Empirical Economic Bulletin, Spring 2008, Vol. 1 by Bryant University

EEB--UNDERGRAD UATE ECONOMICS JOURNAL

EMPIRICAL ECONOMIC BULLETIN

THE CENTER FOR G LOBAL AN D REGIONAL ECONOMI C STUDIES B RYANT UNIVERSITY

..---....

�·��

Spring 2008, Volume 1

EDITOR: Ramesh Mohan DESIGN & LAYOUT: Ramesh Mohan and Rebecca Marcus MISSION STATEMENT: The Empirical Economics Bulletin is the undergraduate journal for the Department of Economics, Bryant University. The journal is produced in conjunction with the Bryant Economic Undergraduate Symposium. The focal point of the Symposium is on training undergraduate students in the art of writing, presenting, and publishing empirical research papers on a range of socio-economic and economic topics. The Symposium’s primary emphasis is empirical studies with policy relevance. Students are then able to publish their empirical papers in the Empirical Economics Bulletin. An objective of the Economics’ Department, Bryant University is to train students to conduct quantitative economic data analysis and to present the results in a coherent and meaningful way. This objective is met through having the Symposium and the publication of the Empirical Economics Bulletin. The first issue was in Spring 2008 and has been published annually with original work from an array of student authors. SUBMISSION GUIDELINES: Students may submit their socio-economic and economic work here: Ramesh Mohan at rmohan@bryant.edu. Limit one submission per author. Each submission should have a title page with the title; name of author; abstract; keywords; JEL classification; author’s email. Previously published work is not accepted. The reading period is September 1 to December 1. Copyright reverts to author upon publication. Any questions may be directed to Professor Ramesh Mohan at rmohan@bryant.edu. © 2008 Empirical Economics Bulletin

Table of Contents Differing Immigrant Wages: Causes of Differences in Immigrants Standards of Living, Sahana Zutshi .... . 5 Determinants of Gasoline Price: Can Consumer Spending Make a Difference?, Jonathan Stachelek .... 20 Causality Relationship Between Foreign Exchange Rates and Stock Market Close: Evidence in Singapore, Jeffrey Allen Shew, Jr. ………………….……………………………………………………………………………………………………… 33 Factors Affecting Health Care Spending, Jennifer Resende .………………………………………………………………… 44 The Importance of Governance to a Stable and Developed Economy, Peter Litvinenko ……………………… 56 An Analysis of the Female Labor Force Participation Rate in the U.S. 1980-2004, Larry Martins ………… 69 Economic and Social Factors that Influence Life Expectancy and Infant Mortality, Kenneth Smith ……… 82 The Interaction between the Stock Market, Monetary Policy and Inflation in Singapore and Malaysia, Lindsey Kahler ………………………………………………………………………………………………………………………………….. 102 The Impact of Sectoral Performance on the Stock Market: Does Volatility Equal Explanatory Power?, Jeffrey Haydock ………………………………………………………………………………………………………………………………. 117 United States Homeownership Rates: The Effect of Macroeconomic Factors on the Domestic Real Estate Market, Alexander N. Grande …………………………………………………………………………………………………………… 133 Causality Between Defense Spending GDP and Economic Growth, Ryan P. Daley ……………………………. 150 U.S. Current Account: Why Is It Increasingly Negative?, Joshua Champagne ……………………………………. 169 The Social and Behavioral Factors That Affect Obesity in OECD Countries, Sanjana Desai …………………. 183

Differing Immigrant Wages: Causes for Differences in Immigrants Standards of Living.

Sahana Zutshia

Abstract This paper looks at the relationship between immigration workers and their differing standards of living based on various variables. The variables chosen were gender, place of origin/ethnicity, number of years living in the United States, education level and finally, the age of the immigrants. The model incorporates all these variables and then finds the connection between them and the earning status of the immigrants. The results show that gender and ethnicity often impact the earnings negatively whereas age, number of years living in the US and education level impact immigrant earnings positively.

