ´× R ´ ´ R R R × × 3A 1A 2A 4A R A R A R A R A 3 1 2 4 R R R R Ρ22B Β´ Ρ33B Β´ Ρ44B Β Ρ11B Β´ ´ ´ ´ R R11B B R R22 B B R R33 B B R R44 B B

RRR ΡΡΡAΑA = RRR RRRBΒAAB = =

33.88 49.19 4.54 4.57 33 88 49 19 54 57 0.803 33...90 88 ´´ 71 49...34 19 ´´ 544...29 54 ´´ 644...30 57 == 0.803 0.803 34 0.803 ´ ´ ´ = 34 34..90 90 71 71..34 34 55..29 29 66..30 30

B B

44 4 4

B B

;; ;;

ΡΡΡBΒB = RRR RRR RRRBΒBB = =

44 4 4

Ρ33B Β R Ρ11B Β R Ρ22B Β R Ρ44B Β R ×R ×R ×R ´ ´ ´ R B 3B 1B 2B 4 R B R B R B ;; R B Ρ111B Β´ Ρ222B Β´ Ρ333B Β´ Ρ444B Β R R R R ; ´ ´ ´ R R11B B R R22 B B R R33 B B R R44 B B

ΡΡΡBΒB = RRR RRR RRRBΒBB = =

44 4 4

34.90 71.34 5.29 6.30 1.000 34 90 71 34 29 30 1.000 34...90 90 ´´ 71 71...34 34 ´´ 555...29 29 ´´ 666...30 30 = = 34 = 1.000 1.000 ´ ´ ´ 34 34..90 90 71 71..34 34 55..29 29 66..30 30

B B

The African Statistical Journal, Volume 20, February 2018

107

Rees Mpofu

ΡC R ΡC R ΡC R ΡC R ΡΡΡC = 4 R1 C ´ × R2 C ´ × R3 C ´ × R4 C ; RRR Ρ1B Β ´R Ρ2 B Β´R Ρ3B Β´R Ρ4B Β; C C C C RRRBΒC = 44 R RRRC = R11B ´ R22 B ´ R33 B ´ R44 B ; B R1B R2 B R3 B R4 B B 753.36 1317.93 84.74 60.07 ΡΡΡC = 4 753.36 ´ 1317.93 ´ 84.74 ´ 60.07 = 15.710 RRR 34.90 71.34 5.29 6.30 .36 ´ 1317 .93 ´ 84 .74 ´ 60 .07 = 15.710 RRRBΒC = 44 753 RRRC = 34.90 ´ 71.34 ´ 5.29 ´ 6.30 = 15.710 B 34.90 71.34 5.29 6.30 B

R1 D R2 D R3 D R4 D RRR D = 4 Ρ R D´ Ρ RD´ Ρ R D´ Ρ RD; D× D× D× D; ´ R44 B ´ R22 B ´ R33 B RRRBD = 44 R11B ΡΡΡ ; ´Ρ ´Ρ ´Ρ RRR D = Ρ R B R B R B R Β Β Β Β 1B 2B 3B 4B B Β R R R R B 1 2 3 4 B 89.45 149.05 7.54 5.55 RRR D = 4 89.45 ´ 149.05 ´ 7.54 ´ 5.55 = 1.610 71.34 89.90 45 ´ 149 .05 ´ 57.29 54 ´ 65.30 55 == 1.610 1.610 RRRBD = 44 34 ΡΡΡ RRR D = 34.90 ´ 71.34 ´ 5.29 ´ 6.30 = 1.610 B Β 34.90 71.34 5.29 6.30 B Interpretation: For every local currency unit spent in country B, 0.803, 15.710, Interpretation: everyunits localare currency unittospent in country B, 0.803, 15.710, and 1.610 local For currency required purchase the same quantity and Interpretation: For every local currency unit spent in country B, 0.803, 15.710, Interpretation: For every local currency unit spent in country B, 0.803, 15.710, and 1.610 local currency units are required to purchase the same quantity and quality of products in countries B,are C, required and D respectively. and 1.610 1.610 local currency units to purchase the same quantity and and local currency units are required to purchase the same quantity and quality of products in countries B, C, and D respectively. qualityofofproducts products countries B, and C, and D respectively. quality in in countries B, C, D respectively. When country C is base, country C becomes the denominator. When When country countryCCisisbase, base,country countryCCbecomes becomesthe thedenominator. denominator. When country C is base, country C becomes the denominator.

RRR ΡΡΡAΑ RRRCA RRRCA C C

= = 444 =4 =

R Ρ33 A Α´R Ρ11A Α R Ρ22 A Α´R Ρ44 A Α R A R4 A R1 A ´ R A × R2 C × R3C ×R C C R A A A A ´ ´ ´ Ρ11C ´ Ρ22 C ´ Ρ33 C ´ Ρ44 C R1C R2 C R3 C R4 C R1C R2 C R3 C R4 C

33.88 49.19 4.54 4.57 RRR A = 4 33.88 ´ 49.19 ´ 4.54 ´ 4.57 = 0.051 36 ´ 1317 .93 ´ 84 74 ´ 60 07 = 0.051 33..88 49.19 4..54 4..57 RRRCA = = 4 753 0.051 ΡΡΡ RRRΑA 4 753.36 ´ 1317.93 ´ 84.74 ´ 60.07 == 0.051 C 753 . 36 1317 . 93 84 . 74 60 . 07 RB R B R B R B C RRR B = 4 R1 B ´ R2 B ´ R3 B ´ R4 B B ´ R444C B ´ R222C B ´ R333C B RRRCB = 44 R111C ´ ´ ´R RRR B = Ρ R C R C R C C Ρ Β Β Ρ Β Ρ Β 1 2 3 4 C ΡΡΡ CB = 4 R11C × R22 C × R33C × R44 C Ρ1C Ρ2 C Ρ3 C Ρ4 C C 34.90 71.34 5.29 6.30 RRR B = 4 34.90 ´ 71.34 ´ 5.29 ´ 6.30 = 0.064 36 ´ 1317 .93 ´ 84 74 ´ 60 07 = 0.064 34..90 71.34 5..29 6..30 RRRCB = 44 753 RRR B = 753.36 ´ 1317.93 ´ 84.74 ´ 60.07 = 0.064 C 753.36 1317.93 84.74 60.07 C 108

10 10 10

Journal africain de statistiques, numéro 20, février 2018

= 0.051 ´ ´ ´ 753.36 1317.93 84.74 60.07

RRR A = C

RB R B R B R B ´ ´ ´ R1C R2 C R3 C R4 C

34.90 71.34 5.29 6.30 = 0.064 ´ ´ ´ 753.36 1317.93 84.74 60.07

3. Computing Consumer Price Power Parities: A Special case for Africa 1 2 Indices3 and Purchasing 4

RRR B = C

ΡΡΡ B = RRR C

R1C R2 C R3 C R4 C Ρ1C ´ Ρ2 C ´ Ρ3 C ´ Ρ4 C 10 ×R ×R ×R R 11C 22 C 33C 44 C R C R C R C R C 4 Ρ1C ´ Ρ2 C ´ Ρ3 C ´ Ρ4 C R R11C C´R R22C C´R R33C C´R R44C C 4 R C R C R C R C 1 2 3 4 C 753 . 36 1317 . 93 84 .74 60.07 = 1.000 ´ ´ ´ RRRC = 4 753 ..36 1317 ..93 84 ..74 60 ..07 753 36 1317 93 84 74 60 07 C ΡΡΡ = 1.000 ´ ´ ´ RRRC = 4 753 ..36 1317 ..93 84 ..74 60 ..07 753 36 1317 93 84 74 60 07 R D R D R D R D 4 3 1 2 4 = C ´ ´ ´ RRR ´ 1317 ´ .93 ´ 84.74 60.07 1.000 RRRCD = = 4 753.36 C R1C R2 C R3 C R4 C R R R R C 3D 1D 2D 4D ´ ´ ´ RRR D = 4 R Ρ Ρ Ρ Ρ R D D R D R D ΡΡΡ DC = 4 R11C ´ × R22 C ´ × R33 C ´ × R44 C RRR Ρ89C.45 R Ρ2 C149R Ρ.05 Ρ7.C R R 54 5.55 3C = 0.103 ´ ´ 4 ´ RRRCD = 4 1 753 . 36 1317 . 93 84 . 74 60 . 07 89.45 149.05 7.54 5.55 C = 0.103 ´ ´ ´ RRR D = 4 89 . 45 149 . 05 7 . 54 5 . 55 753 . 36 1317 . 93 84 . 74 60 . 07 ΡΡΡ DC = 4 = 0.103 ´ ´ ´ RRR 753.36 1317.93 84.74 60.07 C Interpretation: For every local currency unit spent in country C, 0.051, 0.064, and 0.103 local currency units are required to purchase the same quantity and Interpretation: For every local currency unit spent in country C, 0.051, 0.064, quality of productsevery in countries A, B,unit D respectively. Interpretation: localunits currency spent intocountry C,country 0.051, 0.064, and 0.103 localForFor currency are required purchase the same quantity and Interpretation: every local currency unit spent in C, 0.051, 0.064, and 0.103 local currency units are required to purchase the same quantity and and 0.103 local currency units A, are to purchase the same quantity and quality of products in countries B,required D respectively. quality ofofproducts inin countries A, B, quality products countries A, D B,respectively. D respectively. RRRC ΡΡΡCC RRRCC RRRCC

= = = =

4 4

When country D D is is base, base,country countryDDbecomes becomesthe thedenominator. denominator. When country When country D is base, country D becomes the denominator. When country D is base, country D becomes the denominator. R ΡA Α R ΡA Α R ΡA Α R ΡA Α ΡΡΡ AΑ = 4 1 ´ × 2 ´ × 3 ´ × 4 RRR Ρ Ρ Ρ Ρ R R3 D A R R1 D A R R2 D A R R4 D A D RRR A = 4 R1A ´ R2 A ´ R3 A ´ R4 A RRR AD = 4 R11 D ´ R22 D ´ R33 D ´ R44 D RD D .19R3 D4.54 R4 D 4.57 33 .88 R2 49 ΡΡΡ DAΑ = 4 1 0.499 ´ ´ ´ = 0.499 RRR 89 ..45 149 .19 05 74..54 54..55 33 88 49 . 54 57 D ´ ´ ´ = 0.499 RRR A = 4 33 ..88 49..19 47..54 45..57 89 45 149 05 54 55 4 D ´ ´ ´ = 0.499 RRR A = 89 . 45 149 . 05 7 . 54 5 . 55 Ρ Β R B Ρ Β Ρ Β Ρ Β R B R B R B ΡΡΡ DBΒ = 4 1 ´ × 2 ´ × 3 ´ × 4 RRR Ρ Ρ Ρ ΡR4 D R D R D R D R1 B R2 B R3 B R B D RRR B = 4 R1B ´ R2 B ´ R3B ´ R4 B RRR BD = 4 R11D ´ R22 D ´ R33 D ´ R44 D R1 D R2 D R3 D R4 D D The African Statistical Journal, Volume 20, February 2018

109

Rees Mpofu

34..90 90 71.34 55..29 34 29 ´ 66..30 30 = 0.621 ´ ´ RRR = 0.621 ΡΡΡ BBΒ = 44 ´ ´ ´ RRR 89 ..45 149 ..05 77..54 55..55 89 45 149 05 54 55 D D

R C R R R C R Ρ3 C Ρ1C Ρ2 C Ρ4 C R C R RRR ΡΡΡCC = × 2 ´ × 3 ´ × 4 ´ ´ ´ RRR = 44 1 ´ R Ρ11 D Ρ22 D Ρ33 D Ρ44 D R D R R D R R D R R D D D

753..36 36 1317 753 1317..93 93 ´ 84 84..74 74 ´ 60 60..07 07 = 9.756 ´ RRR ΡΡΡCC = = 9.756 ´ ´ ´ RRR = 44 89 . 45 149 . 05 7 . 54 5 . 55 89.45 149.05 7.54 5.55 D D

RRR RRR DD = = 44 ΡΡΡDD = 4 D

R R11 D D´R R22 D D ´ R33 D ´ R R44 D D ´ ´ ´ Ρ D Ρ D Ρ D Ρ4 D D 3D 1D 2D R R R R R11 D × R22 D × R33 D × R44 D Ρ1 D Ρ2 D Ρ3 D Ρ4 D

89 45 149 149..05 05 ´ 77..54 54 ´ 55..55 55 = 1.000 ΡΡΡD = = 44 89.45 ´ RRR = 1.000 ´ ´ ´ D 89 ..45 149 ..05 77..54 55..55 89 45 149 05 54 55 D D Interpretation: ForFor every local local currency unit spent in country D,country 0.499, 0.621, every currency unit spent D, Interpretation: For every local currency unit spent in in country D, 0.499, 0.499, 0.621, 0.621, and 9.756 local currency unitsunits are required to purchase the same quantity andquantity and and 9.756 local currency are required to purchase the same and 9.756 local currency units are required to purchase the same quantity and quality of products in countries A, B, and C respectively. quality quality of of products products in in countries countries A, A, B, B, and and C C respectively. respectively. We turn to the property of transitivity, and illustrate the same for the benefit of practising statisticians. We We turn turn to to the the property property of of transitivity, transitivity, and and illustrate illustrate the the same same for for the the benefit benefit of of practising statisticians. practising statisticians. Transitivity Transitivity means that every indirect multilateral PPP between a pair of Transitivity economies, calculated via a third economy, equals the direct multilateral PPP between the economies. Transitivity Transitivity means means that that every every indirect indirect multilateral multilateral PPP PPP between between aa pair pair of of economies, calculated via a third economy, equals the direct multilateral economies, calculated via a third economy, equals the direct multilateral PPP PPP between between the the economies. economies.

110

Journal africain 12 de statistiques, numéro 20, février 2018

3. Computing Consumer Price Indices and Purchasing Power Parities: A Special case for Africa

ΡΡΡ RRRΒB =

RΒ B Ρ R1Β B Ρ RΒ B Ρ RΒ B Ρ ´ 2 × ´ 3 ×´ 4 ; × R1 Α A Ρ R2 Α A Ρ R3 Α A Ρ R4 Α A Ρ

35.90 71.37 5.29 6.30 =1.245 ´ ´ ´ =1.245 33.88 49.19 4.54 4.57

Α A

ΡΡΡ RRRΒB = Α A

The computation above uses the direct method; yet the indirect method via other The computation above uses the direct the indirect methodensures via countries gives the same result as method; shown yet below. Transitivity internal other countries givesPPPs. the same result as shown below. Transitivity ensures consistency of the internal consistency of the PPPs. Alternatively, the computation of the above PPP assumes the following format: Alternatively, the computation of the above PPP assumes the following format: The indirect method The indirect method RRRB RRRB RRRB RRRB = 3 ΡΡΡ B ´ ΡΡΡC ´ ΡΡΡ D RRRΒA RRRΒA RRRΒA A C Β D C × D B × ΡΡΡΒ = 3 ΡΡΡ ΡΡΡ ΡΡΡ Α Α Α Α Β

1.0000 0.0637 0.6210 = 1.245 ´ ´ RRRB = 3 1 . 0000 0 . 0637 0 . 6210 0 . 8035 0 . 0511 0 . 4989 ΡΡΡΒA = 3 = 1.245 × × 0.8035 0.0511 0.4989 Α The country reversal test and its relevance Suppose the base country wasits B relevance and the PPP for country A was 0.8035. Suppose The country reversal test and there is a base country switch from country B to country A, the new PPP for country B is B base reciprocal the PPP for country Suppose thewhen base country country A was andisthe PPP for of country A was 0.8035. A when the Suppose therewas is a B. base country from as country B to country A, the base country The result switch is obtained follows: new PPP for country B when country A is base is reciprocal of the PPP for country A when 1 country was B. The result is obtained as follows: 1 the base = =1.245 RRR A = 1 B 0.1803 RRR B = =1.245 ΡΡΡ Α = ΡΡΡΒA 0.803 Β The relatives Abetween economies are the same no matter which economy is the base.relatives between economies are the same no matter which economy The is the base. The same logic applies to other PPPs in the comparison. The same logic applies to other PPPs in the comparison.

The African Statistical Journal, Volume 20, February 2018

111

Rees Mpofu

The general elements of theofrelevant matrices matrices based on the The generaland andspecific specific elements the relevant based on the calculations above are as follows: calculations above are as follows: ΡΡΡAΑ éRRR ê A Α êRRR ΡΡΡBΒ ê A Α êRRR ΡΡΡCC ê A Α êRRR ΡΡΡDD êë A Α

RRR ΡΡΡAΑ B Β

RRR ΡΡΡBΒ B Β

RRR ΡΡΡCC

B Β

RRR ΡΡΡAΑ C C

RRR ΡΡΡBΒ C C

RRR ΡΡΡCC C C

RRR ΡΡΡDD

B Β

C C

RRR ΡΡΡAΑ ù  D Dú RRR ΡΡΡBΒ ú  D Dú ú RRR ΡΡΡCC   D Dú ú RRR ΡΡΡDD  ú D Dû

é 1.000 0.803 ê 1.245 1.000 ê ê19.553 15.710 ê ë 2.004 1.610

0.051 0.064 1.000 0.103

0.499ù 0.621úú 9.756ú ú 1.000 û

All the elements or the PPPs are as computed above. All the elements or the PPPs are as computed above. Table 2 shows the PPPs as price ratios as computed above and for conveni-

Table 2 shows the PPPs as price ratios as computed above and for convenience when ence when each country is used as a base. each country is used as a base.

Table 2. Basic PPPs based on Table 5 basic data

Table 2. Basic PPPs based on Table 5 basic data. When country When A country B When country C When country D C Country WhenAcountry is When country B is When country Country is base isbase base. is base.base. is base. is base. Country A 1.000 1.000 0.051 Country A 0.803 0.0510.803 0.499 Country B 1.245 1.000 0.064 Country B 1.245 1.000 0.064 0.621 Country C 19.553 15.710 1.000 Country C 19.553 15.710 1.000 9.756 Country D 2.004 1.610 0.103 Country D 2.004 from Table 11.610 0.103 1.0000 Source: Derived

When co D is

Source: Derived from Table 1

Given the set of PPPs for the basic heading under consideration, the next Given the the set of PPPs for theprocedure basic heading under consideration, step called standardization reduces the above matrix to the a next step called thefour standardization procedure reduces the above matrix to a vector of four vector of elements. elements. The vector of elements is aggregated using national accounts expenditure The vector elements is aggregated using expenditure weights weights fromofbasic headings up to GDP in a fullnational scale ICP.accounts This is equivalent from basic headings up to GDP in a full scale ICP. This is equivalent to to combining elementary aggregate indices when computing CPIs using Household Budget Survey data. However, thecomputing PPPs, the operation is Household combining elementary aggregate indices for when CPIs using slightly complex of the high of data. then, the of complex Budget Survey because data. However, forvolume the PPPs, theEven operation is use slightly limited data would simplify theof process. because of the high volume data. Even then, the use of limited data would simplify the process. From the above matrix, standardization requires four basic steps shown below. the Theabove same steps are standardization applicable to whatever StandardizaFrom matrix, requiressituation. four basic steps shown below. tion is a basic heading level operation. The GEKS procedure is used for is a basic The same steps are applicable to whatever situation. Standardization 112

Journal africain de statistiques, numéro 20, février 2018

3. Computing Consumer Price Indices and Purchasing Power Parities: A Special case for Africa

heading level operation. The GEKS procedure is used for standardization process initially level at lower levels, but is feasible at higher levels the same approach is heading operation. The itGEKS procedure is used for ifstandardization process standardization process initially at lower levels, but it is feasible at higher employed at that level. initially level at lower levels, but is feasible at higher levels the same approach is heading operation. The itGEKS procedure is used for ifstandardization process levels if thelevel. same approach is employed at that level. employed at that initially at lower levels, but it is feasible at higher levels if the same approach is employed at that level. using heading level operation. The GEKS is used formatrix standardization process Standardization firstprocedure column in the above matrix a of Standardization using the firstthe column PPPs inPPPs the above gives agives variety initially at lower levels, but it is feasible at higher levels if the same approach is variety of results. Generally, for the first element, subsequent elements are results. Generally, forthe thefirst firstcolumn element, subsequent elements computed Standardization using PPPs in the above matrix are gives a variety by of employed at that level. computed by successively dividing the geometric mean of all the relevant successively dividing mean ofinallthe theabove relevant PPPs bycomputed one of them results. Generally, forthe thegeometric firstcolumn element, subsequent elements are by Standardization using the first PPPs matrix gives aAvariety PPPs by one of them until exhaustion. The general expression for is as of until exhaustion. The general expression for A is as shown below. By substitution successively dividing the geometric mean of all the relevant PPPs by one of them results. shown Generally, for the first element, subsequent elements are computed by below. Byobtained. substitution the other elements are obtained. the other elements are until exhaustion. The general expression for A is as shown below. By one substitution successively dividing mean ofinallthe theabove relevant PPPs by of them Standardization using the the geometric first column PPPs matrix gives a variety of the other elements are obtained. until exhaustion. The general expression for A is as shown below. By substitution results. Generally, for the first element, subsequent elements are computed by RRR A areΡΡΡ Α the other elements obtained. successively dividing the geometric mean of all the relevant PPPs by one of them Α A RRR =for element ; shown below. By substitution ; A is as = element A until exhaustion. The general expression × ΡΡΡ × ΡΡΡ × ΡΡΡ RRR RRR ´ RRR ´ RRR A C D Α Β 4´ΡΡΡ C D A B 4 RRR A are obtained. = element ; the other elements Α A ´ RRRΑA ´ RRRΑA ´ RRRΑA RRR A C D = element ; A B 4 A A A A B ÁRRR C ´ RRR D 4 RRR A ´ RRR Next, through substitution, specificspecific elements are obtained fromfrom the first to the Next, RRR through substitution, elements are obtained the first to last. A A A A A ; = element the last. The process unfolds as follows: The process unfolds as follows: Next, through substitution, specific elements are obtained from the first to the last. 4 RRR A ´ RRR B ´ RRR C ´ RRR D The unfolds as follows: Next,process through substitution, specific elements are obtained from the first to the last. A• A A of the Astandardization First element vector: • First element of the standardization vector: The process unfolds as follows: 1of .000 • through First element the specific standardization vector: Next, substitution, elements are obtained from the first to the last. = 0.3784 4 1as .000 1.000 ´ 1.245 ´follows: 19.553 ´ 2.004 First element of the standardization vector: The•process unfolds = 0.3784 4 1.000 1.000 ´ 1.245 ´ 19.553 ´ 2.004 = 0.3784 •• Second element of standardization vector: 4 First element of´the standardization vector: 1.000 ´ 1.245 19the .553 ´ 2.004 • Second element of the standardization vector: 1 . 245 • Second element of the standardization vector: 1.000 = 0 . 4711 = 0 . 3784 4 4 1.245 245 élement 1.000 ´óf 19 ´´22..004 • Second vector: 1.000 245 19.the .553 553standardization 004 = 0.4711 4 1.245 1.245 ´1.000 ´19.553 ´ 2.004 = 0.4711 •• Second Third element of the standardization vector: 4 standardization vector: 1.245 élement 1.000 óf 19 .the 553 ´ 2.004 19 .553 • Third element the standardization vector: 1.of 245 = = 70..3985 4711vector: • 44 11..000 Third of the 19of .´ ´11element 245 19 ..553 ´ 22..004 • Third element the standardization vector: 245 ´ ..000 ´553 19 553 ´standardization 004 = 7.3985 4 • Fourth of vector: 19.´553 1.000element ´1.245 19the .553standardization ´ 2.004 = 7.3985 4 2.of 004 • Third element the standardization vector: Fourth element the standardization vector: 1.000 ´1.245 óf 19 .553 ´ 2.004 = 0.7583 4 19 2..004 1.000element ´1.245 ´553 19the .553standardization ´ 2.004 • Fourth of vector: =7 0.3985 7583 4 2.004 1.000 ´1.245 ´19.553 ´ 2.004 0.7583 The•above results are summarised by a =vector with four elements, representing the • Fourth element the vector: 4 Fourth standardization vector: 1.000element ´1.245 óf 19the .of 553 ´ 2standardization .004 four countries being compared. Re-referencing the sameelements, elementsrepresenting by varying the The above results are2.summarised by a vector with four 004 = 0 . 7583 reference country a set Re-referencing ofbyconsistent PPPs. Beyond this the four countries beinggives compared. the same elements byprocess, varying the 4 The above summarised a vector with four elements, representing the 1process .results 000 ´1.are 245 ´19.553 ´ 2.004 weighting computes weighted PPPs. reference country a set Re-referencing of consistent the PPPs. Beyond thisbyprocess, the four countries beinggives compared. same elements varying the weighting process computes weighted PPPs. reference gives a set ofbyconsistent PPPs. process, the the The above country results are summarised a vector with fourBeyond elements,this representing weighting process computes weighted PPPs. four countries being compared. Re-referencing the same elements by varying the 15 reference country gives a set of consistent PPPs. Beyond this process, the 15 weighting process computes weighted PPPs. The African Statistical Journal, Volume 20, February 2018 113 15

Rees Mpofu

The above results are summarised by a vector with four elements, representing the four countries being compared. Re-referencing the same elements by varying the reference country gives a set of consistent PPPs. Beyond this process, the weighting process computes weighted PPPs. Table 3. The transitive vector of PPPs at basic heading level. Country A

Country B

Country C

Country D

0.4711

7.3985

0.7583

0.3784 Source: Derived from Table 2.

A switch of base countries yields consistent and transitive PPPs. PPPs are internally consistent, and any country can be used as the base or reference and does not alter the results. Drawing parallels between the PPPs as multilateral indices and CPIs as time-bound index numbers employing the same set of data. The PPPs are multilateral indices computed for a number of countries simultaneously, while the CPIs are time bound index numbers calculated within a country over time. 3.2.