JEL Classification: J70 Keywords: Immigration, earnings

Department of Economics, Bryant University, 1150 Douglas Pike, Smithfield, RI, 02917. Phone: 203-313-2389 E-mail: szutshi@bryant.edu

Introduction Immigration has always been a very controversial topic and has become even more

recently. Enchautegui, 1998 describes immigrants as “foreign born persons whose original citizenship is that of another country”. Immigration is the relocation of these foreign born citizens to another country. In the long history of immigration most of the emphasis has been on illegal immigration and its effects on the host society. In contrast, this paper looks at immigrants and how their living standards compare to each other, and not citizens. Immigrants make up a significant part of the population, and like citizens, their living standards differ drastically. The paper is to see living standard comparisons among immigrants once they reach the country based on different factors. The variables looked at will be age, gender, ethnicity, time spent in the US and education level. These variables were taken from various studies over the years which are mentioned in the literature review. The point of the paper is to look at the variables and their effects on the earnings status of immigrants. It is also to look at the effects of the variables on each other when placed in the same model. This study aims to enhance understanding of the living standards of immigrants that are here legally and illegally and the factors that affect those living standards. Further valuation of the variables and their importance will give an idea of which one of the variables is most influential to determining earnings, which in this paper is the medium to determine standard of living. The topic is important because it can be used while evaluating current immigrant policy and standard of living among immigrants. Research can be furthered by using the results of this paper and then comparing it to a study of the standard of living of citizens and the factors that

influence those standards. It would be interesting to see the differences between the living standards of immigrants and citizens and the various factors that affect those standards. The study is different from others because there have been no papers that have looked at immigration living standards with these specific variables. There have been papers with variables that are listed in this paper, but no paper in this field has these particular variables together and seen their impact on each other and earnings. The rest of the paper is prepared in five sections. The second part is the historical trends of immigration and gives the reviewers background on the subject. The third part is the literature review, which goes into the articles for this subject and their effects on the economic model. The fourth part of the paper estimates the empirical model and the signs. The fifth section discusses the results and methodology. This section also contains the data and estimation methods analysis. Finally the sixth section is the conclusion, in which the significance of the data is discussed, as well as the purpose of the final model. II)

History and Trends of Immigration

Immigrants have always been attracted by the religious and political freedoms and the great economic opportunities available in the US, thus immigration has always been an important part of the country. Early immigration peaked between 1892-1924 but after that limits began to be passed on how many people could come into the U.S from specific countries. The limits were based on the number of people from that country already living in the U.S. In 1965 immigration quotas were established according to who applied first and preference was given to people with specific skills or relatives of citizens, a trend which continues till today. In 1978 Congress abandoned hemispheric quotas and established a worldwide ceiling, which is the practice 7

currently. The U.S accepts more immigrants than any other developed country, with 11.1% of the country’s population being foreign born. Below are charts showing immigration trends in the U.S thus far in its history.

Immigration Trends from 1820-1980

1 2

www.usimmigration.com T21.ca

III)

Literature Review

There is a fair amount of literature on the subject of legal immigrants in the academic field. This depth, however, is miniscule compared to the amount of research done on illegal immigrant. Borgas (1994), asked three main questions in his study: the first was the immigrants economic performance in the host country; the second question looked at the impact immigrants had on native employment opportunities; the third question looked at which immigration policy benefited the host country the most. According to Borgas (1994), immigration was shown to have fluctuated dramatically in this past, which is consistent with the trends chart above. The author also talked about how the ethnic origins of the immigrant population have also changed with changes in immigration policy and the impact this could have on policy in the future. Of major interest was the fact that the author contended that the gap between immigrant wages and native wages had actually increased from 27.6% in 1980 to 31.7% in 1990. The study attributed this to lower skill level immigrants coming into the U.S and taking cheaper jobs. While Borgas, claimed that immigration wages were actually going down Enchautegui (1998) gave concrete proof to the claim. The study showed that between 1980 and 1994, the number of immigrants without a high-school diploma almost doubled (Enchautegui, 1998). This accounted for Borgas’s research, which showed the gap in wages between immigrants and natives widening between these two time periods. In the study Enchautegui (1998) noted that despite the picture painted of the incoming immigrants, many of the low skilled immigrants that come into the U.S had poverty rates that increase faster than those of the natives, despite relatively solid employment history. The author also made note of the fact that although education was an important determinant of economic success it continued to be the main divider between skilled and low-skilled workers. 9