Table 4 shows countries replaced by time. The months range from January to April 2016. While the PPPs assumed the participation of four countries, the computation of CPI price relatives assumes time-bound indices computed within a country for four months. Both CPIs and HCPIs use national price relatives even though PPPs are international price relatives. Table 4. Time bound price indices based on Table 5 basic data Description

Quantity

Unit

January

February

March

April

33.88

34.90

753.36

89.45

Paint, outdoor use

49.19

71.34

1317.93

149.05

Silicone

300

4.54

5.29

84.74

7.54

Cement

4.57

6.30

60.07

5.55

Geometric mean

13.64

16.97

266.63

27.33

Direct Method

1.000

1.245

19.533

2.004

1.000

1.245

19.533

2.004

Paint, indoor use

Indirect Method Source: Derived from Table 1

114

Journal africain de statistiques, numĂŠro 20, fĂŠvrier 2018

3. Computing Consumer Price Indices and Purchasing Power Parities: A Special case for Africa

Based on the above data, the geometric means of prices are computed for the months of January April. The January geometric mean isof used as ofofthe The TheJanuary January prices prices appear appearup ininto the the denominator denominator ininthe thecomputation computation ofeach each the The The January January prices prices appear appear in in the the denominator denominator in in the the computation computation of of each each of of the the the price reference period or the base month computation of direct The The January January prices prices appear appear in in the the denominator denominator in in the the computation computation of of each each of of the the price price ratios. ratios.prices By By way way ofof in notation notation and and corresponding corresponding substitutions, substitutions, the theof direct direct The The January January prices appear appear in the the denominator denominator in in the the computation computation of of each each of the the price priceCPI ratios. ratios. By By way way wayofof of of notation notation notationand and and corresponding corresponding corresponding substitutions, substitutions, substitutions, the the thedirect direct direct price relatives. price price ratios. ratios. By By and method method proceeds proceeds as asshown shown below. below. and price price ratios. ratios. By Byway way way of ofnotation notation notation andcorresponding corresponding correspondingsubstitutions, substitutions, substitutions,the the thedirect direct direct method method proceeds proceeds as as shown shown below. below. The January appear in the denominator in the computation of each of the method method proceeds proceeds asas as shown shown below. below. method method proceeds proceeds asprices shown shown below. below. The January prices appear in the denominator in thepoint, computation of each price ratios. By way of notation andstarting corresponding substitutions, the isis direct With With the the month month ofof January, January, 2016 2016 asas the the starting point, the the price price relative relative of price ratios. By way of notation and corresponding substitutions, the With Withthe the the month month of of January, January, 2016 2016 as as the the starting starting point, point, the the price price relative relative is is With With the month month of of January, January, 2016 2016 as as the the starting starting point, point, the the price price relative relative is is method proceeds as shown2016 below. computed computed as asfollows: follows: With With the the month month of of January, January, 2016 as as the the starting starting point, point, the the price price relative relative isis direct method proceeds as shown below. computed computed as as follows: follows: computed computed as as follows: follows: computed computedas asfollows: follows: With the month of January, 2016 as the starting point, the price relative is AR AR 1 month of January, 2016 as the starting point, the(the price relative is (the direct direct method) method) RR RR1 1=With =AR AR AR11the computed as follows: AR 11 1 (the (the direct direct method) method) RR RR = = AR AR AR (the (the direct direct method) method) 1 1 1 1 as follows: RR RR =computed 1 1== (the (thedirect directmethod) method) RR RR AR AR1111 11 =AR AR 11 AR AR 11 AR 1 (the direct method) (the direct = price is the price relative relative inin January, January, and and AR the average average price price for for the themethod) same same RRRR1 1 isRR AR1 1isis the 1the is is the the price price relative relative in in January, January,and and and isisthe the the average average price price for forthe the the same same R R R R AR AR AR is the price price relative relative in in January, January, and is average average price price for for the same same 11is the 1is 1 the 1 RR R R R AR AR the price price relative relative in in January, January, and and 2, the the44average average price price for for the the same same month. For For February, February, March March and and April, April, 2,1 31113isisand and are are used used respectively. respectively. The The Rmonth. R1R111 isis the AR AR month. month. For For February, February, March March and and April, April, 2, 2,3133isand 3and and and 4are 4are are are used used respectively. respectively. The The Ρ R is the price relative in January, and AP the average price for the same month. month. For For February, February, March March and and April, April, 2, 2, 4 4 used used respectively. respectively. The The 1 notation notation isisadjusted adjusted accordingly accordingly for for each eachof of four four months ininthe thesample, sample,just just asas month. month. For For February, February, March March and and April, April, 2, 2,the 3the 3 and and 44months are are used used respectively. respectively. The The notation notation is adjusted adjusted accordingly accordingly for for each each of of the four four months months in in the the sample, sample, just just as assame the February, price relative inand January, and is the price for the month. For March April, 2, 3the and 41months are usedaverage respectively. The Rare Ris1isis AR notation notation adjusted adjusted accordingly accordingly for for each each of of the the four four months in in the the sample, sample, just just as as there there are four four countries countries in in the the case case of of PPPs. PPPs. notation notation isis adjusted adjusted accordingly accordingly for forfor each each of ofof the the four four months months in ininthe the sample, sample, just just as as notation is adjusted accordingly each the four months the sample, there there are are four four countries countries in in the the case case of of PPPs. PPPs. month. For February, April, 2, 3 and 4 are used respectively. The there there are four countries countries inin the case case of PPPs. PPPs. there thereare are arefour four four countries countries inthe the theMarch case caseofof ofand PPPs. PPPs. just as there are fourin countries in the case ofofPPPs. notation is adjusted accordingly for the four months in the sample, just as AR AR1each 13 13.64 .64 1 ••there For Forare January January 2016, 2016, the the is is given given by by ==1.000. 1.000.This This RR RR = = RR RR = = AR AR 13 13 . . 64 64 1 AR four countries in the of PPPs. AR 13 .64 64 1 111 is For For January January 2016, 2016, the the is given given by by =1.000. =1.000. 1.000. 1.000. This This RR RR1=1case =AR RR RR1=111==13 =13 ...64 AR AR 13 13 64 . 64 ••••••• For For January January 2016, 2016, the the is given given by by = = This This 1== 111is RR RR For January 2016, the given RR RR 1 1 1 11 = For For January January 2016, 2016, the the RR given by by RR 1.000. This This RR11 ==AR RR AR AR11isis given 13 13 ..64 64== 1.000. 11 = AR 13 13 . 64 . 64 1 1direct AR AR 13 13..64 6413uses isis called called the thedirect directmethod. method. The The methodconsistently consistently January 11 direct AR1 method .uses 64 January isis called called the the direct direct method. method. The The direct direct method method consistently consistently uses uses January January isis isis called called the the direct direct method. method. The The direct direct method method consistently consistently uses uses January January • For January 2016, the is given by = 1.000. This This is called the direct method. The direct method consistently uses RR = RR = 2016 2016 asasthe price reference reference period. called called the theprice direct direct method. method.period. The The1 direct direct method method consistently consistently uses uses January January 1 2016 2016 as as the the price price reference reference period. period. AR 13 . 64 1 January 2016 asreference the price period. reference 2016 2016 asas as the price price reference period. 2016 2016 asthe the the price price reference reference period. period. period. is called the direct method. direct method consistently uses January AR ARThe 16 16.97 .97 22 •• For ForFebruary February 2016, 2016, the the isisgiven givenby byRR =1.245 RR RR2 2=AR =AR RR2 2=16 =16 AR AR 16 16 ..97 97=1.245 2 2 2016 as the price reference period. . 97 . 97 22 For For February February 2016, 2016,the the theRR isgiven isgiven given given by by =1.245 =1.245 RR RR=22===AR RR RR= =16 AR 16 ..97 97 AR 13 13 64 .64 ••••••• For For February February 2016, 2016, isis by by =1.245 =1.245 22== 212 is RR RR 1 is 22 = 22 = For For February February 2016, 2016, the theRR given given by by =1.245 =1.245 For February 2016,the is given RR RR RR RR = AR AR 13 13 . . 64 64 22 =AR 2 2 1 1 AR 13 13 .64 64 11 AR AR 13 13 ...64 6416.97 11 AR 2 • For February 2016, theAR is given by =1.245 RR = RR = AR 266 266 . 63 . 63 2 •• For ForMarch March2016, 2016,the theRR given by byRR =19.553 AR AR33332isisgiven RR3 3=AR =AR RR3 3=266 =266 266 266 ..63 63=19.553 AR 13 . 64 . 63 . 63 1 3 3 is For For March March 2016, 2016, the theRR isgiven given givenby by byRR =19.553 =19.553 AR RR RR=33===AR RR RR=33===266 ForMarch March2016, 2016,the given =19.553 266 .63 63 AR 13 13.64 64=19.553 ••••••• For For =19.553 33isisgiven 33 = 33 = For ForMarch March March2016, 2016, 2016,the the theRR givenby by byRR =19.553 RR RR RR RR AR AR1111isisgiven 13 13 ..64 64=19.553 33 =AR 33 =13 AR 13 . 64 . 64 11 AR AR 13 13..64 64266.63 11 AR 3 • For March 2016, the RR is given by =19.553 = RR = AR AR 27 27 .33 34 4 3.33 ••• For For April 2016, theRR isis is given given by by =2.004 =2.004 RR4 4=AR =AR RR RR = = AR AR 27 27 . . 33 33 ForApril April2016, 2016,the the given by =2.004 AR 13 . 64 4 4 4 4 27 27 .33 33 1 44 For ForApril April April2016, 2016, 2016,the the theRR isgiven isgiven given given by by =2.004 =2.004 RR RR=44===AR RR RR= =27 AR 27 ...33 33 AR 13 13 64 .64 •••••• For For isis by by =2.004 =2.004 44== 414 is RR RR 1 is 44 = 44 = For ForApril April April2016, 2016, 2016,the the theRR given given by by =2.004 =2.004 RR RR RR RR = AR AR 13 13 . . 64 64 44 =AR 4 4 AR 13 13 .64 64 1 111 AR AR 13 13 ...64 6427.33 11 AR 4 • For April 2016, the is given by =2.004 The Theindirect indirect method method involves involves chaining chaining the the price price relatives relatives from from one oneperiod periodtotothe the RR = RR = The indirect method involves chaining the pricerelatives relatives from 4 4from one period The The indirect indirect method method involves involves chaining chaining the the price price relatives from one one period period to to the the AR 13 . 64 The The indirect indirect method method involves involves chaining chaining the the price price relatives relatives from from one one period period to to the the 1 next. next. The The equality equality of of the the direct direct and and indirect indirect methods methods defines defines transitivity. transitivity. The The to the next. The involves equality the direct indirect methods tran- to The The indirect indirect method method involvesof chaining chaining the theand price price relatives relatives from fromdefines one one period period to the the next. next. The The equality equality of of the the direct direct and and indirect indirect methods methods defines defines transitivity. transitivity. The The next. next. The The equality ofof of the direct direct and and indirect indirect methods methods defines defines transitivity. The existence existence of of transitivity transitivity isisof confirmed confirmed by the equality equality ofofthe results results inintransitivity. the the lasttwo twoThe rows rows sitivity. The existence transitivity isthe confirmed by equality of last results next. next. The Theequality equality equality ofthe the the direct direct and andby indirect indirect methods methods defines defines transitivity. transitivity. The The existence existence of of transitivity transitivity is is confirmed confirmed by by the the equality equality of of results results in in the the last last two two rows rows The indirect method involves chaining the price relatives from one period existence existence of of transitivity transitivity is is confirmed confirmed by by the the equality equality of of results results in in the the last last two two rows rows in the last two rows in Table 4. in in Table Table 4. 4. existence existence of of transitivity transitivity isisconfirmed confirmed by bythe theequality equality of of results results in inthe thelast lasttwo tworows rowsto the in in Table Table 4. 4. next. inin in Table 4.4. 4. inTable Table Table 4. The equality of the direct and indirect methods defines transitivity. The existence ofcomputation transitivity is confirmed by thein equality of results in the last two We We use useuse the thethe computation computation ofofof the the price price relative inin April April 2016 2016 to demonstrate demonstrate the therows We the pricerelative relative April 2016 to to demonstrate We We use use the the computation computation of of the the price price relative relative in in April April 2016 2016 to to demonstrate demonstrate the the We We use use the computation of of the the price price relative relative inin in April April 2016 2016 toto to demonstrate demonstrate the the in the Table 4. of transitivity the property the computation of CPIs. We already have property property ofofcomputation transitivity transitivity in in the computation computation ofof CPIs. CPIs. We We already already have have the the price price We We use use the the computation computation of ofthe the thein price price relative relative in April April 2016 2016 to demonstrate demonstrate the the property property of of transitivity transitivity in in the the computation computation of of CPIs. CPIs. We We already already have have the the price price property property of transitivity in the the computation ofof of CPIs. We already already have have the the price theof price relative forin April 2016 as the 2.004 using theWe direct method above. relative relative for for April April 2016 2016 as as 2.004 2.004 using using thedirect direct method method above. above. property property of oftransitivity transitivity transitivity in in the thecomputation computation computation ofCPIs. CPIs. CPIs. We We already already have have the theprice price price relative relative for for April April 2016 2016 as as 2.004 2.004 using using the the direct direct method method above. above. We use the computation of the price relative in April 2016 to demonstrate the relative relative for for April April 2016 2016 as as 2.004 2.004 using using the the direct direct method method above. above. relative relativefor forApril April2016 2016as as2.004 2.004using usingthe thedirect directmethod methodabove. above. property of transitivity in the computation of CPIs. We already have the price relative for April 2016 as 2.004 using the direct method above. The African Statistical Journal, Volume 20, February 2018 115 17 17

Using the indirect method, we note that some terms cancel out in the chaining Rees Mpofu process. The price relative for April is a product of successive chaining beginning Using the indirect method, we note that some terms cancel out in the chaining from January . The price relative in April is a summation of the product of four process. The price relative for April is a product of successive chaining beginning ratios below. Usingillustrated the indirect method, we note that some terms cancel out in the chaining from January . The method, price relative in April is a summation of out the product of four Using indirect weApril note some terms cancel in thebeginning chaining process.theThe price relative for isthat a product of successive chaining ratios illustrated below. process. price relative for April a product of successive chaining beginning Using theThe indirect method, we note that terms cancel out inofthethe chainfrom January . The price relative in April is a summation product of four é AR ù issome AR AR AR 1 2 3 4 ´ ´ ´ R R = ing process. The price relative for April is a product of successive chaining from January . The price relative in April is a summation of the product of four å ê AR below. 7 ú ratios illustrated AR AR AR 1 1 2 3 ë û é ù beginning from AR1 January AR . The AR price AR4relative in April is a summation of the ratios illustrated below. ´ 2illustrated ´ 3 ´ below. RR7 = å ú product of êfour ratios ëéAR AR11 AR AR12 AR AR23 AR AR34 ûù .64´ AR 16.97 63 ùú 27.33 ù ´ AR266 ´ .AR RR7 = å ééê13 AR 1 ´ 2 3 4´ ´ RΡRR AR AR AR AR ´ ´ ´ R777 === å R ú = 2.004 1 1.64 216.97 3 û å êëéëê13 ú . 64 13 266 . 63 AR 13 .64 AR 16.97AR266.AR 633 û 27.33 ûù ´ RR7 = å êë 1 ´ 1 ´ 2 ú = 2.004 13 16.the 97 266 63 ëé13 ûù 13..64 64 16..64 97 266 .63above 27..33 Common terms cancel out and expression R77 == å êé13.64 ´ 16.97 ´ 266.63 ´ 27.33 úù = 2.004reduces the expression below. ΡRR Successive multiplication of terms as .63 shown above reduces to the shorter .64 16.the 97 ´ 13.64 ´ 266 RR7 = återms ëê13 ûú = 2.004 Common cancel out´and expression reduces the expression below. 13.64 13 .64 16terms .97 above 266.63 ë û expression below as common cancel each other. This the defines the important Common terms cancel out and aboveasexpression expresSuccessive multiplication ofthe terms shown reduces above to reduces to the shorter property ofterms transitivity in index numbers. Common cancel outtime-bound and the of above expression reduces the to expression below. sion below. Successive multiplication terms as shown above reduces the expression belowcancel as common terms cancel each other. This defines the important Common terms out and the above expression reduces the expression below. Successive multiplication of terms as shown above reduces to the shorter shorter expression belowinastime-bound common terms cancel each other. This defines property of transitivity index numbers. Successive multiplication of terms termsin as shown above reduces shorter AR4 below expression as common cancel each other. definestothetheimportant the property of transitivity time-bound indexThis numbers. R R7important = expression below as common terms cancel each other. This defines the important property of transitivity in time-bound index numbers. AR 1 ARof property 4 transitivity in time-bound index numbers. ΡR7 = R AR AR1 RR7 = AR44 RR7 = AR1 AR 27 .33 1 =2.004 RΡR77 = 13 . 64 27.33 =2.004 RR7 = we interchange Suppose the average prices for April and January in the 13 .we 64 Suppose interchange the average prices for April and January in the 27 . 33 computation process, =2.004 thereby switching average prices for the two months. R R = 27 . 33 7 computation process, switchingfor average prices forthe thereciprocal two months . Under =2.004 13.we 64 RR7 = the Under new order, thereby the price April becomes Suppose interchange the relative average prices for April and January in the 13 . 64 the new order, the price relative for April becomes the reciprocal of the April of the April price relative underswitching the old arrangement or natural computation process, thereby average prices for theordering. two months . Under price relative the old arrangement naturalfor ordering. Suppose we under interchange the averageorprices April and January in the the new order, the price relative for April becomes the reciprocal of the April 1 for April Suppose we interchange the average prices the computation process, thereby switching average prices for theand two January months . in Under The reciprocal is computed as follows: ΡR = = 0 . 499 7 price relative under the old arrangement or natural ordering. computation process, thereby switching average prices for the two months 2becomes .004 1 the new order, price relative for April the reciprocal of the. Under April The reciprocal isthe computed as follows: RR7 =becomes= the 0.499 the new order, the price relative for April reciprocal of the April CPIs price relatives expressed as percentages. Methods used for compupriceare relative under the old arrangement or natural ordering. 2.004 1 ordering. pricereciprocal relative under the old arrangement orpass natural tation of elementary aggregate must the transitivity The is computed asindices follows: RR = 0.499 and time 7 = 2 . 004 reversal among others. When each becomes the priceused reference CPIs aretests, price relatives expressed as month percentages. for computation 1 Methods The reciprocal isrelatives computed as follows: RR7the = transitivity 1 = 0.499 period, the price are as shown below. of elementary aggregate indices must pass and time tests, The reciprocal is computed as follows: RR7 = 2.004 = 0.499used forreversal CPIs are price relatives expressed as percentages. Methods computation 004 reference period, the price among others. When each month becomes the2.price of elementary aggregate indices must passonthe transitivity and time reversal tests, Table 5. Time price indices based Table 1Methods basic data relatives are asbound shown below. CPIs are price relatives expressed as percentages. used for computation among others. When each month becomes the price reference period, the price CPIs are price relatives expressed as percentages. Methods used computation of elementary aggregate indices%must pass the transitivity and time tests, % % %forreversal Jan=1 Feb=1 Mar=1 Apr=1 relatives are asbound shownprice below. of elementary aggregate indices must pass the transitivity time reversal Table 5. Time indices based on Table 5price basicreference dataand change change change change among others. When each month becomes the period, the tests, price among others. When each month becomes the price reference period, the price % Jan=1 % change Feb=1 % change Mar=1 % change Apr=1 relatives are asbound shown below. January 1.000 0.803 Table 5. Time price indices based 0.051 on Table 5 basic 0.499 data relatives are as shown below.

24.5 1.245 1.000 Feb=1 0.064 0.621 % 24.5 Jan=1 24.5 % change % change24.5 Mar=1 change Table 5. Time bound price indices based on Table 5 basic data 1 470.5 price 1 470.5 1 470.5 March 15.705 1 470.5 1.000 9.757 Table19.553 5. Time based% onchange Table 5 basic data % 18 Jan=1bound % changeindices Feb=1 Mar=1 change -89.8 -89.8 -89.8 April 2.004 1.6010 0.102 1.000 Jan=1 % change Feb=1 % change Mar=1 %-89.8 change 18 Source: Derived from Table 1 February

18 18 116

Journal africain de statistiques, numéro 20, février 2018

Apr=1

Apr=1 Apr=1

% %

3. Computing Consumer Price Indices and Purchasing Power Parities: A Special case for Africa

With January used as the price reference period or base, we obtain four CPI price relatives with the January figure pegged at 1.000. Just as the PPPs can use any other country in the comparison, the CPIs can use any other period as the price reference period. The price relatives for all the periods exhibit the same growth rates. The price relatives for the four months when rereferenced or switched to any of the four months produce CPI price relative equivalent to each other. The result is that the price relatives exhibit the same growth rates because, in each case, division is done using the same scalar meaning the relative magnitude is preserved. PPPs are base country invariant whilst the CPI price relatives are time invariant. The growth rates are the same across selected reference periods. CPI price relatives are not standardized but used straight away for higher-level aggregation purposes. Higher-level aggregation uses household expenditure weights from the latest household budget surveys. Aggregation follows a stepwise pattern. It is important to note that the PPPs average prices and weights refer to the same year. When methods that invoke index number theory are used, the PPPs are superlative indices. On the other hand, the CPIs, by virtue of using current prices and weights that relate to sometime in the past, are Lowe indices. Below we introduce the CDP as an alternative way for computing basic heading PPPs using the same data in Table 1. 3.3 The CPD with no weights and a complete tableau. The basic formula underlying the CPD method is multiplicative, as shown below, and is converted to an additive form by taking logarithms on both sides of the equation. Introduced by Robert Summers (1973) as a tool for filling missing price data gaps. The CPD has become a popular instrument for the computation of basic heading PPPs. The CPD’s regression formulation has been described as a “very simple type of hedonic regression model where the only characteristic of a product is the product itself” (Diewert, 2005: 561). The model in its multiplicative form is also referred to as the law of one price, which suggests that the observed price of a commodity is the product of the international price, Ρi , and the purchasing power parity of currency of country j, ΡΡΡ j . The CPD in multiplicative form is expressed as:

The African Statistical Journal, Volume 20, February 2018

117

form form is is also isalso also referred referred to to astoas the asthe law thelaw law of of one ofone one price, price, which which suggests suggests that that thethe observed theobserved observed form referred price, which suggests that form is also referred to as the law of one price, which suggests that the observed form is also referred to as the law of one price, which suggests that the observed price of a commodity is the product of the international price, , and Rthe price price of a of commodity a commodity is the is the product product of the of the international international price, price, , and , and the th i the price of a commodity is the product of the international price, , and R R R i i i R , and the price of a commodity is the product of the international price, pricepurchasing of a commodity is the product of the international price, R ii , and the power parity of currency of j,country purchasing purchasing power power parity parity of of currency ofcurrency currency of of country ofcountry country j, RRR . j, RRR . j . Rees Mpofu purchasing power parity j, RRR RRR j j .j . purchasing purchasing power power parity parity of of currency currency of of country country j, j, RRR . RRR jj The CPD in multiplicative form is expressed The The CPD CPD in in multiplicative inmultiplicative multiplicative form form is is expressed isexpressed expressed as:as:as: as: The CPD form The The CPD CPD in in multiplicative multiplicative form form is is expressed expressed as: as: rbijub=buka (2)(2)(2) (2) j b iu ij r ijr=ijr ka = = ka ij = j ka ij jijib iiju ij (2) (2) r ka u (2) r ijij = ka jj b iiu ijij