Guillermina Jasso et al (1986), talked further about education especially schooling and language skills of new immigrants and their gains from immigration. The authors stressed the importance of the fact that a key attribute of immigrants is their human capital, one measure of which is years of schooling. The study also claimed that the data showed that there are far more legal immigrants at the top of the education pyramid than at the bottom, which supports the theory that many illegal immigrants arrive here for low-paying jobs that many Americans cannot or will not do. It also shows that legal immigrants come to plug in the gap between the demand for certain types of jobs and the lack of native born supply, such as engineers or physicists (Guillermina Jasso et al 1986). In this study the wages for immigrants were lower than the wages for native born, a fact that remained consistent throughout all authors. The study gives the explanation that this is because many of the immigrants are new, young entrants to the U.S labor market. This study differed from the previous two by showing that new legal immigrants are better schooled than native-born, and that they gain substantial economic advantage by immigrating. One thing to note is that none of the above studies talked about race in their findings. However Waters & Eschbach (1995) contend that in their study they found that employers still use racial and ethnic cues in hiring, which in turn impedes immigrants more than native born. In the study they also found evidence on the substitution on the labor markets of immigrants for native born. Almost as a contradiction to Waters and Eschbach, Bernt Bratsbreg et al (2002) talked in detail about how with naturalization young male immigrants gained access to public-sector, whitecollar, union jobs, and wage growth accelerates that were consistent with removal of employment barriers. In conjugation with wages, Harriet Orcutt Duleep and Mark C. Regets talked about idea of how fast wages were growing and the reason why. In the study the

researchers pointed out that with their research they found out that immigrant wages grow faster than native born wages. However, they did concede that this could be due to potential lower entry wages for immigrants. Duleep and Regets showed that immigrant entry wages usually started lower than the wages of native born but their growth rate was 6.7% compared to native workers 4.4% wage increase. Alan K. Simpson provided in depth background knowledge on the history of immigration, necessary to understand the results of the model better. Overall the author was very informative. The study talked about how the U.S is a major target of immigration due to stagflation in many countries and the opportunities given here (Simpson, 1984). Since the article was written in 1984, there were a great many changes to the policy after, which made it a little less useful than it would have been if it had been written a little later. The author also talked about the different measures institituted to ensure legal employment. All in all, the paper was informative, but in the context of my research it provided more historical background and understanding of my model, and not actual concrete evidence and facts that could be used. Finally Joseph Schaafsma and Arthur Sweetman claimed age was an important factor in wages. The paper was done in Canada, and it was hoped that the results would be replicated in the U.S. As you will see, further into the paper, while the results were not identical the implications were the same. Schaafsma and Sweetman showed evidence that support three underlying scores of the “age” effect: work experience in the “source” country “yields virtually no return in the host country; second, the return to education varies with age at immigration, and, finally, an acculturation effect is observed for immigrants who are visible minorities or whose mother tongue is not English. Further, it is found that educational attainment, and relatedly earnings, varies systematically

across age at immigration with those arriving around age 15 to 18 obtaining fewer years of education” (Schaafsma & Sweetman, 2001). 4.1) Empirical model and Reasoning: Variables, signs and definitions: 𝐸𝑎𝑟𝑛𝑖𝑛𝑔𝑠 = 𝛽0 + 𝛽1 𝐺𝑒𝑛𝑑𝑒𝑟 + 𝛽2 𝐸𝑡ℎ𝑛𝑖𝑐𝑖𝑡𝑦 + 𝛽3 𝑌𝑒𝑎𝑟𝑠 𝑖𝑛 𝑡ℎ𝑒 𝑈𝑆 + 𝛽4 𝐸𝑑𝑢𝑐𝑎𝑡𝑖𝑜𝑛 + 𝛽5 𝐴𝑔𝑒