Inequation, the equation, isprice the price of iproduct i in j,country and iserror the error uerror Rijprice In In the In equation, the the isthe price the price of of product product iniiin country iinincountry country and and the is the term. term.term. ijerror the equation, product j,j, j, and is the error term. u ijuj,isiju Rij is In the equation, the ofof product country and is the RijRis ij is ij is In the equation, is the price of product i in country j, and the error term. u R In the equation, is the price of product i in country j, and is the error term. u ijij Rijij term. In form, additive form, thetakes model takes the following functional In In additive Inadditive additive form, thethe model themodel model takes thethe following thefollowing following functional functional form: form:form: form, takes functional form: In form, the model takes the following functional form: In additive form, the the model model takes takesthe thefollowing followingfunctional functionalform: form: In additive additive form, Ik +InIb nxbi + x ij (3)(3)(3) (3) InR I= na + n+a njibx+ InijIIRnn=ijR InnRII+knnij Ik += IInnIIjak InniaIIb+ +iiijxI+ ijn= j + j + ijx ij (3) R = k + n a + n b + (3) InRijij = Ink + Ina jj + Inbi + x ijij (3) The observed price data are expressed in national the national currencies of th The The observed observed price price data data areexpressed are expressed expressed inthe the national national currencies currencies ofthe the The observed price data are expressed inthe ofthe The observed price data are in theinnational currencies ofcurrencies the par- of The observed price data are expressed in the national currencies of the The observed price data are expressed in the national currencies of the participating countries in the comparison. Dummy variables with values of participating participating countries countries in the inthe comparison. thecomparison. comparison. Dummy Dummy variables variables with with values values of of 1ofand and1 an participating countries in Dummy variables with values 1 11and ticipating countries in the comparison. Dummy variables with values of 1 participating countries in the comparison. Dummy variables with values of and countries ineach the comparison. Dummy variables with values of 1regressio and 0used areto to used to represent each country (j) and product (i). The 0 0are 0participating are used used represent to represent each country country (j)and (j) and and product product (i).(i). (i). The The regression regression are represent each country (j) and product The regression 0 are used to represent each country (j) product (i). The regression 00and are used to represent each country (j) and product (i). The regression are used to represent each country (j) and product (i). The regression coefficients are estimated by ordinary least squares (OLS). It is necessary coefficients coefficients areare are estimated estimated by byordinary ordinary least least squares squares (OLS). It ItisItisnecessary isnecessary necessary to toto t coefficients estimated least squares (OLS). coefficients are estimated bybyordinary ordinary least squares (OLS). It(OLS). is necessary to coefficients are estimated by ordinary least squares (OLS). It is necessary coefficients are estimated by ordinary least squares (OLS). It is necessary to bas acountry base country aproduct base product for the such model, such that if theto specify specify aspecify base base country country and and a abase aand base product product for the forthe the model, model, such that ifthe ifthe the base base specify aaaabase and abase base model, such that ifthe base specify base country and product forfor the model, such that ifthat specify base country and a base product for the model, such that if the base specify country and abase baseproduct the model, such that ifit the base isthen I the then the isfor and follows = 1itfollows country country is isIaisbase Ithen the base base product product isproduct I,then then itfollows follows that that tha country Ithen base product isI,is and that aI,1a=then ab1 1=b=a b111and 1b1and base country country I the then the base product I,I,then then and itit=follows 1= 1a= 1b= country is IIisthen the base product is I, then and it follows that = = 1 country is then the base product is I, then and it follows that a11 = b11 = 1 n.b ln a ln =1aI=1nln = bIn1aIbn=11b= 0=1 .I= 0 1.. = 0 .. ln1that a 0 ln ln a a11 = = IIn nb b11 = =0 0. In the CPD, any other country can bebe used asbe country simply by dividIn the CPD, any other country can used ascountry base country simply by dividin In Inthe Inthe the CPD, CPD, any any other other country country can can be used be used asbase asbase base country country simply simply by dividing bydividing dividing CPD, any other country can used asbase simply by In the CPD, any other country can be used as base country simply by dividing ing every country’s PPP by the new base country’s PPP. In the CPD, any other country can be used as base country simply by dividing every country’s PPP by the new base country’s PPP. every every country’s country’s PPP PPP byby the bythe new thenew new base base country’s country’s PPP. PPP. every country’s PPP base country’s PPP. every every country’s country’s PPP PPP by by the the new new base base country’s country’s PPP. PPP. The following properties underline the CPD: The following properties underline the CPD: The The following following properties properties underline underline thethe CPD: theCPD: CPD: The following properties underline The The following following properties properties underline underline the the CPD: CPD: • Individual product PPPs within basic headings are constantarebetween • Individual product PPPs within basic headings constant between • • Individual • Individual Individual product product PPPs PPPs within within basic basic headings headings areare are constant constant between between any any an product PPPs within basic headings constant between any ••anyIndividual product PPPs within basic headings are constant between any given pair of countries. Individual product PPPs within basic headings are constant between any given pair of countries. given pair pair of of countries. of countries. given pair countries. • given Each country has overall price level that determines the absolute prices given pair of countries. given pair of countries. of products incountry the basichas heading forprice that country. •country Each overall level that determines the absolute o •• • Each •Prices Each country has has overall overall price price level level that that determines determines the absolute theabsolute absolute prices prices ofprices Each country has overall price level that determines the prices ofof vary between countries at the same rate across all products. •• Each country has overall price level that determines the absolute prices of Each country has overall price level that determines the absolute prices of products inbasic the basic heading for that country. products in in the basic thebasic heading heading for that for that country. country. products the heading for that country. • products Prices vary byin at the same rate across countries. products in the heading for that country. products inproduct the basic basic heading for that country. • vary Prices vary between countries atsame the same rate all products. •The • CPD Prices vary between between countries countries at the atthe same the rate rate across across allacross products. all products. • Prices vary between countries at same rate across all products. is an alternative way of dealing with an incomplete matrix of •• Prices Prices vary between countries at the same rate across all products. Prices vary between countries at the same rate across all products. prices — it has been utilized for the computation of PPPs in the early stages of the ICP. 202020 20 20 20

The CPD is regression analysis as shown in Tables 6 and 7 whilst the Jevons-GEKS uses average price levels to compute price relatives or PPPs for countries in the comparison as demonstrated in section 3.1 of the paper. Some regions have in the recent past favored the Jevons-GEKS and its variants given the availability of homogenous or identical products across selected outlets. The CPD is multilateral in which regression analysis is used to obtain transitive PPPs for each basic heading.

118

Journal africain de statistiques, numéro 20, février 2018

3. Computing Consumer Price Indices and Purchasing Power Parities: A Special case for Africa

The CPD method entails preparing a Table like the one below using basic data in Table 1. Given that all products were available in the four countries and hence priced across the same, the relevant table takes the format in Table 6 (below). Table 6. The CPD layout based on basic data in Table 1 Average Prices

Log of Prices

Country A

Country B

Country C

Country D

33.88

3.52

49.19

3.90

4.54

1.51

4.57

1.52

34.90

3.55

71.34

4.27

5.29

1.67

6.30

1.84

753.36

6.62

1 317.93

7.18

84.74

4.44

60.07

4.10

89.45

4.49

149.05

5.00

7.54

2.02

5.55 1.71 0 Source: Prepared using Table 1

We run the regression using the above data and obtain the results displayed in the next Table. The PPPs are derived from the regression results (Table 7) below. The country coefficients are used to derive the PPPs. The base country PPP is set at 1, with the base country omitted from the regression analysis as one of the explanatory variables. For the purposes of sustainable statistical capacity building, it is recommended that readers re-run the regression and confirm the outputs as shown below.

The African Statistical Journal, Volume 20, February 2018

119

Rees Mpofu

The PPPs are derived from the regression results (Table 7) below. The country coefficients are used to derive the PPPs. The base country PPP is set at 1, with the base country omitted from the regression analysis as one of the explanatory variables. For the purposes of sustainable statistical capacity building, it is Table 7. Regression Resultsre-run the regression and confirm the outputs as shown recommended that readers below. Regression Statistics Multiple R 0.999

Table 7. Regression Results R. Squared 0.998 Regression Statistics Adjusted R Squared Multiple R 0.886 0.999 Standard Error 0.235 R. Squared 0.998 Adjusted R Squared 0.886 Observations 16 StandardANOVA Error 0.235 df SS Observations Regression 7 16 252.9 ANOVA df Residual 9 0.50 Regression 7 Total 16 9253.41 Residual Total Coeff 16 SE X.Variable 1 0.219 Coeff 0.166 X.Variable 1 0.219 0.166 X.Variable 2 2.973 X.Variable 2 2.973 0.166 X.Variable 3 0.695 X.Variable 3 0.695 X.Variable 4 3.577 0.155 X.Variable 4 3.577 X.Variable 5 4.116 0.155 X.Variable 5 4.116 X.Variable 6 1.438 X.Variable 6 1.438 0.155 X.Variable7 7 1.321 X.Variable 1.321 0.155 Source: Derived from Table Source: Derived from Table 6 6

Sign F

36.13

655.50

0.0000000

SS 0.06 252.9 0.50 t-Stat 253.41 1.318 SE 0.166 17.909 0.166 4.188 0.166 23.032 0.155 26.506 0.155 9.259 0.155 8.504 0.155

MS 36.13 0.06

F 655.50

Sign F 0.0000000

P-value Lower 95% Upper 95% 0.220 t-Stat

1.318 0.000 17.909 0.002 4.188 0.000 23.032 0.000 26.506 0.000 9.259 0.000 8.504

-0.157 P-value 0.594 Lower 95% 2.5980.220 0.3200.000 0.002 3.225 0.000 3.765 0.000 1.0870.000 0.9690.000

3.349 1.071 3.928 4.467 1.789 1.672

-0.157 2.598 0.320 3.225 3.765 1.087 0.969

The PPPs are the exponents of the country coefficients, and this is 1.00 for the basePPPs country, by exponents definition.of the country coefficients, and this is 1.00 for The are the the base country, by definition. (by definition) RRRA = = 11..000 000 (by definition) ΡΡΡ Α A Α

219 RRR B = 2.7182800..219 = 1.245 ΡΡΡ Β A Α

RRRC = 2.71828 2.973 = 19.553 A

ΡΡΡD = 2.718280.695 = 2.004 Α

Below is the interpretation of the results.

120

Journal africain de statistiques, numéro 20, février 2018

3. Computing Consumer Price Indices and Purchasing Power Parities: A Special case for Africa 0.695 RRR RRRDD ==22.71828 .718280.695==22.004 .004 AA

Below isisthe interpretation ofofthe results. Below thelocal interpretation the results. For every currency unit used in country A to purchase goods and services of constant quality and quantity, 1.245, 19.553, and 2.004 local For every unit used inincountry totopurchase goods services For everylocal local currency unitin used country purchase goodsand and servicesofof currency unitscurrency are required countries B, C,AAand D, respectively, to do constant and quantity, constant quality andcountry. quantity, 1.245, 1.245, 19.553, 19.553, and and2.004 2.004local local currency currency units units are are likewisequality in the base required requiredinincountries countriesB, B,C, C,and andD, D,respectively, respectively,totodo dolikewise likewiseininthe thebase basecountry. country. These are the exact results obtained using the Jevons-GEKS method above. The CPD gives theresults same results as the Jevons-GEKS under special condiThese are exact obtained using the method above. These arethe the exact results obtained using theJevons-GEKS Jevons-GEKS method above.The The tions. This similarity of results obtains when countries have priced all ICP CPD gives the same results as the Jevons-GEKS under special conditions. This CPD gives the same results as the Jevons-GEKS under special conditions. This productsof for aresults given obtains basic heading. suchhave circumstances theICP CPD like similarity when countries products for similarity of results obtains whenUnder countries have priced priced all all ICP products for aa the Jevons – GEKS gives transitive PPPs; the same PPPs pass the country given givenbasic basicheading. heading.Under Undersuch suchcircumstances circumstancesthe theCPD CPDlike likethe theJevons Jevons––GEKS GEKS reversal test asPPPs; well. gives transitive gives transitive PPPs;the thesame samePPPs PPPspass passthe thecountry countryreversal reversaltest testasaswell. well. These are some of the crucial tests from a measurement point of view. These Theseare aresome someofofthe thecrucial crucialtests testsfrom fromaameasurement measurementpoint pointofofview. view. RRR RRRBB 1.245 AA = 1.245 We obtain the We are ableto tocompute computeRRR weobtain obtainthe thePPP PPPvia via ==00.064 Weare areable to compute == we .064; ;we RRRBB == RRR 19.553 .553 RRRCC 19 CC AA

other ororindirectly. Clearly, the method yields the result other PPPs indirectly. Clearly,Clearly, thedirect direct method yields thesame same resultasasthe the PPPPPPs via other PPPs or indirectly. the direct method yields the same indirect method, and this is what transitivity is all about. There is no natural indirect method, and this is what transitivity is all about. There is no natural result as the indirect method, and this is what transitivity is all about. There ordering ofof time with meaning that ensures the ordering time with PPPs, meaning that transitivity transitivity ensures the internal internal is no natural ordering of PPPs, time with PPPs, meaning that transitivity ensures consistency of PPPs. PPPs, however, also need to pass another crucial test: the consistency of PPPs. PPPs, however, also need to pass another crucial test: the the internal consistency of PPPs. PPPs, however, also need to pass another country reversal test. country test. crucial reversal test: the country reversal test. Supposethe thecomparison comparison switches switches AA to B, and all all Suppose base country from Suppose the comparison switchesthe thebase basecountry countryfrom from A to to B, B, and and all PPPs PPPs PPPs computed using country Abase as base already known.We Weeasily easily derive computed using AAasasbase are already known. derive computed usingcountry country areare already known. We easily derivethe thePPP PPP the PPP for toan B existing using an result. existing result. We compute for country AAcountry relative totorelative BBusing compute the for country relativeA using an existing result.We We compute thereciprocal reciprocalofof the reciprocal of a known obtained under a previous base country aaknown PPP under aaprevious base country arrangement. known PPPobtained obtained underPPP previous base country arrangement. arrangement. 11 = 11 = 0.803. (Country reversal) = = 0.803. (Country reversal) RRR RRRAA == RRR .245 RRRBB 11.245 BB 1 1 = 0.803. (Country reversal) ΡΡΡΑ = AA = 1 . 245 ΡΡΡ Β Β Α PPPs InIn other other words, words, PPPs are are base base country country invariant. invariant. When When the the base base country country isis changed changedororinterchanged interchangedthe theresultant resultantPPPs PPPsare arereciprocals reciprocalsofofthe thefirst firstororprevious previous arrangement ororordering. arrangement ordering. In other words, PPPs are base country invariant. When the base country is changed or interchanged the resultant PPPs are reciprocals of the first or previous arrangement or ordering. 23 23

The African Statistical Journal, Volume 20, February 2018

121

Rees Mpofu

ΡΡΡΑ The same ΡΡΡΑ = Β

ΡΡΡΒ

1 = 0.803 (transitivity) 1.245

After utilizing both the Jevons-GEKS and the CPD based on the same set of data and, particularly when the price tableau is full, we make two very important observations to enhance methodological know how. First, transitivity is an important test for both CPIs and PPPs, therefore methods used at both lower and higher levels of aggregation should pass the tests. Second, the time reversal test for time bound indices is equally useful and is equivalent to the country reversal test (or the base country invariance) in the case of the PPPs. Statisticians should endeavour to undertake the two tests using real data. Other tests like the proportionality test are equally important but not illustrated in this paper due to limitations of space among other issues. This paper uses an ideal example where all countries priced all requested ICP items. Statisticians are bound to raise a question as to what happens when there are data gaps. The Jevons-GEKS and CPD relation of equality breaks down. The Appendix section of the paper provides an example of the CPD with missing prices—a slightly different approach to the main example characterised by a full tableau of national average prices. This is part of incremental value addition and total statistical capacity building. Once there are data gaps, the CPD is preferred. The CPD produces transitive PPPs that pass the country reversal test as well even when data gaps exist. If weights are not used at basic heading level, gap filling, using the CPD does not change the results. The CPD derives part of its strength from the use of all the data. The CPD also produces measures of reliability like the standard errors as part of its standard output. The selection of methods, why the Jevons is preferred for aggregation purposes. The above results obtain under special circumstances, with the same units of measurement for specific items within the basic heading and across countries. 3.4

The average price across different items within a given basic heading is computed using the geometric, rather than the arithmetic, mean. The given average prices are arithmetic means, as countries follow a very tight product specification at country level and across countries. Such rigidity in

122

Journal africain de statistiques, numéro 20, février 2018

3. Computing Consumer Price Indices and Purchasing Power Parities: A Special case for Africa

item specification is comparable with the desire to measure price levels and to compare like with like across countries. The use of the arithmetic mean at the stage of averaging country prices to get national averages is meant to ensure compliance with the given product specification, as requested by the ICP within and across countries. The tight ICP product specification regime ensures homogeneity of products across the participating countries. While there is homogeneity at individual item levels within a basic heading, there is heterogeneity across different items within a basic heading or across products within a country. In applying the Jevons-GEKS method, the geometric mean is used for aggregation purposes at the level of basic heading, and not the arithmetic mean. The choice of method matters and is knowledge-driven. The averaging process is across different items with different units of measurement. The geometric mean does well in this operation as it passes the commensurability test; the arithmetic mean (the Dutot) fails the same test. A practical example widely used in price statistics literature demonstrates the importance of using different units of measurement for price levels and price changes. The Carli or the ratio of arithmetic means is one of the methods used to compute the CPIs at the level of elementary aggregates or micro indices. Although the Carli passes the units of measurement test, its failure to pass the transitivity and time reversal tests disqualifies it for applied work in the CPI measurement.

The African Statistical Journal, Volume 20, February 2018

123

Rees Mpofu

Table 8. Units of measurement test and practical lessons to practising statisticians A

M2/M1

Formula Name

Different Units

Price

Price ratio

Pepper (10g)

2.70

3.80

1.4074

Salt (kg)

2.50

2.70

1.0800

Arithmetic mean

2.60

3.25

1.2500

Ratio of arithmetic means

Dutot

Geometric mean

2.60

3.20

1.2329

Ratio of geometric means

Jevons

1.2437

Arithmetic mean of ratios

Carli

1.2329

Geo mean of relatives

Jevons

M2/M1

Same Units

Price

Price ratio

Formula Name

Pepper (kg)

2 700

3 800

1.4074

Salt (kg)

2.50

2.70

1.0800

Arithmetic mean

1 351

1 901

1.4071

Ratio of arithmetic means

Dutot

Geometric mean 82.16 101.29

1.2329

Ratio of geometric means

Jevons

1.2437

Arithmetic mean of ratios

Carli

1.2329 Geo mean of relatives Jevons Source: Practical Guide to Producing Consumer Price Indices (2009).

As the units of measurement vary, the geometric mean price relatives remain the same. The geometric mean price relative remains the same at 1.2329 under different units of measurement; it maintains consistency despite the variation in the units of measurement. It allows for substitution of items. The geometric mean in levels and ratios give the same result. On the other hand, the Dutot breaks down as the units of measurement vary. It changes from 1.2500 to 1.4071, this means that it is sensitive to changes in units of measurement or it fails the units of measurement test. The Dutot performs well when homogeneity prevails. What is good for the national CPIs that rely on the geometric mean to average basic data at the level of elementary aggregates is not necessarily good for the PPPs whose national average prices are prepared using the Dutot at country level. The use of the arithmetic mean to prepare annual averages for various ICP products at country level is an attempt at ensuring homogeneity.

124

Journal africain de statistiques, numĂŠro 20, fĂŠvrier 2018

3. Computing Consumer Price Indices and Purchasing Power Parities: A Special case for Africa

3.5 Similarities and differences between ICP/PPPs and CPIs Statisticians should identify the main similarities and differences and what these mean from a measurement point of view. The specifications of items is fundamental when the attempt is to measure price levels at country level for the purposes of the ICP. The following statement from the UN ICP Manual of 1992 makes a clear distinction between the CPIs and PPPs from a measurement point of view.

In contrast to the measurement of time-to-time price changes, where the necessity of absolutely identical items to be priced nationwide is not pressing, the ICP makes more rigorous demands on comparability of specifications across observations within a country. In consumer prices indices for example â&#x20AC;&#x153;like with likeâ&#x20AC;? comparisons over time for a well specified item are based on a process of averaging price relatives, and there is no need to ensure that the product priced in different shops or outlets, and in various parts of the country, are identical. Indeed, to represent time-to-time changes, the items included in an index in different regions should be more characteristic of local expenditure patterns, including point of purchase. However, for the ICP purposes, the items and the qualities they reflect should be as similar as possible for a proper comparison.

Briefly, the measurement of the CPI depends on comparing like with like a long time and within countries whilst the PPPs depend on comparing like with like across countries. 4. CONCLUSION

The paper demonstrated the equality of the Jevons-GEKS method and the CPD under special conditions. The Jevons-GEKS and the CPD methods are equal when no weights are used, the units of measurement are the same and all countries priced the requested ICP products. While in practice, it is impossible for all countries to price all requested ICP products, item specifications for the same products ought to be similar within and across countries under comparison. What this means is that if the ICP is to utilize common products the identification and specifications of those products need thorough examination and confirmation to arrive at methodologically sound estimates. The process needs close monitoring and periodic independent assessment.

The African Statistical Journal, Volume 20, February 2018

125

Rees Mpofu

Both the PPPs and CPIs use methods that pass the transitivity test, and this is consistent with best practice. The PPPs also pass the country reversal test, whilst the CPIs pass the time reversal test. It means they are all index numbers with the CPIs being time bound yet the PPPs are multilateral indices. They share the same properties, and the estimates coincide when the product specifications are the same, the price tableau is full and no weights are used at basic heading levels. 4.1. Recommendations •

Given the latest 2017 ICP product list, it is important to identify items/products that are common to the ICP and CPIs ahead of data collection. The common products should be genuine 2017 ICP products. Specifications should match requested 2017 ICP specifications, otherwise they are unfit for measurement. Once common products are identified, a strategy to account for the rest of the products needs to be worked out, ensuring its cost effectiveness and sustainability.

•

There are possibilities for CPIs and PPPs sharing identical products or items from time to time within countries and across them. It is important to ascertain that the common products are genuine ICP products. If they are genuine ICP products by specification, they should pass the units of measurement test within and across countries.

•

An understanding of the computation of the CPIs and PPPs from detailed data at lower levels is indispensable, as part of integrating data collection activities for the CPIs and PPPs. Statisticians should have a detailed understanding of various methods used for the CPIs and PPPs at elementary aggregate levels and basic heading levels.

•

Consistent capacity building on PPPs and CPI methods as part of regular activities raises practitioners’ awareness of the similarities and differences between the CPIs and PPPs.

•

PPPs computed using methods that invoke index number theory are superlative indices. The weights and prices refer to the same year. On the other hand, CPIs are Lowe indices using weights from several years back. Because CPIs are weight referenced and PPPs are current year weighted, product specifications may not necessarily coincide across the relevant product coverage range. Qualified inspection would therefore be necessary. Fast evolving products like electronics and some durable goods fit into the category of products that elude CPIs from time to time.

126

Journal africain de statistiques, numéro 20, février 2018

3. Computing Consumer Price Indices and Purchasing Power Parities: A Special case for Africa

•

There is nothing as empowering as being up-to-date with ongoing developments on the ICP at the global or international levels. Part of the acquired knowledge is the grasp of micro foundations like the tests discussed in this paper.

•

It is crucial to note that index number formulas are designed to measure price changes over time (e.g., a Consumer Price Index, CPI) or to measure Prices Levels between countries (i.e., PPPs), but not to measure both aspects simultaneously. Understanding this critical statement defines the distinction between the CPIs and PPPs.

•

Finally, there is no measurement without theory; principles that underline specific methods need mastering and putting into practice. Some statistics are crucial and part of the evolving development agenda, and their methodology need gradual appreciation. Finally, the similarities and differences that come with measuring price levels and price changes need clarity. Economic theory throws light on the choice of methodology in multilateral comparisons.

REFERENCES

African Development Bank (AfDB) (2014). “Comparing the Real Size of African Economies: Highlights of the Main Findings of the 2011 Round of the International Comparison Program in Africa”. African Development Bank (AfDB) (2009). Main Report on the Comparative Outputs, Incomes and Price Levels in African Countries. Final Results of the 2005 Round of International Comparison Program for Africa. ADB (Asian Development Bank) (2007). Purchasing Power Parities and Real Expenditures: 2005 International Comparison Program in Asia and the Pacific. Manila: Asian Development Bank. http://www. adb.org/Documents/ Reports/ICP-Purchasing-Power-Expenditures/PPP-Real-Expenditures.pdf. Asian Development Bank (ADB) (2012). 2009 Purchasing Power Parity Update for Selected Economies in Asia and the Pacific, a Research Study. Balassa, B. (1964). “The Purchasing Power Parity Doctrine: A Reappraisal,” Journal of Political Economy, vol. 6, no. 72, pp.584-96.

The African Statistical Journal, Volume 20, February 2018

127

Rees Mpofu

Commission of the European Communities, International Monetary Fund, Organisation for Economic Co-operation and Development, United Nations, and World Bank. (2008). System of National Accounts 2008. UN: pre-edit version of Volume 1, approved by the Bureau of the UN Statistical Commission, August 2008. http://unstats.un.org/unsd/nationalaccount/sna.asp. Common Market for Eastern and Southern Africa (COMESA) (2011). Harmonized Consumer Price Indices Regulations (Various Harmonization Topics- Stage 2). Common Market for Eastern and Southern Africa (COMESA) (2010). Harmonized Consumer Price Indices Regulations (Various Harmonization Topics-Stage 1). Dalgaard E. and H. Sorensen (2002, “Consistency between PPP Benchamrks and National Price and Volumes”, Paper presented at the 27th General Conference of the Association for Research and Wealth, SwedenAugust 2002. Diewert, W. E. (2005). “Weighted Country Product Dummy Variable Regressions and Index Number Formulae: Review of Income and Wealth,” no. 51, pp. 561-570. Eurostat-OECD. (2013). Compendium of HICP reference documents, Paris: OECD. Eurostat-OECD (2012). Methodological Manual on Compiling Purchasing Power Parities, Paris: OECD. Hill, R.J and T.P Hill (2009), “Recent Developments in the International Comparison of Prices and Real Output”, Macroeconomic Dynamics, 13, Supplement, pp. 194 – 217. Hill, T. P. (1982). Multilateral Measurements of Purchasing Power and Real Gross Domestic Product (GDP). Luxembourg: Eurostat. International Labour Organization (ILO), Eurostat, IMF, OECD, World Bank, and the UN. 2004. Consumer Price Index manual: Theory and practice, ed. Peter Hill. Geneva: International Labour Organization. Available at: http:/ / www.ilo.org/ public/ english/ bureau/ stat/ guides/ cpi/ index.htm.