The dependant variable in the equation is earnings. In this case earnings are the medium for standard of living of immigrants. Earnings are defined as the pay the immigrant is given, including any options or stocks that is part of the salary. There are five independent variables chosen due common theories and research into previous studies. The first variable is the gender of the immigrant; the second variable is the ethnicity or place of origin of the immigrant. Education and age stand for the education level and age level of the immigrant respectively. Years in the US are the number of years the immigrant has spent in the United States. The explanations behind the signs of the model are as follows: For age, theoretically the greater the age of the individual, the greater their experience. This in turn would usually translate into a higher paying job and thus a higher standard of living. That is why the sign for age is positive: it makes sense since it is likely that a 35 year old computer engineer will earn more than a 20 year old student. For years in the U.S the theory is the longer a person has worked here, the greater their chances of getting a better paying job comparatively. Therefore the sign for this variable is positive as well. For gender the explanation is a little different. In theory if the gender of the immigrant is male then he should be paid higher wages than a female immigrant. However if the emphasis is on the female immigrant then the sign should be negative. Taking into consideration both these circumstances the author has decided to let the 12

sign in the model be positive, because it is believed that more male immigrants are coming in because of their labor skills and thus the sign is likely to be positive since male immigrants are paid more that female immigrants. In other words, since males are paid more than females and since it is the theory that there are more male immigrants than female immigrants, than in this study which compares the standard of living for immigrants only, the sign should be positive overall, although it can be negative for the female sub-variable. In the case of ethnicity it is the author’s belief that ethnicity will be a positive sign since it is people of favorable ethnic status in the States that immigrate more here and these people are given higher earnings. Thus the sign for ethnicity remains positive. The table with a summary of the variables, their descriptions and signs is below. Table 1: Variable Description and expected signs Variable

Description

Expected Sign

Data Source

Gender

The sex of the immigrant: in the regression for this variable 0 is for Male and 1 is for females The place of origin/race of the immigrant. Race 1: 1: Asian and White while 0 is Hispanic. Race 2: 1: Hispanic and 0 is Asian and White

+; the theory is that more men than women are entering the U.S; thus the sign is positive

US Census Bureau

+; More immigrants from developed countries therefore a positive sign

US Census Bureau

Age

The Age of the immigrant at the time of arrival

+; The older the immigrant is, the more experienced they are, thus they make more

US Census Bureau

Education

The education degree the immigrant has: Educ: 0: HS; 1:

+; the more time the immigrant spends in the states, the better their chances for

US Census Bureau

Ethnicity

College and Post grad: Educ1: 0: College and Post and 1: HS. The years the immigrant has spent in the US

Years in the U.S

greater pay via permanent visas.

+; the higher the education level, the better the chance of the immigrant to be paid more.

US Census Bureau

4.2) Data: The data for this paper was obtained from the U.S Census Bureau with 148,048 total observations. The data was provided by Professor Jongsung Kim from Bryant University, in Smithfield, RI. The explanation of the data is below. The data is collected every 10 years, the last collected being in 1999. The next data set shall be collected in 2009, and will be distributed with results in 2011. Table 2: Summary Statistics Variable

Std. Error

T-Statistic

P>T

Confidence Interval

Gender

Women: 201.2048 Men: 100.6024 Race 1:445.1174 race2 | 382.693

Women: 95.65 Men: -95.65

W: 0.000 M: 0.000

W:-19640.09 -18851.38 M: 9820.046 9425.688

R1: 20.76 R2: -24.13

R1: 0.000 R2: 0.000

R1: 8370.431 10115.27 R2: -9982.771 --8482.629

Educ: 976.1856 educ1 | 962.6571

Educ: 27.91 Educ 1: 23.53

Educ: 0.000 Educ 1: 0.00

Educ: 25332.8529159.45. Educ 1:

Ethnicity

Education

24537.21 20763.63

Age

34.03381

127.84

0.00 4284.1034417.514

No. Of years in US

9.006215

16.40

0.00 130.0175165.3215

5.0 Empirical Results: Variable3

Gender

Ethnicity

Education

Age

No. Of years in US

Table 3: Regression Results Coefficient Std. Error R2

No of Obs

Men: 9622.867** Women: 19245.73** Race 1: ***9242.853 Race 2***:9232.7 Educ***: 27246.15 educ1***:22650.42 147.6695***

Men: 100.60 Men:.0582 Women: 201.20 Women:.0582

148048

Race1:445.1174 .0090 Race2: 382.693

148048

Educ: 976.1856 Educ1: 962.65

.0984

148048

9.006

.157

148048

1.26***

.155

.157

148048

Please note that *** denotes 1% significance and ** denote 5% significance

5.1: Explanation As you can see from the regression above while most of the variables yielded the expected signs, the variables for sex and ethnicity did not. Both men and women were negative, women more so than men, which was expected. In retrospect, part of the explanation for this sign is simple. While immigrant men do make more money than immigrant women, compared to citizen men, they still make comparatively less wages. The same explanation goes for immigrant women, hence the reason why their wages are even lower than immigrant men. Most immigrant women coming into the country do not typically go into high paying positions, which would explain why their wages are so much lower than immigrant men. However since the comparison was between immigrants only, further research is required to understand the problem. The author is unsure about the negative sign with ethnicity. A likely explanation is the fact that many of the immigrants coming into the country are Hispanic; since the regression shows that on average Hispanic immigrants make less than white and Asian immigrants, the overall sign would be negative. Barring that, the author feels that further research is needed for the subject. Of interest to the author was the coefficient for the number of years living in the U.S. Regarding the result, it would seem that number of years living does not make a significant difference in the earnings of immigrants- an interesting factor given the article on naturalization and how it increases and immigrants wages. These results contradict that data and further research is suggested to get more accurate results. Education was also an interesting result since it showed that many immigrants that came to the U.S with only a high-school diploma fared much worse than those that came who had graduated from college or done post-doctorate work: it coincides with the article reviewed that states that education has a significant impact on earnings. Finally we also

see that while age matter, as was claimed in the Sweetman article, compared to all the other factors, its influence on earnings is relatively low. The above regression also shows that all the factors chosen were important in the equation- they all play a significant part in the increase or decrease of earnings. All the factors fall into the 1% or 5% range which means that there is a 99% or 95% chance that they play a role in the amount of earnings an immigrant makes. Since all the variables are important, this makes the accuracy of the regression more viable. 6.0: Conclusion In today’s day and age immigration is a highly contested issue. However much of the discussion is about illegal immigration and its effects on society and citizens. Not much attention is given to the immigrants themselves and their standards of living compared to each other. This study looks at that part of the population in comparison to each other with their standards of living. The results show that age, gender, ethnicity, education and years of living in the U.S are all significant factors with regards to differences in earnings. They also show that unlike the hypothesized equation, the actual signs for gender and ethnicity are negative, with a misinterpretation of information being the reason. However further research is suggested for more concrete results. The results also show that immigrants are more likely to have higher wages if they are white or Asian, than if they are Hispanic. They also show that immigrants with higher degrees of education earn, on average, far more than those immigrants that come into the U.S with only high school diplomas. All in all the regression is considered accurate and viable and will hopefully further our understanding of this particular subject. This study is different than other studies since it takes many factors that have been specifically studied, such as age and

education and then adds in other factors that should be theoretically relevant such as gender and age. The study shows the effects the variables have on each other as well as on the dependant variable. For example, in Sweetman’s study, age was considered a very important factor. This study shows that age, while significant does not have a significant impact compared to gender and ethnicity. Also, in contradiction to Bernt Bratsbreg, the study shows that naturalization/Years in the US, do not play substantial roles compared to other factors. Bibliography Borgas, George J, (1994), "The Economics of Immigration." The American Economic Review 32: 305-308. Brent, Bratesberg, James F. Ragan, and Zafar M. Nasir, (2002) "The Effect of Naturalization on Wage Growth: a Panel Study of Young Male Immigrants." Industrial and Labor Nations Review 20 : 568-597. Duleep, Harriet O., and Mark C. Regets, (1996), "Are Lower Immigrant Earnings At Entry Associated with Faster Growth?" The Canadian Journal of Economics 29: 130-134. Enchautegui, Maria, (1998), "Low Skilled Immigrants and the Changing American Labor Market." Population and Development Review 24 : 811-824. Jasso, Guillermina, and Mark R. Rosenzweig, (1986) "Immgration and the Family." Journal of Labor Economics 23: 291-311. Schaafsma, Joseph, and Arthur Sweetman, (2001), "Immigration Earnings: Age At Immigration Matters." The Canadian Journal of Economics 34: 1066-1099.