128

Journal africain de statistiques, numéro 20, février 2018

3. Computing Consumer Price Indices and Purchasing Power Parities: A Special case for Africa

Gilbert, M., and I. B. Kravis (1954), An International Comparison of National Products and the Purchasing Power Parities of Currencies. Paris: Organisation for European Economic Co-operation. Ryten, J. (1999). “Report of the Consultant on the Evaluation of the International Comparison Programme”. UN: New York. http://unstats.org/ unsd/statcom/doc99/8-e.pdf Summers, R. (1973). “International Price Comparisons Using Incomplete Data.” Review of Income and Wealth no. 19, pp. 1–6. United Nations (1992), Handbook of the International Comparison Programme, Department of Economic and Social Development Statistical Division/Studies in methods. United Nations (2009). Practical Guide to Producing Consumer Price Indices.. United Nations, New York and Geneva. World Bank (2014), Purchasing Power Parities and the Real Size of World Economies: A Comprehensive Report of the 2011 International Comparison Program. Washington, DC.: The World Bank. World Bank (2013), Measuring the Real Size of the World Economy: The Framework, Methodology, and Results of the International Comparison Program. Washington, DC.: The World Bank. World Bank: (2014). Purchasing Power Parities and the Real Size of World Economies: A Comprehensive Report of the 2011 International Comparison Program. Washington, DC.: The World Bank World Bank (2008). Global Purchasing Power Parities and Real Expenditures, 2005 International Comparison Program. Washington, DC.: The World Bank.

The African Statistical Journal, Volume 20, February 2018

129

Rees Mpofu

APPENDIX

To complement the practical example provided in the main paper, we present another computation characterised by missing prices from all the countries. Done repeatedly, these examples can make a change over time. The characteristic feature of the examples is the use of limited data; it enables a better understanding of steps and efficient application of principles. For the example below, not all countries priced the requested ICP products. We use the CPD to calculate the basic PPPs. The process requires the preparation of Table 10 and running the OLS regression. PPPs are exponents of country coefficients. The derivation of other coefficients utilizes transitivity and country reversal (reciprocals) properties. Table 9. Basic Data on prices from three countries for five products. Country A

Country B

Country C

Product 1

10.00

40.00

100.00

Product 2

12.00

16.00

XXX

Product 3

15.00

30.00

Product 4

25.00

XXX

100.00

Product 5

XXX

20.00

70.00

XXX = missing average price: Source: United Nations Handbook, 1992.

130

Journal africain de statistiques, numĂŠro 20, fĂŠvrier 2018

3. Computing Consumer Price Indices and Purchasing Power Parities: A Special case for Africa

There are five products priced across three countries. Country A did not price product 5, country B did not price product 4, and Country C did not price product 2. We cannot use the Jevons-GEKS straight away. Instead, we use the CPD, the process starts with the preparation of a table of dummy variables. Table 10. CPD layout based on data from Table 10. Price

InP

2.30

2.48

2.71

3.22

3.69

2.77

2.71

3.00

100

4.61

3.40

100

4.61

4.25

Source: Prepared using Table 9

The African Statistical Journal, Volume 20, February 2018

131

Rees Mpofu

Table 11. The CPD Results Regression Statistics Multiple R

0.997

R. Squared

0.994

Adjusted R Squared

0.786

Standard Error

0.417

R. Squared Observations Squared Adjusted RR.Squared ANOVA Adjusted R Squared Standard Error Standard Error Regression Observations Observations Residual ANOVA ANOVA Total Regression Regression Residual Residual Total Total X.Variable 1

X.Variable X.Variable 12 1 X.Variable X.Variable 2 X.Variable 3 2 X.Variable X.Variable 3 X.Variable X.Variable 44 3 X.Variable X.Variable X.Variable 55 4 X.Variable X.Variable6 5 X.Variable X.Variable 6 X.Variable7 6 X.Variable X.Variable 7 X.Variable 7

0.994 12 0.994 0.786 SS 0.786df 0.417 0.417 7 137.9375 12 12 0.8699 df 5 SS df SS 138.8074 7 12 137.9375 7 137.9375 5 0.8699 Coeff SE 0.8699 12 5 138.8074 12 138.8074 Coeff SE 0.4974 0.3153 Coeff SE 0.4974 0.3153 1.4721 0.3153 0.4974 0.3153 1.4721 0.3153 2.8757 0.3019 1.4721 0.3153 2.8757 0.3019 2.8757 0.3019 2.3801 0.3344 2.3801 0.3344 2.3801 0.3344 2.2826 0.3019 2.2826 0.3019 2.2826 0.3019 3.1760 0.3344 3.1760 0.3344 3.1760 0.3344 2.6374 0.4019 2.6374 0.4019 2.6374 0.4019

Source: Derived from Table 10 Source:Derived Derived from 10 10 Source: fromTable Table

Sign F

19.71

113.263

0.0002

0.174 MS

F MS 19.71 113.263 F 19.71 113.263 t-Stat0.174 P-value Lower 0.174 95%

1.577t-Stat 0.17554 P-value -0.3132

t-Stat 4.6691.577 0.00549 1.577 4.669 9.526 4.669 0.00022 9.526 9.526 7.1177.117 0.00085 7.117 7.5617.561 0.00064 7.561 9.4979.497 0.00022 9.497 6.562 6.562 0.00123 6.562

P-value 0.17554 0.6616 0.17554 0.00549 2.0997 0.00549 0.00022 0.00022 1.5204 0.00085 0.00085 1.5066 0.00064 0.00064 0.00022 2.3163 0.00022 0.00123 1.6042 0.00123

Sign F Sign F 0.0002 0.0002

Upper 95%

Lower 95% 1.3079

Lower 95% 2.2826-0.3132 -0.3132 0.6616 3.6517 0.6616 2.0997 2.0997 3.23981.5204 1.5204 3.05861.5066 1.5066 4.03572.3163 2.3163 3.67061.6042 1.6042

Derivation of PPPs from country coefficients using transitivity and country Derivation of PPPs from country coefficients using transitivity and country reversal tests. Derivation reversal tests. of PPPs from country coefficients using transitivity and country reversal tests. The exponents of the first two coefficients from the regression output yield The exponents of the first two coefficients from the regression output yield TheB’s exponents the first coefficients the regression output country and C’sofPPPs. Thetwo first country isfrom not included as one of theyield country B’sB’s andand C’sC’s PPPs. The first country is isnot included asasone ofofthe country PPPs. The first country not included one the explanatory variables in the regression process as a matter of procedure. The explanatory variables asaamatter matterofofprocedure. procedure. The explanatory variablesininthe theregression regression process process as The specific PPP for the base country is 1.000. specific PPP for the base country is 1.000. specific PPP for the base country is 1.000. • • Procedurally country, the thecomputation computationproceeds proceedsasas follows: Procedurallyfor forthe the first first country, follows: • Procedurally for0.0000 the first country, the computation proceeds as follows: RRR A = 2.71828 0.0000= 1.00 (by definition) (by definition) definition) RRR A = 2.71828 = 1.00 (by A

• • For Forthe thesecond secondcountry, country,ititisis RRR B = 2.71828 0.4974 0.4974= 1.64 • For the second country, it is RRR = 1.64 B = 2.71828 A

•

For the third country it is RRRC = 2.718281.4721 = 4.36 1.4721 • = 2africain .71828 = 4.numéro 36 20, février 2018 132 For the third country it is RRRC Journal de statistiques, A

Uppe Upp

= 1.00 (by definition)

RRR A = 2.71828 A

3. Computing Consumer Price Indices and Purchasing Power Parities: A Special case for Africa

•

For the second country, it is RRR B = 2.71828 0.4974 = 1.64 A

• For • For the third country it is RRRC = 2.718281.4721 = 4.36 A

three PPPs country A as basecountry; country;the theother othertwo twocountries countriesare TheThe three PPPs useuse country A as thethe base are comparators. comparators. The derivation of other PPPs to fill the matrix utilizes knowledge on the The derivation of other PPPs and to fill the matrix utilizes on the application country reversal tests. knowledge The derivation of other utilizes knowledge The derivation of other PPPs totofill fill the matrix utilizes knowledge on the The derivationof oftransitivity otherPPPs PPPsto fillthe thematrix matrix utilizes knowledgeon onthe the The derivation of other PPPs to fill the matrix utilizes knowledge on the application of transitivity and country reversal tests. 35 application of transitivity and country reversal tests. application of transitivity and country reversal tests. applicationofoftransitivity transitivityand andcountry countryreversal reversaltests. tests. application   ΡΡΡΑ ΡΡΡΑ ΡΡΡΑ  ùù C  éé  Α Β éé A RRR RRR úúùùù  êêéRRR A A RRR RRR RRR ΡΡΡ ΡΡΡ RRR RRR RRR ΡΡΡ RRR RRR RRR RRR RRR RRR A A A ê Β Β A A A ê A A ê êê  AAAA Α BBAB Β CCACúAúúúúCB  êê A A RRR B B RRR CúCúúú êêêRRR B B B RRR RRR RRR ΡΡΡ ΡΡΡ ΡΡΡ RRR RRR RRR RRR RRR RRR RRR BB C BB C BBúúúC  êRRR ê êêêRRR AB B BB B CBúBú  úúúúC  B C êê  AAAA RRR Α Β BBB RRR CCC êêêRRR C C C úúú RRR RRR RRR ê RRR RRR RRR êRRR RRR RRR RRR RRR CC CC CCú Cú ú AC C BC C CCû ëêêëêêRRR úûûúúûûú B C BBB CCC ëêëëê AAAA The above PPPs correspond to the first column of the above general maThe above PPPs correspond the first column the above general matrix. trix. Once completed, theto matrix is as shownof below. The computation of Once The above PPPs correspond to column of general Once The above PPPs correspond to the first column of the above general matrix. Once The above PPPs correspond tothe thefirst first column ofthe theabove above generalmatrix. matrix. Once The above PPPs correspond to the first column of the above general matrix. Once completed, the matrix is as shown below. The computation of other elements is other elements is exhausted by exploiting transitivity and country reversal completed, the completed, the matrix as shown below. The computation of other elements completed, thematrix matrixis asshown shownbelow. below.The Thecomputation computationof ofother otherelements elementsis completed, the matrix isisisas as shown below. The computation of other elements isisis properties. exhausted by exploiting transitivity and country reversal properties. exhausted by exploiting transitivity and country reversal properties. exhausted by exploiting transitivity and country reversal properties. exhaustedby byexploiting exploitingtransitivity transitivity andcountry countryreversal reversalproperties. properties. exhausted and 1.4721 •• Given ; we compute RRR = 2 . 71828 = 4 . 36 111..4721 .4721 1 . 4721 C 4721 we compute Given we compute • Given ;;;;we compute RRR .71828 71828 .36 36 GivenRRR ;we we compute ••• Given Given compute RRR .71828 ====444.4.36 .36 RRR ===222.2.71828 CC = AC C AAAA

1 1 using the country reversal test RRR == 1111 === 1111 =0.23; A using the country reversal =0.23; using the country reversal test RRR RRR ==RRRC ===4.36 =0.23; =0.23; using the country reversal = 0.23; using the country reversaltest test ΡΡΡ =0.23; using the country reversal test RRR RRR AA = CA A RRR 4 . 36 4 . 36 RRR ΡΡΡ 4 . 36 RRR 4 . 36 RRR C CC C C CCC A A ΑAA A

0.4974 •• Given ; we compute RRR B == 22..71828 ==11..64 000..4974 0.4974 .4974 4974 compute Given we compute • RRR ==222.71828 .71828 71828 .64 64 GivenRRR ;we wecompute compute ••• Given Given we compute RRR . 71828 .64;;;we RRR ===111.64 BBBB= A AAAA

11 1 11 = 1111 = 0.61 ;; using the country reversal test. RRR = 1 A using the country reversal test. using the country reversal test. RRR RRR .61 61 ;using usingthe thecountry countryreversal reversal test. the country reversal test. RRR .61 ;;;using test. RRR ===RRR B ====1.64 ====000.0.61 AA = BA A RRR 1 . 64 RRR 1 . 64 RRR 1 . 64 RRRABBBB 1.64 BBBB A AAA

RRR C RRR RRR RRR RRR CC .36 AC C 4 .36 36 ••• We compute using transitivity. RRR = = 44..36 .36= 2.65; We compute 2.65; using transitivity. A C AAA== 44 We compute 2.65; using transitivity. = RRR = Wecompute computeRRR 2.65;using usingtransitivity. transitivity. ••• We We compute 2.65; using transitivity. RRR = = RRRBCCCC= RRR B = 1.64 ====2.65; RRR 1 . 64 RRR 1 . 64 RRR 1 . 64 RRR 1 . 64 B B B B BB AB B A AAA

1 1 We compute using the country reversal test. RRR == 1111 == 1111 == 0.38; B We compute We compute 0.38; using the country reversal test. RRR RRR ==RRR ===2.65 ===0.38; Wecompute computeRRR 0.38;using usingthe thecountry countryreversal reversaltest. test. We 0.38; using the country reversal test. RRR BB = C CB B RRR 2 . 65 RRR 2 . 65 RRR 2 . 65 RRR 2 . 65 C CC 20, February 2018 The African StatisticalCCCJournal, Volume 133 C C B BBBB

RRRC We compute RRRC =

•

RRR B

4.36 Rees Mpofu = 2.65; using transitivity. 1.64

1 1 00 0 23 0.38; using usingthe thecountry countryreversal reversal test. éé1 We compute compute test. 1..We 00 0..61 61 0 0..RRR 23ùù B = RRR = 2.65 = 0.38; êê1.64 1.00 0.38úú C; C êê1.64 1.00 0.38úú ; B ú êëê4 1 . 00 0 . 61 0 . 23 . 36 2 . 65 1 . 00 ù é û 2.65 1.00 ú ë4.36 ê1.64 1.00 0û.38ú ; ú ê úû the ê The specific elements in 4 . 36 2 . 65 1 . 00 ë The specific elements in the above above matrix matrix correspond correspond to to elements elements in in the the general general matrix above. matrix above. 36

The specific specific elements elementsin in the the above abovematrix matrixcorrespond correspondto to elements elementsin in the the general The matrix above reduces to a vector of standardized PPPs used general matrix above. above. Thematrix matrix above reduces to a vector of standardized PPPs used with with corresponding weights for the same basic heading during the next level corresponding weights for to thea vector same of basic heading PPPs during thewith next level of of The matrix above reduces standardized used corcalculations. The next level of aggregation is outside the scope of this paper. The matrix abovelevel a vector ofduring standardized PPPs calculations. The next of aggregation is outside the scope of this paper. responding weights forreduces the sametobasic heading the next level of used with corresponding the same is basic heading during next level of calculations. Theweights next levelfor of aggregation outside the scope of thisthe paper. Standardization using the first column PPPs in the above matrix calculations. The aggregation scope of thisgives paper.the Standardization usingnext thelevel firstof column PPPsis outside in the the above matrix gives the following results. The general case for the first element is as follows: Subsequent Standardization using the first in the above matrix gives following results. The general casecolumn for the PPPs first element is as follows: Subsequent elements are dividing mean of PPPs into the following results.by The generalthe casegeometric for the first element isrelevant as follows: elements are computed computed by dividing geometric of the the relevant PPPs into the Standardization using theexhaustion. first the column PPPsmean in the above matrix gives one of them in sequence until Subsequent elements are computed by dividing the geometric mean of the onefollowing of them inresults. sequence until exhaustion. Theof general case for theuntil first exhaustion. element is as follows: Subsequent relevant PPPs into one them in sequence elements are computed by dividing the geometric mean of the relevant PPPs into one of them in sequence until exhaustion. RRR A Α RRRΡΡΡ A A Α = A = element = element element RRR ´ RRR ´ RRR ΡΡΡ × ΡΡΡ × ΡΡΡ C A ´Α B ´Β 3 RRR C 3 RRR RRR RRR C A B 3 A A AΑ A Α Α A A A A = element 3 RRR A ´ RRR B ´ RRR C • First element of theA standardization vector: •• First element of A element A First of the the standardization standardization vector: vector: 1.000 000 First1.element of the vector: = ..518676 =0 0standardization 518676 1 . 000 ´ 1 . 64 ´ 4 . 36 1.000 ´1.64 ´ 4.36 1.000 = 0.518676 3 1.000 ´1.64 ´ 4.36 Second element of the standardized vector: Second Second element element of of the the standardized standardized vector: vector: 1 . 64 1.64 = =0 0..852894 852894 3 1 . 000 ´ ´ 3 • 1.Second element 000 ´1 1..64 64 ´4 4..36 36 of the standardized vector: 1.64 = 0.852894 3 1.000 ´1.64 ´ 4.36 Third Third element element of of the the standardization standardization vector: vector:

•

• •• ••

3 3

• 134

Third element of the standardization vector: 37

37 africain de statistiques, numéro 20, février 2018 Journal

3. Computing Consumer Price Indices and Purchasing Power Parities: A Special case for Africa

•

Third element of the standardization vector:

4.36 = 2.260521 1.000 ´1.64 ´ 4.36

Table 12. A vector of standardized PPPs at the level of basic heading Country A

Country B

Country C

Table 12. A vector0.518676 of standardized PPPs0.852894 at the level of basic heading. 2.260521 A Source: DerivedCountry by standardization

0.518676 Source: Derived by standardization

Country B 0.852894

Country C 2.260521

Standardized PPPs are weighted using expenditure data from national acStandardized PPPs weighted are weighted counts to derive PPPs.using expenditure data from national accounts to derive weighted PPPs.

The African Statistical Journal, Volume 20, February 2018

135

4. Partnership, processes and possibilities: the South African experience of integrating Purchasing Power Parity (PPP) and Consumer Price Index (CPI) work Patrick Kelly1, Lekau Ranoto2

Abstract Collection of prices for the compilation of PPPs is often perceived to be timeconsuming and of little immediate value to national statistics offices. Countries may find the integration of survey activities for both the CPI and PPPs a challenge. The fact that international organisations fund the PPP data collection compounds the problem. South Africa has been the base country for ICP Africa for two rounds. The paper compares the concepts, definitions, price collection procedures and quality assurance processes involved in the compilation of the CPI and PPPs. The paper describes how Statistics South Africa has managed to mainstream PPP basic data collection into regular pricing programs at minimum cost. Indeed, rather than being a burden, the integration process has presented opportunities for improving the CPI and providing additional information, such as average prices, to users. Key words: Purchasing Power Parity (PPP), Consumer Price Index (CPI), Products, Items, and Outlets. Abstrait La perception des prix pour la compilation des PPA est souvent perçue comme longue et peu utile pour les bureaux de statistiques nationaux. Les pays peuvent trouver difficile l’intégration des activités d’enquête pour l’IPC et les PPA. Le fait que les organisations internationales financent la collecte de données sur les PPA aggrave le problème. L’Afrique du Sud a été le pays de base du PCI Afrique pour deux tours. Le document compare les concepts, les définitions, les procédures de collecte des prix et les processus d’assurance qualité impliqués dans la compilation de l’IPC et du PPA. Le document décrit comment Statistics South Africa a réussi à intégrer la collecte de données de base PPA dans des programmes régulières à un coût minime. En effet, plutôt que d’être un fardeau, le processus d’intégration a présenté des opportunités pour améliorer l’IPC et fournir des information supplémentaires, telles que les prix moyens, aux utilisateurs. Mots clés : Parité de pouvoir d’achat (PPA), Indice des prix à la consommation (IPC), Produits, articles et points de vente. 1 Executive Manager, Statistics South Africa; patrickke@stassa.gov.za 2 Lekau Ranoto, Statistics South Africa; lekaur@statssa.gov.za

136

Journal africain de statistiques, numéro 20, février 2018

4. Partnership, processes and possibilities: the South African experience of integrating Purchasing Power Parity (PPP) and Consumer Price Index (CPI) work

â&#x20AC;&#x153;The similarity of the basic data collection process for the ICP and CPI should be used to optimize the use of resources. To that end, the ICP should use the statistical infrastructure already in place in a country to the maximum extent possible, and helps improve on it wherever possible. This translates in the ICP using the same outlets, the same price collectors and processing procedures, as long as they comply with ICP requirementsâ&#x20AC;? (African Development Bank, 2015a). 1. INTRODUCTION

The significant improvement in the coverage and availability of PPP adjusted Gross Domestic Product (GDP) and its principal components since 2005 has enhanced the measurement of real sizes of economies. While the World Bank compiles and publishes PPPs, national statistics offices facilitate this by collecting and assembling the relevant price and expenditure data. The collection of prices for selected ICP products is typically the responsibility of CPI units in national statistics offices. The price collection process and related activities can be viewed as some form of inconvenience and additional work or as an opportunity for integrating the same with regularly pricing programs and as a learning process. Although the objectives of CPIs and PPPs are characterised by differences and similarities leading to methodological variations, the common classification and general price collection makes their harmonisation possible and useful. In South Africa, the statistics office uses its own staff, processing systems and resources to collect ICP price data. In so doing, it identified and adopted elements of the ICP processes to improve the national CPI. These improvements will further streamline the collection of ICP price data in future rounds. 2. BACKGROUND

Price data collection methods are dependent on the index computation methods at lower levels for both the CPI and PPPs. While the prices fed into a CPI and PPP are typically the same, their intended uses indicate differentiated requirements and hence data collection requirements. For this reason, it is useful to compare the definitions and purposes of a CPI and PPP.

The African Statistical Journal, Volume 20, February 2018

137

Patrick Kelly and Lekau Ranoto

CPIs are temporal price indices, mainly used to measure price changes over time. The CPIs utilize a representative set of products purchased by a reference population over time. The CPIs primarily measure inflation in an economy, or track changes in the cost of living. The relative importance of these two objectives guides price statisticians in their choice of methodology. In practice, many national CPIs, including South Africa’s, aim at measuring inflation with a view of informing monetary policy design. The same CPIs also serve, among other purposes, as an escalator of contracts, wages and benefits (Statistics South Africa, 2013). PPPs, on the other hand, are spatial or multilateral price indices —they show the ratio of the prices in national currencies of the same good or service in different economies. PPPs answer a simple question: For every Rand spent in South Africa on a basket of goods and services of fixed quality and quantity, how much is required to do the same in Rwandan local currency (Rwandese Francs) and across all other comparator countries. PPPs primarily enable the international comparison of national economies and their growth. They allow cross-country measurement of real GDP per capita and GDP (Word Bank, 2013). The first ICP took place in 1970, covering only 10 countries. More recent ICP rounds have taken place in 2005, 2009 (interim round), 2011 and 2015 (interim round). The ICP has been coordinated at a global level by the World Bank. The surveys are organised by region and coordinated by regional implementing agencies. The rationale for regional surveys is that the products are generally homogeneous within the regions, and the expenditure patterns are likely to be similar. There are also operational advantages in having the ICP carried out by agencies that are in close proximity to the economies they are coordinating. A relatively large African representation characterised the 2005 and 2011 rounds of the ICP. Under the coordination of the African Development Bank (AfDB) the number of participating African countries stood at 48 and 50 for the 2005 and 2011 rounds, respectively. The African countries were distributed across four sub-regional organisations (ECOWAS, COMESA, SADC and AFRISTAT) for administrative purposes. The 2005 ICP round marked South Africa’s first participation in the global statistical initiative. Participation in this round was limited, as it coincided with the substantial overhaul of the national CPI. Consequently, certain ICP methods, such as the Structured Product Description (SPD), were utilized as part of the across the board revamping of the CPI data collection methods.

138

Journal africain de statistiques, numéro 20, février 2018

4. Partnership, processes and possibilities: the South African experience of integrating Purchasing Power Parity (PPP) and Consumer Price Index (CPI) work

During the 2011 round, South Africa was the base country for the African region. While the choice of the base does not affect the results, as the PPPs are base country invariant, the choice of South African currency reflected its extensive use across currency markets. Furthermore, South Africa is the most industrialised country on the continent, with a wide variety of consumer goods and services across various outlets. These features enhance comparability with other regional base countries. The AfDB funded ICP activities in African countries, however some countries viewed the same activities as a separate work program. With the funding by the Bank, the participation rate was generally high. The funding issue led to the recognition of the ICP as an additional effort, thereby reducing the incentive to mainstream the same activities into regular pricing programs of the national statistical systems. South Africa used its own permanent staff for ICP operations and did not request or receive AfDB funding. Improved integration of the ICP work with the regular CPI work was therefore necessary to ensure the success of both operations. The AfDB provided extensive support to national statistics offices, including the compilation of manuals for data collection forms, data entry sheets, data validation tools and continuous training. 3.

PRODUCT SELECTION

The ICP faces two competing challenges in product selection. First, the ICP products are supposed to be comparable across countries and be representative within the same countries. Second, it is necessary to link different ICP regions in the process of calibrating ICP results from individual countries to derive global PPPs - denominated in American dollars. The South African ICP basket was determined by first checking the common products in the CPI and the given ICP baskets. The relevant data was extracted from the CPI database while adhering to set parameters or guidelines (e.g. outlet type, units of measurement, brand, etc.) for all the products. This process ensures that the exact ICP product are linked with specific CPI products, where applicable. The other products might be available in the country but not necessarily in the CPI basket. Simultaneous price collection for such products with CPI products ensures there is no additional cost. It was possible to accommodate this additional work in the schedules of the permanent CPI price collectors.