Simpson, Alan K, (1984), "The Politics of Immigration Reform." International Migration Review 18: 486-504. Waters, Mary C., and Karl Eschbach, (1998), "Immigration and Ethnic and Racial Inequality in the United States." Annual Review of Sociology 21 : 419-446.

Determinants of Gasoline Price: Can Consumer Spending Make a Difference? Jonathan Stacheleka

Abstract: This paper investigates the determinants to gasoline prices in the United States. The regression contains five different independent variables: the price of imported oil per barrel, the number of barrels of oil imported to the United States, the number of barrels of oil produced domestically in United States, the commodity price of oil, the number of automobiles purchased monthly in the United States, and the dependent variable, Gasoline Price. By using these variables in a linear regression model, the results show that the price of gasoline is primarily driven by the price of imported crude oil, followed shortly thereafter by the domestic oil produced.

Department of Economics, Bryant University, 1150 Douglas Pike, Smithfield, Rhode Island, 02917. Phone: 860-874-1369, Email: jstachel@bryant.edu

1.0 Introduction Each month, millions of cars are sold nationwide to eager new owners looking for an easier method to get to and from where they need to go. Unfortunately, what most people don’t realize is that the price of a vehicle only starts after it has been purchased. Today, gasoline prices are soaring to record highs and becoming one of the major concerns to consider when purchasing a vehicle. But what exactly is the cause for the price increases? In this regression, five different independent variables will be included to see what exactly is the cause of recent gasoline price increases. Gasoline Price will be used as the independent variable while Imported Oil Price, the number of barrels ordered per month, the commodity price of oil, the number of barrels produced domestically per month, and the number of auto sales per month will be dependent variables. This study hopes to examine these variables to see the effects and importance that each one has on the overall price of gasoline in the United States. With the current trends continuing and no end seeming to be in sight, the price of gasoline is a very real concern for a country that is highly dependent on foreign oil. This topic is incredibly important to our economy, because if we can discover the main cause behind gasoline price, efforts can be made so that we can stabilize, or even decrease the price of gasoline so money can be spent elsewhere to further boost the economy. This paper differs from others that have been observed, in that this contains the inclusion of the number of automobiles purchased monthly in the OLS regression with the variables pertaining to oil. It seems common sense that the increase of automobile sales will increase gasoline prices as the demand for more fuel becomes present. However, is this truly the main cause for the increase in prices? McManus (2007) believes that the increase in gasoline

prices may not be correlated with the number of automobiles sold. He states that since September 11th automobile sales have decreased (on average compared to previous trends), while the price of gasoline and oil has steadily increased. This paper will answer not only, which of the five variables run in the OLS regression is the most significant factor, but it will also show which variables have positive or negative impacts upon gasoline price, and to what degree they affect gasoline price. 1.0 Trends The current trends that are emerging in the oil markets have been an increase in production amount and price almost every year. Most of these increases are due in large part to economic shocks such as war, embargos, and governmental regulations. In order to obtain the true determinants of price for gasoline, four years prior and after September 11, 2001 have been chosen so that the impact of the actual event can be spread over a greater period of time when the averages are computed. The following graph contains data showing the Price of Crude Oil to the United States from the years 1947-2007.

As evidenced by the graph, the most important factor that has been determining the price of Crude Oil has been shocks to the economy caused by war or forms of embargo. The biggest spike was caused by the Iran/Iraq war which placed large strains on all OPEC countries to produce and export oil. Likewise, after September 11th oil prices once again soared and continue to increase today. The next chart shows the amount of crude oil production by OPEC countries, who together export the largest amount of crude oil.

The price of crude oil has steadily increased over the past 34 years regardless of the fact that production levels are constantly changing. This price has obviously been transferred to the consumers who have seen gas prices steadily and incrementally rise every month for years. 1.0 Literature Review Each month it seems that the price of gasoline continues to increase no matter what we seem to do (buy hybrids, protest gas sales etc). Lee and Zyren (2007) say that the traditional price theory in economics, which determines an equilibrium price by balancing supply and demand, may not fully explain the current price behavior of crude oil and petroleum products. This is important because if the traditional theory of supply and demand does not hold for crude oil prices, what exactly determines the equilibrium price? Abosedra and Radchenko (2006) find that gasoline prices tend to increase quicker when crude price increases, than they decrease when crude price decreases. If this is 24

truly the case, then can we blame outside oil companies and countries who sell the oil we so desperately need? As a corollary to this, Lewis (2004) found that there is little or no empirical evidence identifying the market characteristics responsible for changes in gas price. Borenstein, Cameron, and Gilbert (1997) find asymmetric passthrough from crude to retail oil prices in the U.S. and they attribute the asymmetry partly to retailer market power and tacit collusion. It is their opinion that the price of crude oil has more to do with the collusion between oil tycoons rather than the actual marketplace. Borenstein, Cameron, and Gilbert (1997) also hold that wholesale gasoline prices respond about as equally quickly to decreases as to increases in spot prices (the commodity price) for generic gasoline. McManus (2007) holds that the price of gasoline may not impacted by the sale of automobiles. Due to this, the variable of Automobiles sold per month was added to the regression to test this hypothesis. Based upon the opinions of these authors and some of the conflictions in their opinions, the variables of imported oil price, the number of barrels imported, commodity price, domestic oil produced, and the number of automobiles sold were run in an OLS regression, in order to determine which of these variables does in fact have the greatest effects on gasoline price. 2.0 Data and Empirical Methodology 2.1 Definition of Variables 𝑃𝑅𝐼𝐶𝐸 = 𝛽0 + 𝛽1 𝐼𝑀𝑃𝑂𝐼𝐿𝑃𝑅𝐼𝑖 + 𝛽2 𝑁𝑈𝑀𝐵𝐴𝑅𝐼𝑀𝑃𝑖 + 𝛽3 𝐶𝑂𝑀𝑃𝑅𝐼𝑖 + 𝛽4 𝐷𝑂𝑀𝑂𝐼𝐿𝑃𝑅𝐼𝑖 + 𝛽5 𝐴𝑈𝑇𝑂𝑆𝐴𝐿𝐸𝑖 The dependent variable, PRICE, represents the price for a single gallon of gasoline based on the United States national average based on monthly data in cents. The independent variables which affect the price include the following: IMPOILPRI is the price per barrel of

imported oil. The independent variables, NUMBARIMP and DOMOILPRI, signify the number of barrels ordered per month for imported and the latter correspond to domestic orders. The next variable, COMPRI stands for the Commodity Price of Oil in the Stock Market. To statistically symbolize the number of auto sales the next variable is labeled AUTOSALE. To maintain accurate data for the ordinary least square (OLS) regression model, the data for the five independent variables will be collected by monthly, national averages. All of the variables are expected to have a positive sign, as it is believed that each variable does have a positive impact on the final price of gasoline. 2.2 Data This study uses data that was collected on month by month nationwide average for the years 1996-2006. The data for the variables of Average monthly Gasoline Price (PRICE), Domestic Oil (DOMOILPRI), the Commodity Spot Price of Oil (COMPRI), and the Imported Crude Oil (NUMBARIMP) all come from information listed on the Energy Information Administration (EIA) website. The information containing the number of automobiles sold in the years 1996-2006 monthly was taken from the Wards Automotive Group’s yearly journals depicting all information regarding auto sales around the world. Predictions: It is the belief of this author that the most important factors which determine gasoline price are the imported oil price, and the number of barrels imported. Since the United States relies almost entirely on imported oil, the belief that the price of this imported oil would be the largest determinant seems to be well founded. The statistics created from this information are contained in the tables below:

(Table 1) Regression Analysis Data

Std.

Expected

Coefficient

Error

t-Statistic

Prob.