The African Statistical Journal, Volume 20, February 2018

139

Patrick Kelly and Lekau Ranoto

In the ICP 2011 round, 1,032 household products were identified for possible price collection. South Africa collected 585 products, of which 369 were common products between the CPI and ICP baskets. The remaining 216 were collected as a dedicated set for ICP purposes beyond the usually CPI activities. Despite the similarities of items across the ICP and CPI at any particular point in time, item selection for the two statistical endeavours is not necessarily the same. For household goods and services, the use of a common classification, Classification of Individual Consumption According to Purpose (COICOP) eases applied work. Because the CPI monitors price changes of the items over time, it is not concerned with the brand, size, etc. of each product under consideration. In other words, there might be different items at different price levels in the CPI basket, but that does not induce any bias in the elementary index compilation. For example, there are different varieties of rice represented by different brands, sizes and price levels. Each month the CPI computes price changes of each item. The suitable index formula is used to obtain an aggregate price change at the level of elementary aggregates. The ICP, on the other hand, focuses on average price levels, not average price changes. Any variation from a given set of specifications distorts the results. It is for this reason that the products within the ICP basket have tight specifications aimed at ensuring the comparison of like with like across the participating countries. Because of the tight product specification for the ICP, it is possible to get a number of ICP products under a single CPI product description. For example, although Basmati, Jasmine and white rice are separate ICP products, they belong to the same CPI category called rice. Consequently, a manual matching process is undertaken, during the first month of the ICP round price collection, to identify and link relevant items with items priced as part of monthly CPI consumer price surveys. Owing to the greater flexibility of product choice and specifications for the CPI, there is a free text field for brand and product names. This gives rise to numerous spelling errors, and hampers deeper analysis of price behaviour of identical products. However, the rigid ICP approach illustrates the benefits of ensuring standardised capturing of names and product sizes. South Africa has revised the CPI forms and capturing system to provide for drop

140

Journal africain de statistiques, numĂŠro 20, fĂŠvrier 2018

4. Partnership, processes and possibilities: the South African experience of integrating Purchasing Power Parity (PPP) and Consumer Price Index (CPI) work

down lists following the ICP style. Such improvements will enhance ICP validation in future rounds. 4.

OUTLET SELECTION

Together with the price and the product specifications, the outlet type is one of the fundamental elements of ICP price collection. To maintain comparability across countries, price collection takes place from requested outlet types. In the context of the CPI, the outlet is not part of product specification, and purposive sampling guides the outlets selection. It means that any outlet that stocks relevant CPI products stands an equal chance for selection. The restriction on the type of outlet in the case of the ICP limits the use and number of products from the CPI. Even if the products are available in the CPI basket or in the country, they might not be included in the country’s ICP basket because of an outlet mismatch. For example, during the 2011 round, informal markets were potential outlets, yet the CPI does not have such outlets. Another example is the requirement for products from ‘Butchery without refrigerator’, which do not exist in the country. It was for this reason that South Africa decided to include a field for ‘type of outlet’ in both the household expenditure survey and the CPI collection forms. This led to the introduction of categories of chain store, independent stores and informal markets. Although these outlet types are not identical to the ICP (supermarket, specialised stores, department stores, wholesale stores, discount stores etc.), they do provide a step towards better alignment and allow for the analysis of pricing behaviour across different outlet types. There is still opportunity for expanding the outlet types for CPI price collection to ensure that price data covers as many outlet types as possible across the country. 5.

PRICE COLLECTION

Price collection entails fieldworkers visiting selected outlets each month to record prices that the reference population actual pay. Field based data collectors do price collection mainly for goods, while head office staff collect prices for services.

The African Statistical Journal, Volume 20, February 2018

141

Patrick Kelly and Lekau Ranoto

Below are the three main survey forms used in price collection for the CPI a. The Outlet Cover Page captures data on the outlet status, address, the contact person for collection and quality control. b. The SPD initiates a new product selected for pricing following given specifications. c. The pricing form enables time to time price collection for given products selected for regular pricing. These forms date back to the general revision of the CPI in 2005. The use of the SPD followed its adoption from the ICP 2005 round. It has been extremely valuable for product initiation for the CPI. A variation of the SPD was designed and tailor made for the producer price index (PPI). Even though the ICP uses spreadsheets for collection, the fields in the spreadsheet correspond to the survey forms designed for the CPI. For the ICP products that are not necessarily part of the CPI, the SPD forms are prepared, dispatched, and used for price collection in similar outlets used for the CPI purposes. Suppose the same ICP products are not necessarily available in the usual CPI outlets, recruitment of relevant outlets takes place. Fieldworkers apply the same CPI publicity approach as part of regular price collection procedures when approaching new outlets. Integration between CPI and ICP is cost-effective since there is no need to employ extra data collectors and build new systems. However, data collectors need thorough training to be able to appreciate the typical tight product specifications associated with the ICP, relative to the generally loose product specifications for the CPIs. Poor understanding on the part of fieldworkers causes delays in the field, may compromise data quality as well as intermediate and final results. 6.

DATA PROCESSING

After data collection, different regions dispatch both CPI and ICP forms to head office for data capture and subsequent activities. Initially, all the data is captured into the CPI database, where the editing process follows the CPI methodology. At the end of editing, a clean database is ready for any type of analysis. Although the ICP and CPI use different product codes, a correspondence table enables the linkage between specific CPI and ICP items, the extraction uses a custom-built SAS program. The ICP comes with a specific data-entry sheet for collected prices. This procedure aims

142

Journal africain de statistiques, numĂŠro 20, fĂŠvrier 2018

4. Partnership, processes and possibilities: the South African experience of integrating Purchasing Power Parity (PPP) and Consumer Price Index (CPI) work

at ensuring a homogeneous structure of the datasets across all countries. Following data capturing, the data are merged, using an additional tool, to produce a country data file. To avoid creating different databases for CPI and ICP, South Africa does not use the standard ICP data entry sheet for data capturing. Instead, all the relevant data is captured into the CPI database, and entries required for the ICP are extracted to the ICP data-entry sheet. That way, only data that pass CPI quality-validation processes are exported into the ICP data-entry sheet. This process results in speedier validation of ICP data. The ICP data entry sheet provided useful practical lessons for the CPI processes. Since the standard ICP data capture sheet contains protected cells to avoid mistakes, and guides users during data entry sessions, its desirable functions have since been adopted for regular CPI processes, with dropdown lists of options. The approach reduces errors. 7.

DATA VALIDATION / QUALITY ASSURANCE

The quality of the basic data is a vital prerequisite for good results and subsequent statistical analysis. The CPI and ICP activities use different data validation methods. The CPI requires comparison of prices of like with like items from month to month. Data validation attempts to ascertain that the attributes of products priced each month are the same for purposes of consistency for the ICP, the identity of products is very detailed to ensure comparable products are identified and consistently priced across countries. Consequently, the ICP data validation process aims to ensure that the prices collected by each of the countries are indeed comparable. Quality assurance for the CPI data is done by editing data in the databases. Data editing, aimed at detecting and correcting errors in data, consists of four steps: a. Validation checks to ensure that the correct item status codes were used during data capturing and allocation of missing unit codes; b. Logical edits to ensure that the current and previous monthâ&#x20AC;&#x2122;s data do not have contradictory values; c. Range edits to identify whether the data item value falls inside a predetermined acceptable range; and d. Final checks of the accuracy of item codes.

The African Statistical Journal, Volume 20, February 2018

143

Patrick Kelly and Lekau Ranoto

Upon the completion of the editing process, the data are exported to a SAS (Statistical Analysis System) software application for compilation. The SEMPER software, primarily designed for ICP purposes, validates and calculates average price levels for the ICP-related products. The SEMPER analyzes the observed prices and provides descriptive statistics (measures of central tendency and dispersion) in a few seconds enabling comprehensive intra data validation. A junior statistician requires less than three days each month to process and validate all related data. The SEMPER performs the following functions: a. Conformity checking: to code the product name, quantity, unit of measurement and other characteristics related to the description of the products. The software rejects any irrelevant products. b. Factor conversion: to convert the unit of measurement (e.g. grams, millilitres) from that collected to what is desirable for ICP comparison. c. Recalculation of prices (range edits): to identify outliers. The country data file is used to correct errors detected by SEMPER. The SEMPER re-run is iterative until a consensus on data consistency is achieved. Participating countries submit validated Country Data files to the AfDB for cross-country data validation and eventual computation of PPPs. The annual average prices and the expenditure weights from the Gross Domestic Product (GDP) provide basic data to compute PPPs at basic and higher levels. Following the introduction of ICP-inspired enhancements discussed above, South Africa started calculating averages prices from the CPI data, initially at national level (in 2000), and by province from 2010. The national average prices are not published, but they are available on request. There has been substantial demand from government departments for the same data for purposes of benchmarking of procurement activities. 8.

POSSIBILITIES FOR INTEGRATION AT A REGIONAL LEVEL

The AfDB, from time to time, invites representatives from participating African countries to data validation workshops to ensure cross country consistency. The data validation exercises provide an opportunity for continuous statistical capacity building on ICP as part of their core operations.

144

Journal africain de statistiques, numĂŠro 20, fĂŠvrier 2018

4. Partnership, processes and possibilities: the South African experience of integrating Purchasing Power Parity (PPP) and Consumer Price Index (CPI) work

Despite their close involvement in ICP data collection, national statistics offices play no role in the computation, analysis and dissemination of ICP results. Deeper involvement in presenting and explaining results may require additional training, but should lead to enhanced integration and commitment by countries. 9. CONCLUSION

In many countries there is general resistance integrating CPI and ICP, and the related changes that may come from that. However, the integration between the CPI and ICP in South Africa has proven to be a valuable practical exercise. A major benefit of this integration is ICP price collection at minimum cost with no additional resources. It has positively improved the CPI processing system, enhancing the quality and range of data collected during price collection. The ICP has further provided opportunities for research work such as calculating sub-national PPPs (Kgantsi, 2012). 10. REFERENCES

African Development Bank Group (2014). Comparing the Real Size of African Economies: Results of the 2011 International Comparison Program 2011 for Africa. Abidjan. Kgantsi, E. M. (2012). A comparative study of purchasing power parities for the food component using the consumer price index data in the South African provinces. Dissertation submitted in fulfilment of the requirements for the degree of Master of Science. University of the Witwatersrand, Johannesburg. Statistics South Africa (2013). The South African CPI Sources and Methods Manual. Pretoria. World Bank (2013). Measuring the Real Size of the World Economy: The Framework, Methodology, and Results of the International Comparison Programâ&#x20AC;&#x201D;ICP. Washington, DC.

The African Statistical Journal, Volume 20, February 2018

145

5. Counting What Counts: Africa’s Seminal Initiative on Governance, Peace and Security Statistics Marie Laberge,1 Yeo Dossina 2, François Roubaud 3

Abstract This paper documents the practical experience of eleven African national statistical offices that tested and eventually institutionalized a methodology for producing official harmonized statistics in governance, peace and security statistics between 2012 and 2017. This took place whilst the rest of the world was still debating the rationale for including this new domain in the next global development agenda. The paper documents Africa’s successful GPS-SHaSA experiment in the context of the continent’s long-standing desire to “achieve political sovereignty through data autonomy”. The paper also presents some strategic advantages of the GPS-SHaSA methodology, provides illustrations using selected targets of Africa’s Agenda 2063 and Sustainable Development Goal (SDG) 16 on how the four types of data generated by the methodology can inform policymaking. Finally, the paper identifies methodological, institutional, financial and communicational investments necessary for a sustainable GPS statistical production by NSOs in Africa and beyond. Key words: Measurement, Indicators, Harmonized, Sustainable Development Goals, Agenda 2063, Household Surveys. Abstrait Ce document présente l’expérience pratique de onze instituts nationaux de statistique africains qui ont testé et finalement institutionnalisé une méthodologie pour produire des statistiques harmonisées officielles dans les statistiques de gouvernance, paix et sécurité entre 2012 et 2017. Ceci a eu lieu alors que le reste du monde discutait encore pour inclure ce nouveau domaine dans le prochain agenda de développement mondial. Le document présente l’expérience réussie du GPS-SHaSA en Afrique dans le contexte de la volonté de longue date du continent de “parvenir à la souveraineté politique par l’autonomie des données”. Le document présente également certains avantages stratégiques de la méthodologie GPS-SHaSA, fournit des illustrations en utilisant des cibles sélectionnées de l’Agenda 2063 et de l’Objectif de développement durable (ODD) 16 sur la manière dont les quatre types de données générées par la méthodologie peuvent 1 Marie Laberge, French Institute of Research for Sustainable Development (IRD-DIAL), e-mail: marie.laberge.34@gmail.com. 2 Yeo Dossina, African Union Commission (AUC), Statistics Division, Addis Ababa, Ethiopia e-mail: DossinaY@africa-union.org. 3 François Roubaud, French Institute of Research for Sustainable Development (IRDDIAL), e-mail: Roubaud@dial.prd.fr

146

Journal africain de statistiques, numéro 20, février 2018

5. Counting what counts: Africa’s seminal initiative on governance, peace and security statistics

éclairer l’élaboration des politiques. Enfin, le document identifie les investissements méthodologiques, institutionnels, financiers et communicationnels nécessaires pour une production statistique GPS durable par les Instituts Nationaux de Statistique (INS) en Afrique et au-delà. Mots clés : Mesure, Indicateurs, Harmonisés, Objectifs de développement durable, Agenda 2063, Enquêtes auprès des ménages. 1.

AN AFRICAN PARADOX

Very few people are aware that some African statistical offices had long been producing official statistics on governance well before European statistical offices did. Madagascar first published a comprehensive set of governance statistics in 1995, followed by seven Francophone West African countries in the first half of 2000, several of which — notably Mali and Benin — have been doing periodic updates since then4. It was not until 2013 that Eurostat began experimenting with governancerelated questions in its core EU survey module on Income and Living Conditions (EU-SILC) applied by national statistical offices (NSOs) across the European Union5. The Quality of Life Expert Group, mandated in 2012 by EU Directors of Social Statistics to develop multi-dimensional measures of quality of life, concluded in its final report that with respect to “Governance and Basic Rights”, several topics continue to be “difficult for official statistics to approach” (European Union, 2017). While these topics — namely, satisfaction with public services, discrimination, and voice and accountability — are yet to feature in Eurostat’s survey modules, they are routinely measured by a dozen African statistical offices. They may well be measured continent-wide in the near future, as according to the African 4 The periodic repetition of governance surveys by some African countries is all the more striking because, according to the Mo Ibrahim Foundation, only half of Africa’s population live in a country that has conducted more than two comparable surveys in the past 10 years (Mo Ibrahim Foundation, 2016). 5 The EU Statistics on Income and Living Conditions (EU-SILC) survey, notably its ad hoc module 2013 on subjective well-being, included three survey questions on trust (in the legal system, the political system and the police), and the EU-SILC ad hoc module 2013 on social and cultural participation included one question on “active citizenship” (i.e. participation in activities of a political party or local interest group, participation in a public consultation, signing a petition, writing a letter to a politician or to the media, participation in a demonstration, etc.). (See http://ec.europa.eu/eurostat/web/incomeand-living-conditions/data/ad-hoc-modules).

The African Statistical Journal, Volume 20, February 2018

147

Marie Laberge, Yeo Dossina and Francois Rouband

Union’s second edition of the Strategy for the Harmonization of Statistics in Africa (AU, UNECA, AfDB, 2017). Africa, regarded by some as a continent doomed to perpetual crisis and bad governance, is the world leader in measuring progress on governance. This paper aims to shed light on this apparent paradox by showing that the adoption by African Heads of States in 2011 of an official commitment to harmonized official statistics on governance, peace and security (GPS) was the culmination of two decades of efforts to reclaim data sovereignty. We examine the motivations that led eleven African statistical offices, from 2012 to 2017, to pilot-test (and for some, to institutionalize) a methodology for producing official statistics in an emerging domain. Given that world leaders have adopted SDG 16 on Peaceful, Just and Inclusive Societies, a review of the SHaSA methodology on Governance, Peace and Security Statistics (GPS-SHaSA), developed by the community of African statisticians, is timely. Insights and lessons emerging from this African experience will benefit national statisticians worldwide, as preparations for reporting on Goal 16 take effect. Amongst several such insights, three are of particular relevance to ongoing efforts to establish a global monitoring mechanism on SDG 16. First, the conclusive results of the GPS-SHaSA pilot in Africa prove that nationally produced survey-based statistics on GPS that are comparable across countries are feasible. Second, the high diversity amongst participating countries — including post-conflict settings such as Mali’s, countries-in-crisis such as Burundi (at the time), “new” democracies such as Tunisia, and top-ranking democracies like Cape Verde — shows that NSOs in both transitional and consolidated democracies are politically, financially and methodologically capable, and willing, to produce GPS statistics. Third, the GPS-SHaSA dataset demonstrates the policy value of combining administrative and survey data sources. The first section of this article, which takes a historical perspective, puts into context Africa’s successful GPS-SHaSA experiment within the continent’s commitment to achieve political sovereignty through data autonomy. The second section presents some strategic advantages of the GPS-SHaSA methodology, as we illustrate how the four types of data generated by the GPS-SHaSA instruments can inform policymaking, using an example on the ‘free and fair elections’ target of Africa’s Agenda 2063, and on the ‘no discrimination’ target of the world’s 2030 Agenda for Sustainable

148

Journal africain de statistiques, numéro 20, février 2018

5. Counting what counts: Africa’s seminal initiative on governance, peace and security statistics

Development. The third section analyzes the political and institutional contexts of national pilots and the strategies applied by NSOs to secure the buy-in of their political leadership and to create broad-based demand for GPS statistics in their respective countries. The final section identifies methodological, institutional, financial and communicational investments necessary for the sustainable production of GPS statistics by NSOs. 2.

THE GPS-SHASA ORIGINS

Africa’s bold decision to embark on a continent-wide statistical program on governance, peace and security is best understood in the broader context of various commitments over the past fifteen years to reclaim sovereignty through data autonomy, especially in the highly strategic domain of governance and peace. 2.1

Inviting citizens’ feedback: The African Peer Review Mechanism (APRM) When the AU created the New Partnership for Africa’s Development (NEPAD) in 2002 as part of the African Renaissance initiative, it wanted to set a new stage for managing its own development agenda and for finding “African solutions to African problems”. To this end, a homegrown governance evaluation system was needed, which would allow countries to come together as equals, to engage in peer reviews and to share lessons learnt and best practices amongst themselves. Established in 2003 as a voluntary mechanism for African countries to selfassess on governance, the APRM was pioneering in at least two respects (APRM/NEPAD, 2003). First, the APRM’s self-assessment questionnaire was structured around adherence to a set of continental and regional standards and codes for democratic, economic and corporate governance. This was a bold assertion of African sovereignty at a time when a mushrooming industry of international rating and ranking indices were basing their assessment of individual countries’ performance on externally determined criteria of ‘good’ governance (Arndt et al., 2006). Second, the APRM’s invitation of people’s participation in the evaluation of governance was a daring experiment. When developing their self-assessment report, states had to gather the perspectives of a broad range of non-state actors, including ordinary citizens. While such efforts may have been imperfect, they were, nonetheless, “a powerful political and moral symbol” (Corrigan et al., 2017: 11) of Africa’s adhesion to the idea that governance

The African Statistical Journal, Volume 20, February 2018

149

Marie Laberge, Yeo Dossina and Francois Rouband

reform needed to be informed by the experience of ordinary citizens. In this regard, the APRM established an important precedent: it formalized public participation in governance assessment processes, and helped popularize the idea that without it, the legitimacy and credibility of governance evaluations may suffer. 2.2 Including GPS in official statistics: the SHaSA African Heads of States adopted the SHaSA in 2011 to accelerate the African integration agenda — a process which “requires quality statistics — statistics that are accurate, objective, timely, consistent, harmonized i.e. comparable across time and space, and produced efficiently and regularly” (AU et al., 2011). The SHaSA was first and foremost a response to African frustrations with the fact that “statistics are produced using methodologies that do not always reflect African realities […], partly because international statistical references and standards do not always take into account continental specificities (i.e. the nature of African economies, the cultural habits of local populations, etc.)” (AU et al., 2011). Underpinning the SHaSA was also the realization that sound national statistics reinforce a country’s sovereignty. This idea was subsequently championed by the Mo Ibrahim Foundation’s 2012 report, which lamented the paucity of African data and advocated for statistical autonomy within African countries, and affirms data autonomy as integral to political sovereignty.6 The SHaSA was also adopted at a time when international governance indices were harshly criticized for the subjectivity inherent in the selection and interpretation of the data used in their construction. Their lack of transparency and comparability over time, and their limited use for policymakers who struggled to find what action to take based on a single composite score, were also frown upon (Arndt et al., 2006). These well-known biases and limitations of international governance indicators made it imperative for national statistical systems to start generating their own data. As explained by Zakari Mwangi, Director-General of Kenya’s National Bureau of Statistics: “Governance in Kenya is being assessed by some twenty organizations — and not one of them is Kenyan! This proliferation of externally-led, uncoordinated data-collection drives not only marginalizes our national statistical agencies but also creates confusion by applying different methods to measure the same things” (Mwangi cited in UNDP, 2017). It is in this context that the methodological approach proposed by the SHaSA Group on GPS statistics was unanimously adopted by the Committee of Directors-General of African NSOs at its first annual meeting after the adoption of the SHaSA, 6 The Mo Ibrahim Foundation also started investing in nationally generated data sources for its index that year, in a partnership with the Afrobarometer and Global Integrity.

150

Journal africain de statistiques, numéro 20, février 2018

5. Counting what counts: Africa’s seminal initiative on governance, peace and security statistics

in 2012, in Yamoussoukro. Shortly after, twenty7 NSOs responded to the AU Statistics Division’s call for expressions of interest in the piloting of the GPS-SHaSA instruments. At their subsequent annual meeting in 2013 in Johannesburg, DirectorsGeneral formally requested the GPS-SHaSA Group to “secure funding for a regional project to support NSOs in the institutionalization of GPS data collection across the continent” (AUSTAT, 2013). On the eve of the adoption of Agenda 2063 — The Africa We Want, the continent’s new development framework for the next fifty years, the Directors-General also underlined the timeliness of the GPS-SHaSA initiative, and welcomed it as a direct contribution towards Agenda 2063’s vision of “a more united and strong Africa, […] a global power to reckon with […], speaking with one voice” (AU, 2013). Rooted in pan-Africanism and setting the road towards an “African Renaissance”, Agenda 2063 further legitimized the GPS-SHaSA endeavor, consolidated statisticians’ buy-in and triggered a race to the top in the tightly knit community. 2.3

Advocating for GPS data sovereignty globally: The African position on the post-2015 agenda African Heads of State once again demonstrated Africa’s leadership in governance monitoring when the Open Working Group on SDGs considered relegating issues on governance and peace to a set of cross-cutting development enablers in the preamble of the new development agenda (as was done in the Millennium Declaration adopted in 2000). Africa’s common position on the post-2015 Agenda (AU/UNECA, 2014), which included a stand-alone pillar on ‘Peace and Security’8 addressing a broad range of governance issues (AU/UNECA, 2014), was a game changer in at least three respects. First, as Cling et al. (2018) point out, it was instrumental to the global consensus on SDG 16, despite the strong opposition of a powerful faction in the Group of 77 — led by China, India and Russia.9 Second, by championing the inclusion of a dedicated goal 7 The twenty countries that officially expressed interest to the African Union in piloting the GPS-SHaSA instruments in 2013 were the following: Benin, Burundi, Cameroon, Cape-Verde, Chad, Congo-Brazzaville, Democratic Republic of Congo, Gabon, GuineaConakry, Côte d’Ivoire, Kenya, Madagascar, Malawi, Mali, Niger, Senegal, Seychelles, Togo, Tunisia and Uganda. 8 This ‘Peace and Security Pillar’ underlined the importance of “addressing the root causes of conflict” through tackling a broad range of governance issues such as reducing social inequality, exclusion and discrimination and encouraging democratic practices (AU/ UNECA, 2014). 9 Cling et al. (2018: 5) explain that “the countries most reticent to this SDG [on Governance and Peace] (China, India and Russia) found themselves increasingly isolated, to the point where they had to comply with the majority, notably following the change in

The African Statistical Journal, Volume 20, February 2018

151

Marie Laberge, Yeo Dossina and Francois Rouband

on governance and peace in the 2030 Agenda, with corresponding targets and indicators, African member states were also signalling to the world their confidence in the measurability of such a goal. This confidence was largely derived from their own recent experiments in this area, notably through the APRM and the GPS-SHaSA initiative,10 which were showcased at various high-level events11 leading to the adoption of the final 2030 Agenda in September 2015. Third, the AU’s advocacy for the adoption of governance and peace as a stand-alone goal further promoted “GPS data sovereignty” across the continent — and indeed, across the world — since the adoption of the goal would then require countries to produce national statistics to report on progress. 3.

THE GPS-SHASA METHODOLOGY AND ITS POLICY RELEVANCE

Below, we briefly describe the main features and advantages of the harmonized statistical instruments developed by the SHaSA Group on GPS, and field-tested by eleven pilot countries between 2013 and 2017. We illustrate, using examples, the policy relevance of statistics generated by these instruments and their suitability for reporting on SDG 16 and on Aspirations 3 (on good governance) and 4 (on peace and security) of Agenda 2063. 3.1 Strategic advantages of the GPS-SHaSA methodology The GPS-SHaSA methodology includes four instruments: two survey modules (two one-page questionnaires, one on governance and the other on peace and security, for a total of around 60 questions) and likewise, two administrative data collection instruments (again, one schedule of position of African countries”. They further note that Africa’s assertive stand contributed to “tip[ing] the scale” (p.5) in the politics of SDG16. 10 “It became obvious during the debates that the positions of African countries were based on their experience using household surveys in the context of the GPS-SHaSA programme” (Cling et al., 2018: 12). For instance, during subsequent negotiations on the selection of indicators for SDG 16, Africa once again asserted its leadership when expressing support for survey-based indicators, citing conclusive results from the GPS-SHaSA experience, despite serious reservations of several developed countries about the use of survey data for the measurement of governance, because of a lack of experience in this area and the general belief that NSOs should not get involved in this area. 11 For instance, the Joint UNDP/AU High-Level Event on SDG16 and the GPSSHaSA – “Towards Regional and National Statistical Capacities for Measuring Peace, Rule of Law and Governance: An Agenda for the Post-2015 Sustainable Development Goals Framework”, June 11-12, 2014, at the African Union Commission Addis Ababa, Ethiopia; and the Joint UNDP/AU High-Level Event on SDG16 and GPS-SHaSA, New York, December 2014.

152

Journal africain de statistiques, numéro 20, février 2018

5. Counting what counts: Africa’s seminal initiative on governance, peace and security statistics

administrative items on governance, and the other on peace and security). Additionally, supporting methodological tools12 were developed — among them, an interviewer training manual, metadata sheets to guide the collection of administrative data, survey results tabulation plans, as well as indicator matrices classifying GPS-SHaSA indicators by theme and sub-themes, and showing the complementarity of survey-based and administrative data. The deeply entrenched ownership of the methodological design process by the dozen or so African statisticians constituting the SHaSA Group on GPS is perhaps the most distinctive feature of this initiative (Razafindrakoto and Roubaud, 2015). At various steps in the process, expert inputs were invited and subsequently debated within the Group — notably from the Afrobarometer, the Mo Ibrahim Foundation, the Small Arms Survey, UNODC and UN Women. This strong ownership of the methodology within the GPS-SHaSA Group translated into a similarly strong endorsement by the Committee of Directors General of African NSOs throughout the data production cycle, including at the (critical) publication stage. The scientific provenance of the governance survey methodology that the SHaSA Group adopted for GPS contributed significantly to creating the confidence level required for its swift adoption by the broader community of African statisticians. By 1995, the national statistical office of Madagascar (INSTAT), assisted by the French research institute IRD-DIAL, was already pioneering survey-based measurements on democratic governance (Herrera et al., 2007). INSTAT developed a compact module on governance questions that could be appended to household surveys, often donor-funded for such purposes as health or agriculture. Based on this successful initiative, the IRD-DIAL researchers simultaneously replicated the approach in seven Francophone West African countries in the first half of 2000, and in seven countries of the Andean community in South America in the second half of the year. Subsequent scientific analyses of these initiatives demonstrated the reliability of the governance indicators produced, thanks to the established rigorous standards of the NSOs and to the large samples they could undertake. The utility of such indicators to policymakers, researchers and civil society, and the country ownership over the data production process, also represented real value-added over externally generated international indices on governance (Razafindrakoto and Roubaud, 2005; Herrera et al., 2007; Giang et al., 2011). These features contributed to the retention of this approach for the survey component of the GPS-SHaSA methodology in 2012. The combination of survey modules and administrative data collection instruments provided five main strategic advantages. 12 These methodological tools were available in English, French and Portuguese.

The African Statistical Journal, Volume 20, February 2018

153

Marie Laberge, Yeo Dossina and Francois Rouband

3.1.1 Combining survey data and administrative sources to get a complete picture Conducting surveys may be operationally easier in developing countries, relative to maintaining up-to-date administrative records13. Nonetheless, the GPS-SHaSA methodology was deliberately designed to show the links between ‘inputs’ — capabilities and efforts by the state to be inclusive, accountable and effective in managing public affairs, best measured through administrative sources — and ‘outcomes’ — notably, the lived experiences of citizens and their trust in institutions, best measured through survey data. The SHaSA Group on GPS argued that investing in administrative data collection systems on governance and peace was no less important, even if the investments required are more consequential.14 In Côte d’Ivoire, for instance, administrative statistics were used to help contextualize survey data on people’s trust in the courts of justice. When pairing levels of trust with the ratio of judges per 100,000 people, taking into account budgetary allocations to legal aid services and the proportion of defendants who had legal representation in courts, policymakers were able to identify some of the reasons why people in different regions of the country were more or less satisfied with court services (UNDP, 2017). 3.1.2 The imperative of capturing people’s “voice” when assessing GPS When monitoring GPS, the very nature of the issues at stake — how peaceful and inclusive are societies, how just and accountable are institutions — makes it especially important to integrate people’s voices into GPS measurements. In other words, the measurement approach used to monitor official commitments to improve governance needs to be true to these commitments’ inherent values and principles, such as the accountability of the state to its citizens. It is in this context that the GPS-SHaSA methodology placed strong emphasis on the use of survey-based evidence to capture peoples’ assessment of governance practices and peace dynamics in day-to-day life. As a senior advisor to the GPS-SHaSA pilot initiative remarked, “For the 13 Only half of GPS-SHaSA pilot countries tested the administrative data collection instruments, namely Kenya, Cote d’Ivoire, Malawi, Cape Verde and Burundi, while all eleven of them tested the survey modules. 14 This consideration was also shared by Directors-General. At their 2016 annual meeting in Cote d’Ivoire, some Directors noted that. “while much progress has been made on the survey component of the GPS-SHaSA methodology, the importance of also investing in administrative statistics should not be underestimated: the assessment of the ‘demand’ side […] needs to be complemented by an assessment of the ‘supply side’ […]. The more extensive investments required to establish administrative data collection systems in ministries and agencies (in time, in human resources, technologically and financially) should not deter the STG 1 from moving forward on this front while continuing its excellent survey work” (Laberge, 2016: 3).

154

Journal africain de statistiques, numéro 20, février 2018

5. Counting what counts: Africa’s seminal initiative on governance, peace and security statistics

vast majority of people — the uneducated, low-income laborers living in rural areas who rarely get a chance to participate in national policymaking, except for casting a ballot once every four or five years, but even that may turn out to be useless — for the vast majority of these people, participating in a governance survey represents a rare chance to have their voice heard by power-holders, particularly in countries where civil society or other intermediary bodies are poorly organized” (UNDP, 2017).15 3.1.3 Leveraging the statistical advantages of using nationally representative household surveys Piggybacking survey modules on a large support survey allows for the precise identification of the population groups — women, university graduates, northerners, urbanites, the unemployed, the poorest quintile, young people, etc. — most affected by the dysfunctions of governance systems. This is a major advantage of working with NSOs, compared with other types of organizations running governance surveys on smaller samples. The Afrobarometer surveys, for instance, are run on samples of approximately 2,400 respondents and as such have considerably higher margins of sampling error than GPS-SHaSA surveys, which have sample sizes that can go up to 40,000 households (see Table 1). As shown in Table 1, GPS-SHaSA modules in most pilot countries were grafted on general living conditions surveys or labor force surveys. In all countries, GPS-SHaSA survey modules were administrated to a representative sample of adults (above 18 years of age) randomly selected from the support survey.16 Attaching GPS survey modules to large-sample official surveys offers the added advantage of mobilizing other socio-economic variables available in the support survey to investigate interactions between measures of governance and broader measures of development outcomes, such as health-related data collected by a demographic and health survey, or food security data collected by a living conditions survey. 15 Quote by Mark Orkin, Senior Advisor to GPS-SHaSA, former Statistician-General, Statistics South Africa, and Associate Fellow, Department of Social Policy and Intervention, Oxford University 16 The selection of adults for the GPS-SHaSA survey was carried out at two levels. A first selection was done at the household level: in 40 percent of pilot countries, only a subset of the households sampled by the support survey were selected for the GPS-SHaSA modules. A second selection was done at the individual level: in 70 percent of pilot countries, only a subset of all adults living in a household were surveyed (in most cases, only one adult per household was randomly selected, using a variety of methods such as the Kish Grid, the nearest birthday, cards, etc.).

The African Statistical Journal, Volume 20, February 2018

155

156

12,848

10,303

Nb. of HH (theor.)

Nb. of HH (final)

Adult

50%

5,102

Universe

Nb. of HH (from S. S.)

Nb. of ind. (from HH)

Nb. of ind. (final)

3,771

n.a.

50%

Adult

2013

8,804

9,918

n.a.

Adult

2013

Pilot

Test

Specific

GATS

Kenya

14,198

All

Adult

2015

14,198

12,700

699

HLS/123

WMS

Malawi

10,600

All

Adult

2014

n.a.

4,470

298

GoV

GPD

Tunisia

Journal africain de statistiques, numéro 20, février 2018

Burundi

39,991

All

Adult

2015

21,402

22,080

911

HLS/123

13,116

All

Adult

2013-14

7,006

7,128

415

HLS /123

EMICoV ECVMB

Benin

P1-E123

Mada gascar

EMOP

Mali

3,082

25%

Adult

2015

n.a.

12,816

1 068

7,166

All

Adult

2015

4,020

220

13,835

All

Adult

2014-15

n.a.

5,466

911

HLS/123 HLS/123 HLS/123

ENV

Côte d’Ivoire

Other Countries (Self-starters)

1,036

33%

Adult

2013

n.a.

3,750

375

GoV

UNGBS

Uganda

Note: The Kenya National Bureau for Statistics (KNBS) was only able to test the GPS-SHaSA survey modules on a small sample of 74 individuals, as the donor funding the larger support survey declined the NSO’s request to graft the GPS-SHaSA modules to that survey.

Sources: GPS-SHaSA modules, 2013-2015, NSOs; authors’ calculations.

2014

Year of Survey

GPS-SHaSA Modules

1,024

HLS/123 HLS/123

Type of Survey

Number of PSUs

ECAM 4

IMC

Cape Verde

Name of the Survey

Support Survey

Cameroon

Pilot Countries

Table 1. Overview of sampling strategies applied to the GPS-SHaSA survey modules

Marie Laberge, Yeo Dossina and Francois Rouband

5. Counting what counts: Africa’s seminal initiative on governance, peace and security statistics

It is beyond the scope of this paper to elaborate on the robustness of the GPS-SHaSA survey results already discussed at length in Razafindrakoto and Roubaud (2015). Their analysis (based on a review of measurement and sampling errors in national GPS-SHaSA datasets, and on a review of the internal and external consistency of survey results) concludes that GPSSHaSA survey results are robust and reliable, in some cases of higher statistical quality even than traditional survey-based statistics on the labor force, living conditions or demographics. Importantly, GPS-SHaSA results produced by national statistical agencies were found not to differ significantly from those of the Afrobarometer survey, produced by a research network (Razafindrakoto and Roubaud, 2015; Calvo et al., 2018). This finding refutes the view that governance cannot be reliably measured through surveys run by public institutions, due to their assumed lack of independence. As Razafindrakoto, Senior Advisor to GPS-SHaSA, explains: “The Afrobarometer survey, at the end, asks respondents who they think is running this survey. Even if it is clearly mentioned by enumerators in their introduction that the survey is conducted by a non-governmental research outfit, more than half of the respondents still assume that the survey is run by the government. So, if Afrobarometer survey results are seen as independent and impartial even if most respondents think it is run by the government, then why would similar surveys run by public institutions such as NSOs be any less reliable?” (Razafindrakoto cited in UNDP, 2017). 3.1.4 Meeting the pan-African harmonization objective while respecting national specificities Survey questions were drawn from a variety of past or ongoing surveys with proven robustness. The surveys include the democratic governance survey module developed by IRD-DIAL in the early 2000s, the well-established Afrobarometer survey of national public attitudes in Africa on democracy and governance, and standardized victimization surveys (UNODC/UNECE, 2010). In keeping with the pan-African harmonization objective of the SHaSA, the African Charter on Democracy, Elections and Governance (AU, 2007) — the foremost pan-African normative framework on governance signed by 45 African member states to date — was used to define the thematic scope of the instruments and flesh out their substantive content. Out of the numerous options available, questions were also selected for their resonance with diverse national contexts across the continent, as well as for their analytical relevance. While the core GPS-SHaSA modules must be applied verbatim in each survey to ensure the comparability of data across countries over time, the GPS-SHaSA methodology encourages countries to add a few questions on

The African Statistical Journal, Volume 20, February 2018

157

Marie Laberge, Yeo Dossina and Francois Rouband

other aspects of governance and peace that were not addressed in the core questionnaire but may be important in a given national context. Uganda, Tunisia, Benin, Madagascar, Tunisia and Uganda added country-specific questions when they piloted the core survey modules. 3.1.5 The sine qua non consideration: Sustainability of the methodological approach The primary concern of African statisticians involved in the design of the GPS-SHaSA survey methodology was keeping the methodology as ‘light’ as possible. Concise survey modules tend to generate higher quality data, as the respondent’s full attention can be mobilized when interview time is kept to a minimum. The material, financial, and human resources required for implementation can determine the feasibility, reliability and sustainability of a survey. In this regard, the ‘add-on’ modular survey technique makes good economic sense, in that it saves cash-strapped African NSOs the cost of setting up additional, stand-alone surveys on GPS. The use of regularly conducted socio-economic surveys as vehicles for the GPS modules also helped ensure from the outset that GPS surveys would get repeated periodically — another critical feature for establishing time series on GPS statistics with which meaningful observations can be extracted for policy formulation. 3.2

The policy relevance of GPS-SHaSA indicators for monitoring SDG 16 and Africa’s Agenda 2063 Monitoring progress on 17 SDG global goals and 20 African goals is no small feat. For efficiency’s sake, the monitoring frameworks for both Agendas had to reduce the number of indicators to one or two ‘catch-all’ proxies per target. Yet, there is a broad consensus17 that reliance on single, standalone indicators can produce misleading assessments on progress in meeting any particular target. On the other hand, indicator baskets18 combining different types of (perception-based, experience-based and administrative-record-based) indicators enhance the evaluation of the effectiveness of a policy response by shedding light on a range of factors impacting and impacted by a certain issue. Importantly, the use of indicator baskets can reduce the tendency of

17 See SDG16 Data Initiative, 2017; Bertelsmann Stiftung & Sustainable Development Solutions Network, 2017; Transparency International, 2017; Institute for Economics and Peace, 2016. 18 ‘Indicator baskets’ typically combine experience-based indicators to monitor the actual occurrence of a phenomenon, ‘input’ and ‘output’ indicators – often drawn from administrative records – to track concrete steps taken to address the problem, as well as public perceptions indicators to see whether the public feels that an improvement is truly occurring, or what their attitudes are towards certain issues.

158

Journal africain de statistiques, numéro 20, février 2018

5. Counting what counts: Africa’s seminal initiative on governance, peace and security statistics

states to improve performance on a few indicators without implementing real change in people’s lives. Recent independent efforts to monitor progress on SDG 16 using a basketapproach have drawn from a wide range of data sources — a time-consuming process for the entity having to quality assure this data produced by third parties before aggregating it. In this regard, the GPS-SHaSA methodology offers a substantial advantage. It centralizes data collection in national statistical offices, where the vetting, integration and quality assurance of various types of survey-based and administrative data can be performed by experts in statistical methods and standards. To illustrate the value of applying a basket-approach to measuring targets under Agenda 2030 and Agenda 2063, we present in this section a few results from selected GPS-SHaSA pilot countries that produced both survey-based and administrative data. As explained by Ben Paul Mungyereza, the Executive Director of the Uganda Bureau of Statistics: “It’s a myth that policymakers are not interested in, or distrust, data because it’s based on citizen perceptions rather than on ‘real’ experiences or other ‘objective’ information. The fact of the matter is, regardless of whether a government institution is actually a hotbed of nepotism (for example), the popular perception that it is one is probably more important than the actual state of affairs—because this perception shapes citizens’ behaviour and attitudes towards the government” (Mungyereza cited in UNDP, 2017). In this section, we demonstrate how four types of data generated by the GPSSHaSA instruments — perception data tracking people’s own assessments and appreciations, experience data measuring people’s experiences and behaviors, data on the values and norms they hold, and data from administrative sources compiled by various governmental entities — can be combined to produce rich policy insights. This will be illustrated through actual applications of the basket-approach for monitoring Agenda 2063’s target on ‘free and fair elections’ and Agenda 2030’s target on ‘non-discrimination’. 3.2.1 Monitoring Agenda 2063’s target on “Free and Fair Elections” While the global monitoring framework for SDG 16 deliberately omitted reference to elections, this core aspect of democratic governance is addressed in a dedicated target under Africa’s Agenda 2063 Priority Area 1 on “Democratic Values and Practices are the Norm,” which states that by 2023, at least 70 percent of the public should perceive elections as free, fair and transparent.

The African Statistical Journal, Volume 20, February 2018

159

Marie Laberge, Yeo Dossina and Francois Rouband

Figure 1. Free and fair elections: Do you consider it as “essential”? Is it “respected” in this country? Benin 100 Uganda

Burundi

80 60 40

Tunisia

Cameroon

20 0

Mali

Cape Verde

Malawi

Côte d’Ivoire Madagascar Essential

Respected

Sources: GPS-SHaSA modules, 2013-2015, NSOs, various countries; authors’ calculations. Note: Tunisia did not include “free and fair elections” amongst the “key characteristics of democracy” listed for this question.

As illustrated by Figure 1 above, there is generally a wide gap (except in Malawi and to a lesser extent Mali) between the level of popular aspirations (the extent to which people say that free and fair elections are “an essential characteristic of democracy”) and people’s perception of the extent to which elections are indeed free and fair in their country. Such measures can be utilized by policymakers to ‘localize’ global or continental targets, taking into account national circumstances, as envisaged by both Agendas. For instance, the AU 2023 target of at least 70 percent of the population finding elections as free and fair might not be a realistic target in countries such as Benin, Cote d’Ivoire and Madagascar where this figure currently stands at 40 percent or less.

160

Journal africain de statistiques, numéro 20, février 2018

5. Counting what counts: Africa’s seminal initiative on governance, peace and security statistics

Figure 2. Voting behavior in Burundi: Did you vote in the last election? If not, why? (Urban vs. rural populations) 100 80

89.7%

Did you vote in the last election?

81.4%

If not, what is your reason?

60 48.0% 40 28.5% 20 6.2% 0 % Who voted in the last election (June 2010) Urban

12.0% 3.6%

25.4% 16.2%

6.5%

% Eligible voters % Registered voters Voting does not No candidate/party NOT registered who did NOT vote make a difference represented my views

Rural

Source: GPS-SHaSA module, 2014, ISTEEBU, Burundi; authors’ calculations. Note: The GPS-SHaSA survey took place before the elections cycle of 2015-16.

In the left panel, Figure 2 uses experience-based data from the GPS-SHaSA survey module to depict the voting behavior of Burundians in the 2010 elections, and reveals a higher electoral participation rate in the rural population. The panel on the right uses perception data to draw policymakers’ attention to possible explanations for the lower electoral turnout rate in urban areas: nearly half of urban Burundians who did not vote felt that “voting does not make a difference”, and one in four felt that no candidate or party represented their views. Table 2 below applies the four-dimensional basket approach introduced above to the cases of Burundi and Cote d’Ivoire to illustrate how the four types of data generated by the GPS-SHaSA instruments can be used to examine the electoral climate in any given country, and help inform policy. While people’s perception of the freeness of the 2010 elections in both countries is almost the same (60 percent in Burundi and 62 percent in Cote d’Ivoire), the voter turnout in Cote d’Ivoire is significantly lower than in Burundi. This observation is validated by two types of data: a 5-point difference between the two countries in experimental survey-data (“Did you vote in the last election?”) and a seven-point difference in the electoral roll

The African Statistical Journal, Volume 20, February 2018

161

Marie Laberge, Yeo Dossina and Francois Rouband

tallies of both countries. To investigate possible reasons behind the lower voter turnout in Cote d’Ivoire, three factors can be examined. First, Table 2 shows a much lower level of satisfaction with democracy in Cote d’Ivoire (56 percent) than in Burundi (76 percent), a variable which may generate a certain level of voter apathy. Second, less than a third (31 percent) of the population in Cote d’Ivoire feels that politicians take their concerns into account. When comparing Members of Parliament and traditional leaders based on their ability to “listen to people like them”, parliamentarians elected through the ballot box scored 30 points lower than traditional leaders. Third, 92 percent of Ivoirians consider the principle of “free and fair elections” as essential to a democracy’s effective functioning, compared to 98 percent of Burundians. This could be explained, at least in part, by the generally negative views held by Ivoirians about electoral practices and elected officials. This is an important point for policymakers to ponder on, as people’s perceptions can ultimately affect their adherence to certain norms and values (including their relative “preference” for democratic regimes over other types of more authoritarian regimes) if their aspirations are consistently frustrated by day-to-day experiences. Burundi’s higher voter turnout and considerably higher level of satisfaction with democracy are only part of the story. Policymakers also need to pay attention to much more negative views held by urban populations (a 20-point gap with rural populations, compared to an 8-point gap in Cote d’Ivoire). Policymakers intending to understand the possible factors fueling such feelings of dissatisfaction amongst urban populations in Burundi will note the 33-point difference between urban populations saying that parliamentarians (23 percent) and traditional leaders (56 percent) “pay particular attention to people like themselves”. Other elements of response are in the large share of voters who did not vote because “voting does not make a difference” (33 percent in Burundi compared to 7 percent in Cote d’Ivoire) or because no candidate or party represented their views (18 percent in Burundi compared to 8 percent in Cote d’Ivoire).

162

Journal africain de statistiques, numéro 20, février 2018

5. Counting what counts: Africa’s seminal initiative on governance, peace and security statistics

Table 2. Four-dimensional indicator basket to monitor Agenda 2063 target on “Free and Fair Elections” Agenda 2063 Target 2023: “At least 70% of the public perceive election to be free, fair and transparent by 2020” Comparative illustration for Burundi and Cote d’Ivoire Data type

GPS-SHaSA indicators

Data for Burundi

Data for Cote d’Ivoire

1. Perceptions

% of population saying that the principle of free and fair elections is respected in the country

60% (urban: 42.1% vs. rural: 62.3%)

61.9% (urban: 58% vs. rural: 65.9%)

% of population saying they did NOT vote because: (1) Voting does not make a difference; (2) No candidate / party represented their views

32.5% 18.1%

6.8% 8%

% of population saying that politicians take into account citizens’ concerns

36% (urban: 30.8% vs. rural: 36.6%)

29.8% (urban: 30.4% vs. rural: 29.2%)

% of population saying that: (1) Members of Parliament (2) Traditional leaders listen to people like themselves

21.5% (urban: 23.1% vs. rural: 21.3%) 65.7% (urban: 56.4% vs. rural: 66.9%)

30.5% (urban: 27.2% vs. rural: 33.8%) 60.8% (urban: 56.7% vs. rural: 65%)

% of population saying that they 76.2% are satisfied with how democracy (urban: 59.8% vs. works in their country rural 78.3%)

56.4% (urban: 51.5% vs. rural: 61.5%)

2. Experiences

% of population saying that they 88.8% voted in the last general election

83.5%

3. Norms/ values

% of population indicating that 98% free and fair elections is an essential characteristic of democracy

91.9%

4. Administrative sources

Proportion of registered voters who voted during the last presidential elections

83.7% (first round, 2010) 81.1% (second round, 2010)

The African Statistical Journal, Volume 20, February 2018

90.5% (2010)

163

Marie Laberge, Yeo Dossina and Francois Rouband

3.2.2 Monitoring SDG Target 16.B on Non-Discrimination Global indicator 16.b.1 (Proportion of population reporting having personally felt discriminated against or harassed in the previous 12 months on the basis of a ground of discrimination prohibited under international human rights law) does not measure the actual incidence of discrimination and harassment occurring in a given population. Rather, this indicator measures the proportion of the population having identified (subjectively) that they had been discriminated against and/or harassed, and willing to disclose this information to data collectors. Africa’s Agenda 2063 also has a focus on human rights and a national target19 for 2023 that requires the monitoring of discrimination. In at least three respects, the four-dimensional indicator baskets that can be assembled with GPS-SHaSA data (Table 3) provide richer data on discriminatory patterns than the use of global indicator 16.b.1 in isolation. First, GPS-SHaSA survey data make it possible to distinguish between ten different types of discrimination — ethnicity, sex, language, religion, regional origin, foreign origin, economic situation (poverty), disability, political affiliation and homosexuality. Second, the GPS-SHaSA survey modules allow for the monitoring of perceptions of discrimination alongside lived experiences of discrimination. Figure 3 reveals much higher levels of perceived discrimination, when compared with experienced discrimination. This observation is fairly consistent across the ten pilot countries (Cote d’Ivoire being a notable exception), with gaps between these two types of measures found to be widest in Cape Verde and Tunisia, for poverty-based discrimination. Important policy insights can be gained from these large differences between measures of perception and actual experience. For instance, such gaps can arise from inadequate communication between the state and citizens. Policymakers should pay attention to high levels of perceived discrimination, whether or not they reflect reality, because perceptions drive behavior — people perceiving widespread discrimination are more likely to adopt discriminatory practices themselves, thus fueling a sort of self-fulfilling prophecy (UNDP, 2017).

19 Situated under Priority Area 2, this target states: “At least 70% of the people perceive the entrenchment of the culture of respect for human rights, the rule of law and due process.”

164

Journal africain de statistiques, numéro 20, février 2018

5. Counting what counts: Africa’s seminal initiative on governance, peace and security statistics

Figure 3. Perceived and experienced discrimination: Ethnicity vs. economic situation 80 70 60 50 40 30 20 10

Perceived discrimination / ETHNICITY Experienced discrimination / ETHNICITY

AVERAGE for 10 countries

Cameroon

Uganda

Malawi

Tunisia

Côte d’Ivoire

Benin

Cape Verde

Madagascar

Burundi

Mali

Perceived discrimination / ECONOMIC SITUATION Experienced discrimination / ECONOMIC SITUATION

Sources: GPS-SHaSA Module, NSOs, various countries; authors’ calculations. Note: The questions are formulated as follow: “People are sometimes discriminated against on various grounds. In this country, do you think there is discrimination related to [this characteristic]? In the past 12 months, have you ever been victim of discrimination due to [this characteristic]?”

Third, the Peace & Security component of the GPS-SHaSA methodology allows for the specific investigation of discrimination perpetrated by security forces. As shown in Table 3, the markedly higher perceptions on discrimination by security forces in Cote d’Ivoire (26 percent, compared to 17 percent in Burundi) correlate with higher levels of experiences of discrimination by security forces in Cote d’Ivoire (17 percent, compared to 5 percent in Burundi). Policymakers looking for advice on how to tackle this issue need to take into account contrasting scenarios in the two countries in terms of where such discriminatory practices by security forces are most prevalent. While in Burundi self-reported experiences of discrimination are mainly concentrated in urbans areas (8 percent in urban areas vs. a lower national average of 5 percent), the opposite holds true in Cote d’Ivoire (only 12 percent in the capital Abidjan vs. a national average of 17 percent). Policymakers

The African Statistical Journal, Volume 20, February 2018

165

Marie Laberge, Yeo Dossina and Francois Rouband

would also need to pay attention to the administrative data collected on “the proportion of security personnel prosecuted over the total number of reported cases of misconduct”. Data provided by Cote d’Ivoire show that, while some level of prosecution was taking place in 2014 for misbehaving military personnel (69 percent of reported cases were prosecuted) and gendarmerie personnel (31 percent of reported cases were prosecuted), no police personnel was prosecuted for misconduct that year — even though people typically interact more frequently with the police than with the military. Table 3. Four-dimensional indicator basket to monitor SDG target 16.B “Promote and enforce non-discriminatory laws and policies for sustainable development”20 Indicator 16.b.1: Proportion of population reporting having personally felt discriminated against or harassed in the previous 12 months on the basis of a ground of discrimination prohibited under international human rights law Comparative illustration for Burundi and Cote d’Ivoire Data type

GPS-SHaSA indicator

Burundi

Cote d’Ivoire

Perceptions

% of population saying that the principle of non-discrimination is respected in the country

79.8%

55.2%

20.9% 13.2% 18.5% 16.3% 22.5%

18.9% 13.2% 14% 9.2% (23.4%

16.9%

25.6%

% of population saying that there is discrimination due to20: (1) Ethnicity (2) Religion (3) Region (4) Gender (5) Economic situation % of population saying that some people are discriminated against by public security services

20 Other dimensions of discrimination explored by this question in the governance survey module include: Language/dialect, being foreign, disability, political affiliation, and sexual orientation.

166

Journal africain de statistiques, numéro 20, février 2018

5. Counting what counts: Africa’s seminal initiative on governance, peace and security statistics

Indicator 16.b.1: Proportion of population reporting having personally felt discriminated against or harassed in the previous 12 months on the basis of a ground of discrimination prohibited under international human rights law Comparative illustration for Burundi and Cote d’Ivoire Data type

GPS-SHaSA indicator

Burundi

Cote d’Ivoire

Experiences

% of population saying that they have been a victim of discrimination due to: (1) Ethnicity (2) Religion (3) Region (4) Gender (5) Economic situation

(1) Ethnicity: 4.5% (2) Religion: 2.4% (3) Region: 3.1% (4) Gender: 2.4% (5) Economic situation: 7%

(1) Ethnicity: 14.1% (2) Religion: 9.4% (3) Region: 8.8% (4) Gender: 6% (5) Economic situation: 13.7%

% of population saying that they 4.8% (Urban have been discriminated against areas: 8.1%) by public security services on the basis of at least one type of discrimination

17,1% (Abidjan: 11.6%; other urban areas: 20.1%)

Norms/ values

% of respondents indicating that ‘Absence of discrimination’ is an essential characteristic of democracy

83.4%

Administrative sources

% of fifteen core international (1) 78% (2014) and African conventions [+ (2) 67% (2014) regional conventions as relevant] on governance and human rights which were: (1) ratified AND enacted in national legislation (2) for which a first compliance report (at a minimum) was submitted to treaty bodies to report on implementation

98.4%

(1) 53.3% (2014) (1) 0%21 (2014)

21 Error in interpretation: 0% indicates the proportion of these conventions for which a progress report was submitted during the year (2014) the GPS-SHaSA data collection took place — whereas the indicator aims to measure the proportion of conventions for which a progress report has been submitted — at any point in time since their adoption/ ratification (which is what the figure of 67 percent for Burundi represents.)

The African Statistical Journal, Volume 20, February 2018

167

Marie Laberge, Yeo Dossina and Francois Rouband

GPS-SHaSA indicator

Burundi

Cote d’Ivoire

Proportion of security personnel prosecuted over total number of reported cases of misconduct

Data not available Military personnel: 69% (2014) Gendarmerie personnel: 31% (2014) Police personnel: 0% (2014)

THE GPS-SHASA INITIATIVE IN DIVERSE INSTITUTIONAL AND POLITICAL CONTEXTS

A variety of motivations prompted national statistical agencies to embark on GPS statistical production, each national context presenting a unique configuration of challenges and opportunities. In this section, we analyze the political and institutional contexts in which national pilots took place, and we investigate a number of factors which may have facilitated NSOs’ entry in an area traditionally seen as sensitive in most African countries. 4.1 A variety of national scenarios 4.1.1 NSOs with no previous experience with GPS statistics Lack of experience with GPS statistics was in no way a barrier to entry for neophytes. In countries where the production of GPS statistics was a new venture, the leadership of statistical offices essentially adopted one of two strategies. First, in countries where a government entity had previously requested some type of governance data, the statistical agency could conveniently refer to this earlier request to introduce the GPS-SHaSA pilot. In Uganda, for example, the Ministry of Public Service had asked the Uganda Bureau of Statistics (UBOS) for data to help address the scourge of corruption and abuse of public office in service delivery. In response to this request, UBOS partnered with the School of Statistics and Planning of Makerere University to map existing governance data producers in Uganda amongst government, civil society and research institutes, and to assess the quality of existing data sources (UBOS, 2014). By the time the GPS-SHaSA survey modules were

168

Journal africain de statistiques, numéro 20, février 2018

5. Counting what counts: Africa’s seminal initiative on governance, peace and security statistics

brought to the attention of the Director General of UBOS, the statistical office had already designed (with support from UNDP) its own “National Baseline Survey” on governance to collect data on the specific concerns raised by the Ministry. On realizing the importance of the harmonization objectives underpinning the GPS-SHaSA modules, the DG promptly secured additional funding from UNDP to attach the GPS-SHaSA modules to the (Uganda-specific) Baseline Survey. With its specific component on peace and security, the GPS-SHaSA questionnaire offered a useful complement to the Baseline Survey on governance, especially in view of remaining tensions in the North of Uganda. The piloting of the GPS-SHaSA modules in Uganda enabled the NSO to go beyond addressing the initial specific request of the Public Service Ministry. Equipped with the GPS-SHaSA questionnaire, UBOS was able to embrace a larger ambition — supplying policymakers “measurable governance indicators to inform the National Development Plan, the Uganda Vision 2040, the East African Community Integration Agenda, the African Union Vision 2063, and the post-2015 Sustainable Development Goals” (UBOS, 2014). In the second category, where no pre-existing government request could be leveraged to “justify” the launch the GPS-SHaSA pilot in the country, statistical agencies had to be more opportunistic in creating demand for GPS data from the political leadership. In Burundi, for instance, the NSO (which reports to the Ministère à la Présidence Chargé de la Bonne Gouvernance et du Plan, a ministry responsible for good governance) introduced the GPS-SHaSA pilot as a monitoring tool aligned with existing official commitments on governance found in various national planning frameworks. In the preface to the first national report on GPS statistics, the Director-General of the NSO invoked the country’s national vision, Vision Burundi 2025, which describes governance as a critical lever for economic development and for the improvement of citizens’ living conditions (ISTEEBU, 2014). Reference to such official commitments to governance proved to be an effective strategy for statisticians operating in young democracies to secure the buyin of politicians who would otherwise be averse to the type of democratic accountability enabled by nationally representative surveys on governance that translate people’s voice in unambiguous numbers. Other NSOs capitalized on ongoing efforts by the executive branch to establish national governance monitoring systems to introduce the GPS-SHaSA pilot. Empowered by the recent adoption of the GPS-SHaSA methodology by the African community of statisticians, these NSOs demanded roles in such national governance monitoring systems, invoking their credentials as public institutions formally entrusted with the mandate to produce official

The African Statistical Journal, Volume 20, February 2018

169

Marie Laberge, Yeo Dossina and Francois Rouband

statistics on all major areas of citizens’ lives, including governance. This was the case in Malawi where initial plans to establish a national monitoring system for the Democratic Governance Sector Strategy (Government of Malawi, 2012) had assigned overall responsibility for this system to an executive entity in the Office of the President and Cabinet. Similarly, in Tunisia, the NSO introduced GPS-SHaSA instruments to officials of the Ministry of Foreign Affairs and the Ofice of the President and Cabinet, which was at the helm of a UNDP-supported initiative (UNDP, 2016) to demonstrate the measurability of SDG 16 through national data collection systems. The critical role played by the Tunisian statistical agency in demonstrating the feasibility of reporting on SDG 16 through (GPS-SHaSA-inspired) survey-based indicators had significant ripple effects at the global level — the feasibility of the GPS-SHaSA approach beyond Sub-Saharan Africa had now caught the attention of the international community.22 4.1.2 NSOs with some prior experience with GPS statistics In contrast to the first group of NSOs attempting to produce GPS statistics for the first time, others had an existing record of accomplishment in this area. These statistical agencies had, to some degree, already secured the blessings of the political leadership. They saw their participation in the GPS-SHaSA initiative as a way of consolidating or complementing ongoing efforts to produce governance statistics. The statistical office of Cote d’Ivoire, for instance, established in 2007 a dedicated department on governance statistics (Département de l’Organisation de l’Information pour la Gouvernance), which regularly publishes governance statistics, compiled mainly from survey data on perceptions and experiences of corruption. A request by the Haute Autorité pour la Bonne Gouvernance (National Commission on Good Governance) for a broader range of governance statistics to help inform the Commission’s annual report on governance led the NSO to a test of the GPS-SHaSA instruments and the expansion of the scope of its governance statistics beyond corruption. Similarly, the Kenyan statistical bureau used the GPS-SHaSA pilot as a stepping stone to the conversion of its Crime and Justice Statistics Unit23 into a fully-fledged Governance Statistics Section, which now publishes a wide range of governance statistics (albeit from administrative data sources only) in the Annual Economic Survey and the Statistical Abstracts. This 22 UNDP and selected governments, High-Level Side-Event SDG 16 Pilots, Participation by Tunisia, 2015. 23 When Kenya’s Vision 2030 was launched in 2008, with a strong focus on tackling crime and improving security across the country, the government requested the Kenyan National Bureau of Statistics to start producing statistics on crime, justice and security.

170

Journal africain de statistiques, numéro 20, février 2018

5. Counting what counts: Africa’s seminal initiative on governance, peace and security statistics

was also the case in Cape Verde where a Justice and Security Statistics Unit had been created in 2011, at a time when security concerns ranked high on the national agenda. The GPS-SHaSA survey modules enabled the Cape Verdean statistical agency to supplement existing administrative data collection with survey data, and to widen the range of reported issues beyond justice and security. In Benin, the (self-funded) GPS-SHaSA pilot was launched in the wake of the publication of the 2014 Mo Ibrahim Index, which showed a decline24 in the country’s overall governance performance since 2011 (Mo Ibrahim Index, 2014). As explained by a statistician seconded to the Prime Minister’s Office, the opportunity to take part in the GPS-SHaSA pilot came just when the political leadership in Benin was questioning the validity of external ‘expert’ perceptions making up a large part of the Mo Ibrahim Index. In this context, the GPS-SHaSA survey provided a welcome countervailing approach to the use the country’s own data to ‘cross-check’ the declining trends recorded in the Mo Ibrahim Index. Finally, the Cameroonian NSO, mindful of the limited appetite for GPS statistics shown by authorities in previous years, presented the GPS-SHaSA pilot as an opportunity to further consolidate the country’s leadership position in Central Africa (and long-standing reputation as a regional pole of statistical excellence) by being the first GPS-SHaSA pilot in the subregion. A peer from the statistical office of the neighboring Republic of Congo remarked at the launching event that Brazzaville was hoping to follow soon “Yaoundé’s pioneering example”.25 4.2 Hedging sensitivities by adopting a ‘big-tent’ approach 4.2.1 National validation workshops The NSOs implemented the GPS-SHaSA pilots in close consultation with numerous government actors, civil society organizations, academia and research institutions. Their intention was to foster broad-based ownership of the initiative, and to create a bottom-up demand for GPS statistics. In the words of Dorcas Nabukwasi, a Ugandan statistician, “the perception that people had of [the Ugandan Bureau of Statistics] was greatly enhanced when they started to see it as an institution in tune with their daily struggles and aspirations; suddenly, the numbers started to make sense to them” (Nabukwasi cited in UNDP, 2017). 24 Benin registered a deterioration of -0.2 in its Mo Ibrahim Index score between 2014 and 2011. See http://static.moibrahimfoundation.org/u/2015/10/02201305/03_Benin. pdf. 25 The GPS-SHaSA pilot was launched in Yaoundé, Cameroon, in August 2013.

The African Statistical Journal, Volume 20, February 2018

171

Marie Laberge, Yeo Dossina and Francois Rouband

Most countries launched the pilot by hosting a ‘National Validation Workshop’ where the GPS-SHaSA methodology was presented to all stakeholders. Such events brought together likely users of GPS statistics — including the country's political leadership, parliamentarians, relevant ministries, departments and agencies, oversight institutions such as anti-corruption commissions and audit institutions, civil society and academia — and data producers in relevant government entities. All actors had the opportunity to propose amendments or additions to the four data collection instruments to better suit the local context.26 More advanced peers from other pilot countries sometimes also attended to share their experience with national stakeholders.27 Experience-sharing by other pilot NSOs with a broad national audience proved to be an effective way to establish the scientific credibility of the GPS-SHaSA methodology and to build national confidence around its feasibility, while at the same time building a strong team spirit amongst pilot NSOs. 4.2.2 Multi-stakeholder “Steering Committees on GPS Statistics” As a direct outcome of these validation workshops, several pilot countries established multi-stakeholder “Steering Committees on GPS Statistics” (known by other names in different countries). These committees were mandated to keep potential users of GPS statistics engaged throughout the process, thus increasing the likelihood of the use of GPS-SHaSA statistics by the institutions represented on such committees. Coordinated by the NSO, the membership of these committees included both statistical focal points in relevant government entities and representatives from civil society and academia.28 In Uganda, for instance, the School of Statistics and Planning of Makerere University was a key member of the Technical SubCommittee on Governance Statistics, and played a critical role in training survey enumerators. 26 In Cameroon, for instance, a question was added on linguistic discrimination against the Anglophone community, and a new sub-question on the elderly was added to a question on victimization, to capture abuses perpetrated against older women accused of practicing witchcraft. In Kenya, additional indicators on natural resource governance were added to the administrative data collection instrument. 27 For instance, a Malian statistician took part in Tunisia’s validation workshop, and a Cape Verdean statistician contributed to validation workshops in Malawi and Cameroon. 28 For example, the Technical Sub-Committee on Governance Statistics established by the Ugandan Bureau of Statistics included representatives from the School of Statistics and Planning of Makerere University, civil society organizations, the media, the Electoral Commission, the National Human Rights Commission, the Office of the Prime Minister, the National Planning Authority, institutions from the Justice, Law and Order Sector, and development partners such as the UN and DFID (UBOS, 2014).

172

Journal africain de statistiques, numéro 20, février 2018

5. Counting what counts: Africa’s seminal initiative on governance, peace and security statistics

Steering committees also acted as guarantors of methodological rigor throughout the process, which was essential for GPS statistics to advertise themselves as reliable and trustworthy. In Malawi, for instance, the Committee performed a quality assurance function at three levels — Committee members verified the accuracy of the questionnaire translation into the two national languages, Chichewa and Tumbuka. They also contributed to the training of enumerators and integrated teams of senior statisticians to observe interviews during fieldwork (UNDP, 2017). Regarding administrative data collection, Committee members served as a critical interface with their respective institution: they were responsible for assessing data availability and quality, as well as for securing the full collaboration of their agency in sharing the requested data within agreed timeframes. They also helped identify specific capacity-building needs in their respective agencies to enhance data collection practices, which the NSO would then address through targeted training. The Cape Verde NSO was particularly effective in this regard: “cooperation protocols”29 between the statistical office and various government entities30 became the order of the day. The protocols stipulated the format and frequency of data-sharing with the NSO, and reiterated the strict application of rules on information confidentiality. The Kenyan Bureau of Statistics was particularly successful in establishing a Technical Working Group on GPS statistics comprising of representatives from as many as thirty government entities31. This Working Group was fur29 In view of the high interest generated by the Cape Verdean experience, a ‘sample protocol’ (UNDP/AUC/INECV, Sample Collaboration Protocol for GPS-SHaSA Administrative Data Production, 2013, available in French and English) was designed and made available in both French and English to other pilots, as a tested model to help formalize collaboration between NSOs and relevant government entities. In most other pilot countries however, NSOs’ efforts to implement such protocols were halted due to a lack of sufficient financial resources to implement the elaborate capacity-building plans elaborated in such protocols. 30 At the time of the GPS-SHaSA pilot, the Cape Verde statistical office had formalized such protocols with the Ministry of Justice, the Ministry of Internal Affairs, the Prosecutor’s Office, the Superior Council for the Judiciary, and the Judicial Police. 31 The Kenya Technical Working Group on GPS Statistics included representatives from the following 30 institutions: the Kenya Police Service, the Judiciary, the Kenya Prisons service, the Probation and Aftercare Services, the Public Prosecution, the Ethics and Anti-Corruption Commission, the National Registration Bureau, the Independent Electoral & Boundaries Commission, the Immigration department, the Children Service Department, the National Assembly, the Institute of Development Studies, University of Nairobi, the National Environment Management Authority, the Kenya Wildlife Service, the Law Society of Kenya, the Office of the Attorney General, the National Gender and

The African Statistical Journal, Volume 20, February 2018

173

Marie Laberge, Yeo Dossina and Francois Rouband

ther subdivided into three subcommittees — on criminal justice statistics, governance and security statistics, and environmental governance statistics (a Kenya-specific addition). The Working Group met on a quarterly basis to assess progress in data collection, and was particularly effective in facilitating the design of new data collection protocols to help harmonize data collection practices within sectors (between the courts and the police, for instance) and to increase data-sharing amongst institutions. The Working Group also developed a joint annual work plan, which empowered individual members to integrate specific data collection activities in the work program of their own institution (KNBS, 2014). 5.

LOOKING AHEAD: WHAT IT TAKES TO INSTITUTIONALIZE GPS STATISTICS

Institutionalizing the production of GPS statistics at country-level will require at least four types of investments — institutional, methodological, financial and communicational. The specific actions, finances and skill sets required across these four domains were fleshed out in the five-year work plan and budget tabled by the SHaSA Group on GPS statistics at the annual meeting of the Committee of Directors General of Statistics in Yamoussoukro, in 2012 (AU, 2012). Below are the details, supplemented by lessons emerging from pilot experiences. There is a consensus among all eleven NSOs that piloted the GPS-SHaSA instruments on the need to institutionalize expertise on GPS statistics within the NSO — for instance, by creating a dedicated unit on GPS statistics rather than relying on a few statisticians scattered across other departments. The production of GPS statistics is a new area for most African NSOs, and requires staff properly trained on the subject matter and working full-time on this agenda, especially to cultivate the strong institutional partnerships needed across government to collect administrative statistics. Merely nominating a ‘GPS focal point’ (as was the case during the pilot phase), with pre-existing responsibilities and insufficient time to dedicate to this new area, is unlikely to lead to sustainable results. At a methodological level, the institutionalization of GPS statistics will require the permanent integration of GPS-SHaSA survey modules in a dedicated household survey (such as a living conditions survey or a labor Equality Commission, the Ministry of Defense, the Kenya National Commission on Human Rights, the Communication Commission of Kenya, the Monitoring and Evaluation Directorate, Ministry of Devolution & Planning, and the National Crime Research Centre.

174

Journal africain de statistiques, numéro 20, février 2018

5. Counting what counts: Africa’s seminal initiative on governance, peace and security statistics

force survey) to be conducted regularly. For some NSOs, the piggybacking strategy for the survey modules proved challenging. Ironically, the intended cost-saving advantages using the same approach sometimes failed, as the more fundamental challenge of securing sufficient funding for the support survey turned out to be a challenge. This was an issue in Cote d’Ivoire where the two surveys considered as candidate support surveys for attaching the GPS modules were repeatedly postponed due to insufficient funding. In Kenya, the NSO failed to convince the donor funding the candidate support survey to attach the GPS modules to it. Some countries therefore considered running the GPS survey modules independently, with national resources. Although this alternative approach would reduce the sample size, it would help ensure the regularity of the surveys over time. However, the piggybacking tactic continues to be the most feasible in most countries. In Uganda, for instance, the statistical office has selected from the GPS-SHaSA survey modules a small subset of questions to be integrated in the largest household survey run by the NSO every three years. Financial needs are a lingering concern, not only for data collection activities, but also for investment in the requisite institutional architecture within NSOs and in data-producing government entities. The needs are particularly acute for administrative data collection. All five countries (Burundi, Cape Verde, Cote d’Ivoire, Kenya and Malawi) that tested the administrative data collection instruments confirmed the feasibility of the exercise. Among the five is Burundi, which according to the World Bank indicator of statistical capacity, has the least developed statistical system amongst all pilot countries, yet could still measure 80 percent of GPS-SHaSA administrative indicators (ISTEEBU, 2017). These five countries, nonetheless, called for the deployment of an extensive training program for statistical units in government agencies. Most NSOs observed that the main challenge is not that ministries or agencies refuse to share data, but rather that data are unavailable, or of poor quality.32 This is due to low budgetary allocations to monitoring, evaluation and statistical production across government agencies. It is therefore no small feat that some pilot countries, such as Kenya, successfully leveraged the GPS-SHaSA initiative to secure new budgetary allocations towards the production of GPS statistics (the Kenyan police, for instance, secured funding for the statistical office to help create statistical units in police establishments across the country). The establishment or improvements of fully-fledged statistical units in GPS-related ministries and agencies is a 32 For example, the Kenyan pilot noted incomplete datasets, arithmetic errors, manual processing of data in the field leading to various errors, etc. (KNBS, 2014).

The African Statistical Journal, Volume 20, February 2018

175

Marie Laberge, Yeo Dossina and Francois Rouband

potential practical solution to high staff turnover, which deprives NSOs of their focal points in ministries after having invested considerable time and efforts in building their capacity. Finally, the sustainability of GPS statistical production hinges on the effective dissemination of results and conversion of data into policy-relevant findings. In Cape Verde, for instance, the appetite for GPS data in government entities is explained by the active role played by public servants and government officials in analyzing survey results. Instead of analyzing these results behind closed doors, the Cape Verde statistical office organized GPSSHaSA retreats where government agencies and civil society organizations mingled in sector-wide working groups and examined datasets from the viewpoint of their own policy interests. At the heavily publicized launch in the National Assembly, the people’s ‘voices’ were conveyed unaltered to their elected representatives, in a powerful show of direct democracy. A few days later, the GPS-SHaSA survey caught people’s attention again when the President of Cape Verde, at the country’s 39th independence anniversary, raised concerns about some democratic shortcomings unveiled by the survey, notably in relation to popular perceptions of the unequal treatment of citizens before the law (UNDP, 2017). The dissemination of GPS-SHaSA survey results had equal impact on the more fragile settings of Mali and Burundi. For instance, the Director General of the Malian NSO was invited to present results in Parliament and, similarly, the Burundian statistical agency was invited to present results to senior decision-makers in government. In both cases, these presentations resulted in explicit government requests that the NSO repeat the survey to allow for the tracking of trends. Drawing exclusively from national resources, the Malian statistical office has implemented four rounds of the GPS-SHaSA annually since 2014, and the modules are now an integral part of the annual living conditions survey. With both countries invested in vast peacebuilding efforts, the use of GPS-SHaSA data for ‘early warnings’ of potential flashpoints was found to be particularly valuable. In Burundi, provincial governors specifically requested the statistical agency to disaggregate results by province to reveal regional discrepancies on various aspects of governance, and to guide peacebuilding interventions. Beyond the mere publication of GPS-SHaSA datasets, NSOs have highlighted the need to encourage independent research institutions to mine the datasets and to produce accessible policy briefs and data summaries on issues of public interest. However, concerns about the lack of a ‘data culture’ in government agencies continue to loom large over the prospects for GPS data uptake

176

Journal africain de statistiques, numéro 20, février 2018

5. Counting what counts: Africa’s seminal initiative on governance, peace and security statistics

by policymakers. This further highlights the important ambassadorial role played by individual members of Steering Committees on GPS Statistics, who can communicate on the initiative in their respective institution, and help identify ahead of time where the demand is for GPS data among their peers. The more in tune with country-specific policy priorities is the GPS statistical production cycle in any given country, the greater the chances that GPS data will find their way into decision-making fora. 6.

CONCLUSION

The year 2017 started with the AU Assembly of Heads of State and Government entrusting the APRM with an expanded mandate “to play a monitoring and evaluation role for the African Union Agenda 2063 and the United Nations Sustainable Development Goals Agenda 2030” (AU, 2017). In January 2018, the Assembly of African Heads of States and Government adopted the SHaSA II — Africa’s expanded Strategy for the Harmonization of Statistics for the period 2017-2026 — which reiterates the vital importance of GPS statistics charting Africa’s trajectory towards the 2063 horizon. This renewed emphasis on the APRM as an Africa-owned mechanism for self-assessment, combined with a renewed commitment to harmonize statistical production across the continent, offers an ideal conjunction of political will and technical means to help propel the institutionalization of GPS statistics across the continent. It is also a unique opportunity for Africa to reaffirm her global leadership role in promoting a nationally owned, scientifically robust and policy-useful approach to governance monitoring. Taking advantage of the groundwork laid by the African community of statisticians, the APRM could collaborate with NSOs to generate the data needed to report on SDG 16 and on Aspirations 3 and 4 of Agenda 2063. The GPS-SHaSA instruments also offer tested and proven solutions to the challenges of the earlier APRM experiment, by addressing some of its earlier shortcomings. These include the bias towards urban, more highly educated elites (because of the lack of a mechanism to capture a nationally representative sample of people’s voices) and the absence of a ‘light’ monitoring methodology to enable regular and cost-effective tracking of progress over time (Corrigan et al., 2017). The African statistical community may well be challenging the adage that “not everything that counts can be counted, and not everything that can be

The African Statistical Journal, Volume 20, February 2018

177

Marie Laberge, Yeo Dossina and Francois Rouband

counted counts.” Eleven African countries demonstrated that governance, peace and security ‘count’ in their national context (so much so that six of these countries self-financed the piloting of the GPS-SHaSA methodology, and have since continued to regularly produce GPS statistics), and contrary to common wisdom, that it can be ‘counted’. Concerted action by a critical mass of countries is now required to demonstrate the feasibility of mainstreaming GPS statistical production in the work program of the global statistical community. Fifty-five African countries could turn the tide. 6. REFERENCES

African Peer Review Mechanism /NEPAD (2003). APRM Base Document, Addis Ababa. African Union, UNECA and AfDB (2011). Strategy for the Harmonization of Statistics in Africa (SHaSA), Addis Ababa. African Union (2017). Decisions, Declarations and Resolution of the Assembly of the Union Twenty-Eight Ordinary Session, Addis Ababa. African Union and UNECA (2014). Common African Position on the Post2015 Development Agenda, Addis Ababa. African Union (2013). Agenda 2063 — The Africa We Want, Addis Ababa. African Union (2007). African Charter on Democracy, Elections and Governance, Addis Ababa. Arndt, C. and C. Oman (2006). “Uses and Abuses of Governance Indicators”. Paris: OECD Development Centre. AU Statistics Division (2012). Progress Report of the STG1 on GPS to the Committee of the Directors General of African NSOs. Yamoussoukro, Cote d’Ivoire. AU Statistics Division (2013). Progress Report of the STG1 on GPS to the Committee of the Committee of the Directors General of African NSOs. Johannesburg, South Africa. Bertelsmann Stiftung and Sustainable Development Solutions Network (2017). SDG Index & Dashboards Report, New York. Calvo, T., M.

178

Journal africain de statistiques, numéro 20, février 2018

5. Counting what counts: Africa’s seminal initiative on governance, peace and security statistics

Razafindrakoto and F. Roubaud (2018). “Are National Statistical Offices’ Governance Indicators Really Biased? A Comparison of Results from Afrobarometer surveys and NSOs household surveys in Sub-Saharan Africa”, DIAL Working paper, No.2018-04. Cling, J.-P., M. Razafindrakoto and F. Roubaud (2018). “SDG 16 on Governance and its Measurement: Africa in the Lead”, DIAL Working Paper no. 2018-02 [A first version of this paper was published in French: “L’ODD 16 sur la gouvernance et sa mesure. L’Afrique en tête”, Afrique Contemporaine, 2016/2, 258, pp. 73-93]. Corrigan T. and S. Gruzd (2017). “Can the APRM be an Effective Tool to Monitor Agenda 2063 and The SDGs?”, SAIIA Occasional Paper 251. Giang, D., T. K. V. Nguyen, ,T. H. Nguyen, M. Razafindrakoto, F.Roubaud, F. and M. Salomon (2011). Youth integrity in Vietnam: Piloting transparency international’s youth integrity survey, Cecodes, DIAL, Live, and Learn Transparency International, Towards Transparency, Hanoi. European Union (2017). Final Report of the Expert Group on Quality of Life Indicators, Statistical Reports, Luxembourg. Ferrin, M. and H. Kriesi (2014). “How Europeans View and Evaluate Democracy.” Comparative Politics, Oxford University Press. Herrera, J., M. Razafindrakoto and F. Roubaud (2008). “Poverty, Governance and Democratic Participation in Francophone Africa and the Andean Region”, OECD Journal on Development, vol. 9, no. 2, pp.99-118. Herrera, J., M. Razafindrakoto and F. Roubaud (2007). “Governance, Democracy and Poverty Reduction: Lessons Drawn from Household Surveys in Sub-Saharan Africa and Latin America”, International Statistical Review, vol. 75, no. 1,pp. 70-95. Institute for Economics and Peace (2016). Sustainable Development Goal 16 — From now to 2030: What is Needed to Measure Goal 16. Kaufmann, D., A. Kraay and M. Mastruzzi (2010). “The Worldwide Governance Indicators. Methodology and Analytical Issues”, World Bank Policy Research Working Paper 5430, Washington, D.C.: World Bank.

The African Statistical Journal, Volume 20, February 2018

179

Marie Laberge, Yeo Dossina and Francois Rouband

Kurtz, M.J. and A. Schrank (2007). “Growth and Governance: Models, Measures, and Mechanisms”, Journal of Politics vol. 69, no. 2, pp. 538-554. Government of Malawi (2012). Democratic Governance Sector Strategy 2013-17. Lilongwe, Malawi. Government of Kenya (2008). Kenya Vision 2030. Nairobi, Kenya ISTEEBU (2017). Donnees administratives GPS-SHaSA. Bujumbura, Burundi. Laberge M. (2016). Mission Report for AU/EU Pan-African Statistics (PAS) Program: Side-Event on GPS Statistics. Grand-Bassam, Cote d’Ivoire. Laberge M. and M. Orkin (2014a). Strategy for the Harmonization of Statistics in Africa (SHaSA): Governance Indicators Metadata. Addis Ababa: AUC and UNDP. Laberge M. and M. Orkin (2014b). Strategy for the Harmonization of Statistics in Africa (SHaSA): Peace and Security Indicators Metadata. Addis Ababa: AUC and UNDP. Mo Ibrahim Foundation (2012, 2014, 2016). Reports on the Ibrahim Index of African Governance. London. OHCHR (2016). A Human-Rights Based Approach to Data — Leaving No One Behind in the 2030 Development Agenda. Geneva. Orkin, M., M. Razafindrakoto and F. Roubaud (2015). “Governance, Peace and Security in the Strategy for the Harmonization of Statistics in Africa (GPS-SHaSA)”, Addis Ababa: AUC and UNDP. Razafindrakoto, M. and F. Roubaud (2015). “Les modules Gouvernance, Paix et Sécurité dans un cadre harmonisé au niveau de l’Afrique (GPS-SHaSA): développement d’une méthodologie d’enquête statistique innovante”, Statéco, no. 109, pp.122-158 [available in English: “The Governance, Peace and Security modules of the Strategy for the Harmonisation of Statistics in Africa (GPS-SHaSA): development of an innovative statistical survey methodology”, DIAL Working Paper no.2015-20].

180

Journal africain de statistiques, numéro 20, février 2018

5. Counting what counts: Africa’s seminal initiative on governance, peace and security statistics

Razafindrakoto, M., and F. Roubaud (2014a). Survey Manual for the SHaSA Survey Modules on Governance, Peace and Security. Addis Ababa: AUC and UNDP. Razafindrakoto, M. and F. Roubaud (2014b). Tabulation Plan for the SHaSA Survey Modules on Governance, Peace and Security. Addis Ababa: AUC and UNDP. Razafindrakoto M. and F. Roubaud (2006). “Governance, Democracy and Poverty Reduction: Lessons drawn from the 1-2-3 surveys in francophone Africa”, African Statistical Journal, vol. 2, May, pp. 43-82. Razafindrakoto, M. and F. Roubau (2010). “Are International Databases on Corruption Reliable? A Comparison of Expert Opinions Surveys and Household Surveys in Sub-Saharan Africa”, World development, August, vol. 38, no. 8, pp. 1057-1069. Razafindrakoto M. and F. Roubaud (2006). Etude sur les perceptions et attentes citoyennes concernant l’exercice du pouvoir à Madagascar. Collateral Creations, INCIPALS. Antananarivo: European Union. Razafindrakoto, M., F. Roubaud and E. Sentamba (2016). “Expériences, perceptions et aspirations citoyennes à l’aube de la crise au Burundi”, Revue Tiers-Monde, No. 228 (Octobre-Décembre), pp.67-100. Rotberg, R. (2014). On Governance: What It Is, What It Measures and Its Policy Uses. Ontario: Centre for International Governance Innovation. SDG16 Data Initiative (2017). 2017 Global Report. New York. Transparency International (2017). Monitoring Corruption and Anti-corruption in the Sustainable Development Goals: A Resource Guide. Berlin. UBOS (2014). Uganda National Governance Baseline Survey. Kampala, Uganda. UN (2014). A World That Counts: Mobilising the Data Revolution for Sustainable Development. Report prepared for the UN Secretary-General by the Independent Expert Advisory Group. New York. UNDP (2016). Final Report on Illustrative Work to Pilot Governance in the Context of the SDGs. New York

The African Statistical Journal, Volume 20, February 2018

181

Marie Laberge, Yeo Dossina and Francois Rouband

UNDP (2017). Voices from the Field: African Experiences in Producing Governance, Peace and Security Statistics — Recommendations for National Statistical Offices for Monitoring Goal 16 on Peaceful, Just and Inclusive Societies. Norway: UNDP Oslo Governance Centre. UNODC and UNECE (2010). UN Manual on Victimization Survey. Geneva.

182

Journal africain de statistiques, numéro 20, février 2018

S ND

FO AF RIC

EVELOPMEN T

EM ENT

ND CA

AF RI

D UN

BANQ UE

INE DE DEVE

EN T EM PP

A IC FR

LO AIN DE DEVE

African Statistical Journal Call for Papers The African Statistical Journal (ASJ) is currently accepting manuscripts for publication in French or/and English. The ASJ was established to promote the understanding of statistical development in the African region. It focuses on issues related to official statistics as well as application of statistical methodologies to solve practical problems of general interest to applied statisticians. In addition to individual academic and practicing statisticians, the Journal should be of great interest to a number of institutions in the region including National Statistical Offices, Central Banks, research and training institutions and sub-regional economic groupings, and international development agencies. The Journal serves as a research outlet and information sharing publication among statisticians and users of statistical information mainly in the Africa region. It publishes, among other things: • • • • •

articles of an expository or review nature that demonstrate the vital role of statistics to society rather than present technical materials, articles on statistical methodologies with special emphasis on applications, articles about good practices and lessons learned in statistical development in the region, opinions on issues of general interest to the statistical community and users of statistical information in the African region, notices and announcements on upcoming events, conferences, calls for papers, and recent statistical developments and anything that may be of interest to the statistical community in the region.

All manuscripts are reviewed and evaluated on content, language, and presentation. The ASJ is fully committed to providing free access to all articles as soon as they are published. We ask you to support this initiative by publishing your papers in this journal. Prospective authors should send their manuscript(s) to ASJ-Statistics@afdb.org The ASJ is also looking for qualified reviewers. Please contact us if you are interested in serving as a reviewer. For instructions for authors and other details, please visit our website – http:// www.afdb.org/en/knowledge/publications/african-statistical-journal/

The African Statistical Journal, Volume 20, February 2018

183

S ND

FO AF RIC

EVELOPMEN T

EM ENT

ND CA

AF RI

D UN

BANQ UE

INE DE DEVE

EN T EM PP

A IC FR

LO AIN DE DEVE

Journal africain de statistiques Demande de soumission d’articles Le journal africain de statistiques (JSA) accepte actuellement des manuscrits pour la publication en anglais ou en français. Le JSA a été établi pour favoriser la compréhension du développement statistique dans la région africaine. Il se concentre sur des questions liées aux statistiques officielles aussi bien que l’application des méthodologies statistiques pour résoudre des problèmes pratiques d’intérêt général pour les praticiens de la statistique. En plus des universitaires et des statisticiens de métier, le Journal devrait revêtir un grand intérêt pour les institutions de la région, notamment les offices nationaux de statistiques, les banques centrales, les instituts de recherche et les organisations économiques sous-régionaux et les agences internationales de développement. Le Journal constitue un document de recherche et d’information entre les statisticiens et les utilisateurs de l’information statistique, principalement dans la région africaine. Il publie entre autres: • • • • • •

des articles sur le plaidoyer en matière de statistique qui démontrent le rôle essentiel des statistiques dans la société, plutôt que de présenter le matériel technique, des articles sur les méthodologies statistiques, avec un accent particulier sur les applications, des articles sur les meilleures pratiques et les leçons tirées sur le développement de la statistique dans la région, des avis sur des questions d’intérêt général pour la communauté statistique et les utilisateurs de l’information statistique dans la région africaine, des informations et des annonces sur les prochains événements, les conférences, les appels à contribution pour des papiers, et les développements statistiques récents et tout autre aspect susceptible d’intéresser la communauté statistique dans la région.

Tous les manuscrits sont passés en revue et évalués sur le contenu, la langue et la présentation. Le JSA s’engage entièrement à fournir le libre accès à tous les articles dès qu’ils sont publiés. Nous vous demandons de soutenir cette initiative en publiant vos articles dans ce journal. Les auteurs éventuels devraient envoyer leur manuscrit(s) à ASJ-Statistics@afdb.org

184

Journal africain de statistiques, numéro 20, février 2018

Le JSA cherche également les critiques qualifiés. Veuillez nous contacter si vous êtes intéressé à contribuer en tant que critique. Veuillez visiter notre site Web http://www.afdb.org/en/knowledge/ publications/african-statistical-journal/ pour les instructions aux auteurs et autres détails.

The African Statistical Journal, Volume 20, February 2018

185

Editorial policy The African Statistical Journal (ASJ) was established to promote the understanding of statistical development in the African region. It focuses on issues related to official statistics as well as application of statistical methodologies to solve practical problems of general interest to applied statisticians. Of particular interest will be the exposition of: how statistics can help to illuminate development and public policy issues like poverty, gender, environment, energy, HIV/AIDS, etc.; development of statistical literacy; tracking national and regional development agendas; development of statistical capacities and effective national statistical systems; and the development of sectoral statistics, e.g. educational statistics, health statistics, agricultural statistics, etc. In addition to individual academic and practicing statisticians, the Journal should be of great interest to a number of institutions in the region including National Statistical Offices, central banks, research and training institutions, sub-regional economic groupings, and international development agencies. The Journal serves as a research outlet and information sharing publication among statisticians and users of statistical information mainly in the African region. It publishes, among other things, articles of an expository or review nature that demonstrate the vital role of statistics to society rather than present technical materials, articles on statistical methodologies with a special emphasis on applications, articles about good practices and lessons learned in statistical development in the region, opinions on issues of general interest to the statistical community and users of statistical information in the African region, notices and announcements on upcoming events, conferences, calls for papers, and recent statistical developments and anything that may be of interest to the statistical community in the region. The papers, which need not contain original material, should be of general interest to a wide section of professional statisticians in the region. All manuscripts are peer reviewed and evaluated on content, language and presentation.

186

Journal africain de statistiques, numĂŠro 20, fĂŠvrier 2018

Ligne éditoriale Le Journal statistique africain a été établi pour favoriser la compréhension du développement statistique dans la région africaine. Il se concentre sur des questions liées aux statistiques officielles aussi bien que l’application des méthodologies statistiques pour résoudre des problèmes pratiques d’intérêt général pour les statisticiens de métier. L’intérêt particulier est de montrer comment les statistiques peuvent aider à mettre en exergue les problèmes de développement et de politique publique tels que la pauvreté, le genre, l’environnement, l’énergie, le VIH/ SIDA, etc.; le développement de la culture statistique ; la prise en compte des questions de développement régional et national; le développement des capacités statistiques et des systèmes statistiques nationaux efficaces; et le développement des statistiques sectorielles comme les statistiques d’éducation, de santé, des statistiques agricoles, etc. En plus des universitaires et des statisticiens de métier, le Journal devrait revêtir un grand intérêt pour les institutions de la région, notamment les offices nationaux de statistiques, les banques centrales, les instituts de recherche et les organisations économiques sous-régionaux et les agences internationales de développement. Le Journal constitue un document de recherche et d’information entre les statisticiens et les utilisateurs de l’information statistique, principalement dans la région africaine. Il publie entre autres: des articles sur le plaidoyer en matière de statistique qui démontrent le rôle essentiel des statistiques dans la société plutôt que la présentation des outils techniques, des articles sur les méthodologies statistiques, avec un accent particulier sur les applications, des articles sur les meilleures pratiques et les leçons tirées de la région, des avis sur des questions d’intérêt général pour la communauté statistique et les utilisateurs de l’information statistique dans la région africaine, des informations et des annonces sur les prochains événements, les conférences, les appels à contribution pour des papiers, et les développements statistiques récents et tout autre aspect susceptible d’intéresser la communauté statistique dans la région. Les articles, qui n’ont pas besoin de contenir du matériel original, devraient intéresser une grande partie des statisticiens professionnels dans la région. Tous les manuscrits seront passés en revue et évalués sur le contenu, la langue et la présentation.

The African Statistical Journal, Volume 20, February 2018

187

Guidelines for manuscript preparation and submission Submissions Manuscripts in English or French should be sent by email to ASJ-Statistics@ afdb.org Title The title should be brief and specific. The title page should include the title, the author’s name, affiliation and address. The affiliation and address should be given as a footnote on the title page. If the manuscript is co-authored, the same information should be given for the co-author(s). Abstract, Key Words, and Acknowledgments A short abstract of about 150 words must be included at the beginning of the manuscript, together with up to 6 key words used in the manuscript. These key words should not repeat words used in the title. Acknowledgments, if any, should inserted as a new section at the end of the paper and before the References. Sections and Numbering Major headings in the text should be numbered (e.g. 1. INTRODUCTION). Numbered subheadings and sub-subheadings (e.g. 1.1 The establishment of the NSDS and 1.1.1 Bodies comprising the NSDS). Main body text in the form of paragraphs should not be numbered. Formatting Please use minimal formatting as this will facilitate harmonization of all the papers. As your default, keep to “normal” (12 pt. Times New Roman) for main text with a single line space between paragraphs. Do not apply “body text” as an inbuilt style. The levels of heading need to be easily identifiable. We recommend all capitals bold for the first level of heading in the main text (e.g. “1. INTRODUCTION”); thereafter bold upper and lower case for subheadings (e.g. “1.1 The establishment of the NSDS”) and italic not bold (e.g. 1.1.1. Creating a culture of cooperation) for sub-subheadings. Please refer to the latest volume of the AJS as a guide. House Style The Bank’s house style is U.S. rather than British spellings (e.g. “organization” not “organisation”; “program” rather than “programme”, “analyze” rather than “analyse” etc.). Use percent rather than per cent or % although the percentage sign should be used in figures and tables and double rather than single quotation marks. Dates should be U.S. style (e.g. December 11, 1985 not 11 December 1985).

188

Journal africain de statistiques, numéro 20, février 2018

Tables and Figures Tables and figures should be numbered and given a title. These should be referred to in the text by number (e.g. “See Table 1”), not by page or indications such as “below” or “above”. Equations Any equations in the paper should be numbered. The numbers should be placed to the right of the equation. References A list of references should be given at the end of the paper (to precede the Annexes, if included). The references should be arranged alphabetically by surname/name of organization. Where there is more than one publication listed for an author, order these chronologically (starting with the earliest). The references should give the author’s name, year of publication, title of the essay/book, name of journal if applicable. Use a, b, c, etc. to separate publications by the same author in the same year. Titles of journals and books should be in italic; titles of working papers and unpublished reports should be set in double quotation marks and not italicized. Examples: Fantom, N. and N. Watanabe (2008). “Improving the World Bank’s Database of Statistical Capacity,” African Statistical Newsletter, 2 (3): 21–22. Herzog, A. R. and L. Dielman (1985). “Age Differences in Response Accuracy for Factual Survey Questions,” Journal of Gerontology, 40: 350–367. Kish, L. (1988a). “Multipurpose Sample Designs,” Survey Methodology, 14 (3): 19–32. Kish, L. (1988b). “A Taxonomy of Elusive Populations,” in Proceedings of the Annual Meeting of the American Statistical Association. January 1988. Parpart, J. L., M. P. Connelly, and V. E. Barriteau (eds.) (2000). Theoretical Perspectives on Gender and Development. Ottawa: International Development Research Center. World Bank (2006). Statistical Capacity Improvement in IDA Countries – Progress Report. Washington, DC: The World Bank.

The African Statistical Journal, Volume 20, February 2018

189

Cross References In the main body of the article, cross-references should be Harvard-style, e.g. (Kish, 1988a; Herzog and Dielman, 1985: 351). For cross-references to three or more authors, only the first surname should be given, followed by et al., although the names of all the authors must be provided in the References entry itself. Abbreviations ibid. and op. cit. should be avoided.

190

Journal africain de statistiques, numĂŠro 20, fĂŠvrier 2018

Instructions pour la préparation et la soumission de manuscrits Soumission Les manuscrits en anglais ou en français doivent être envoyés à : ASJ-Statistics@afdb.org Titre Le titre devrait être bref et détaillé. La page de titre doit inclure le titre du papier, le nom de l’auteur, l’affiliation et l’adresse. L’affiliation et l’adresse doivent figurer comme note de bas de page. Si le manuscrit est produit par des coauteurs, la même information doit être donnée pour les coauteurs. Résumé, mots clés et remerciements Un résumé court d’environ 150 mots doit être inclus au début du manuscrit ainsi qu’environ 6 mots clés utilisés dans le manuscrit. Les mots clés ne doivent pas répéter les mots utilisés dans le titre. Les remerciements, s’il y en a, doivent être insérés à la fin de l’article, avant les références bibliographiques. Section et numérotation Les principaux titres doivent être numérotés (par exemple “1. INTRODUCTION“). Les sous-titres et sous sous-titres numérotés (par exemple “1.1 L’élaboration de SNDS” et “1.1.1 Créant une culture de coopération”) peuvent être employés. Le corps principal du texte sous forme de paragraphes ne devrait pas être numéroté. Formatage Veuillez utiliser le formatage minimal car ceci facilitera l’harmonisation de tous les articles. Garder par défaut le format “normal” (12 pt. Times New Roman) pour le texte principal avec l’espace d’une seule ligne entre les paragraphes. Ne pas appliquer le “corps de texte “ en tant que modèle intégré. Les niveaux du titre doivent être facilement identifiables. Nous recommandons les majuscules en gras pour le premier niveau titre dans le texte principal (par exemple “1. INTRODUCTION“) ; ensuite les lettres minuscules en gras pour les sous-sections (par exemple “1.1 l’élaboration de la SNDS”) et ensuite l’italique pour les sous sous-titres (par exemple “1.1.1 Créant une culture de coopération”). Veuillez vous référer au dernier volume du JSA comme guide. Tables and Figures Les tableaux et les graphiques doivent être numérotés et comporter un titre. Ceux-ci devraient être mentionnés (par exemple “voir Tableau 1” ) dans le texte par le nombre correspondant, et non par une indication de page ou par d’autres indications telles que “ci-dessous” ou “au-dessus de”.

The African Statistical Journal, Volume 20, February 2018

191

Équations Toutes les équations dans le papier doivent être numérotées. Les nombres doivent être placés à la droite de l’équation. Références bibliographiques Une liste de références doit être fournie à la fin de l’article (avant les annexes, le cas échéant). Les références doivent être classées par ordre alphabétique selon le nom de l’auteur ou de l’organisation. Là où il y’a plus d’une publication listée pour un auteur, elles doivent être classées chronologiquement (en commençant par les plus récents). Les références doivent donner le nom de l’auteur et l’année de publication, le titre du livre, le nom du journal le cas échéant. Utiliser a, b, c, etc. pour séparer les publications du même auteur au cours der la même année. Les titres des journaux et des livres devraient être en italique ; les titres des documents de travail et des rapports non publiés devraient être placés dans de doubles guillemets et ne pas être imprimés en italique. Exemples : Fantom, N. and N. Watanabe (2008). “Improving the World Bank’s Database of Statistical Capacity,” African Statistical Newsletter, 2 (3): 21–22. Herzog, A. R. and L. Dielman (1985). “Age Differences in Response Accuracy for Factual Survey Questions,” Journal of Gerontology, 40: 350–367. Kish, L. (1988a). “Multipurpose Sample Designs,” Survey Methodology, 14 (3): 19–32. Kish, L. (1988b). “A Taxonomy of Elusive Populations,” in Proceedings of the Annual Meeting of the American Statistical Association. January 1988. Parpart, J. L., M. P. Connelly, and V. E. Barriteau (eds.) (2000). Theoretical Perspectives on Gender and Development. Ottawa: International Development Research Center. World Bank (2006). Statistical Capacity Improvement in IDA Countries – Progress Report. Washington DC: The World Bank. Renvois Dans le corps principal de l’article, les renvois devraient suivre le modèle de Harvard, par exemple (Kish, 1988a ; Herzog et Dielman, 1985 : 351). Pour des renvois à trois auteurs ou plus, seulement le premier nom de famille devrait être donné, suivi par et al., bien que les noms de tous les auteurs

192

Journal africain de statistiques, numéro 20, février 2018

doivent être fournis dans la Bibliographie elle-même. Les abréviations ibid. et op. cit. ne devraient pas être employées dans le texte ou dans les notes de bas de page.

The African Statistical Journal, Volume 20, February 2018

193

African Statistical Journal Journal africain de statistiques

African Statistical Journal / Journal africain de statistiques

ENT

S ND

AF RIC

P LO AIN DE DEVE

2233 2820

Volume 20 – February / février 2018

ELOPME

PE M

DEV

B A N QU EA F

E DE DEV EL

ENT EM PP

IN CA RI

5. Counting what counts: Africa’s seminal initiative on governance, peace and security statistics Marie Laberge, Yeo Dossina and Francois Rouband

Volume 20 – February / février 2018 African Development Bank Group Groupe de la Banque africaine de développement