Sign

COMPRI

0.90404

0.964087

1.006552

0.3173

DOMOILPRI

0.000246

0.000119

2.06194*

0.0425

IMPOILPRI

1.873107

0.810884 2.309958* 0.0235

NUMBARIMP

0.000102

5.910005 1.719535* 0.0895

AUTOSALE

1.0005

6.540006

1.535018

0.1288

2.942649

25.8583

0.113799

0.9097

Table 1 shows us the Coefficients, Std. Errors, t-Statistics, Probability, and Expected Signs that were found for each variable tested in the OLS regression. The asterisk’s next to each t-Statistic denote variables that were shown to be significant in obtaining the price associated with a single gallon of gasoline. As expected, each variable did in fact have a positive sign associated with it. (Table 2) Regression Analysis Advanced Output

R-Squared Adj R-Squared S.E. of Regression Sum Squared resid

0.83008 0.819188

Mean dependent var S.D. dependent var

158.8478 19.08307

8.114508

Akaike info criterion

7.093933

5135.928

7.267563

Log Likelihood

-291.945

F-Statistic

76.2079

Schwartz criterion Hannan-Quinn criterion Durbin-Watson Statistic

7.163731 1.01127

Table 2 contains the information retrieved during the regression such as the R-Squared value, the F-Statistic, and the Durbin-Watson Statistic. For the complete Output listing from E-Views, please refer to Appendix A. 3.0 Empirical Methodology As we can see from the data presented in table 1, the greatest determinant to gasoline price in the United States is the price per barrel of oil that is imported. Since we are in fact a country that depends almost entirely on oil produced in other places in the world, other oil producing countries seem to have the biggest amount of control when it comes to determining our gas prices at the pump. Due to the evidence procured in the Trends section, the supply shocks of war and governmental regulations also play a large role in determining the final price of crude oil. Limiting these events, will most likely cause the equilibrium price of oil to stabilize at a moderate rate. Also, since the data proves that the amount of domestically produced oil is only a fraction compared to what is imported, this is not an unbelievable result for the imported oil price to contribute more to the price of gasoline than the domestic price. The second most important variable to gasoline price is the price of domestic oil produced. This result was different than what was perceived earlier in the regression, as the amount of oil produced domestically was not believed to be large enough to have a large effect on the price of gasoline. If we can find a way to increase the use of domestic sources of oil, the price of gasoline will surely decrease in response to the decreased dependence on foreign sources. By opening up United States oil reserves the price of gasoline would greatly decrease as the abundance of US reserves is more than enough to create gasoline for at least the next decade. The third most significant value is the number of barrels of oil that were imported. This directly correlates with the price of imported oil, as the 28

increase in price will cause the amount imported to fluctuate. The other variables such as the number of automobiles sold was not truly significant and this proves what was said by McManus (2007). This regression found an opposite reaction to what was held true by Borenstein, Cameron, and Gilbert (1997). In this regression the spot price does not truly have a significant effect on the price we see at the pump. 4.0 Conclusion In this paper, data was collected from 1996-2006 in order to determine which of the five independent variables have the greatest impact on gasoline price. Through these results we can see that the price of imported oil is truly the greatest determinant of gasoline price. Due to the fact that the United States as a nation relies so heavily on imported oil, fluctuations to the price per barrel that we import will have the greatest effect on changing prices at the pump. In order to bring down or stabilize the price of gasoline, the United States must concentrate on either reducing our dependence on foreign oil, or finding a way in which we can lessen the cost to import crude oil. The sale of automobiles were found to be relatively insignificant, therefore current trends may continue without much of an impact, other than the personal cost of how often one must actually fill up their automobile. Similarly, the commodity spot price of crude oil does not have a significant impact on the final price of gasoline at the pump, which means that constant fluctuations in the stock market can be relatively harmless in the short run. Domestic Oil Prices were also found to be significant, which shows that if we can find a way in which to decrease our dependence on foreign oil and increase our use of domestic oil, then the price will surely fluctuate.

References Abosedra, Salah, and Radchenko, Stanislav (2006), “New evidence on asymmetry in gasoline price: Volatility versus Margin” Organization of the Petroleum Exporting Countries (2006). Borenstein, S., Cameron A.C. and R. Gilbert (1997), “Do gasoline prices respond asymmetrically to crude oil price changes?” The Quarterly Journal of Economics 112, pp. 305–39. Lee, Thomas K., Zyren, John (2007), “Volitility Relationship between Crude Oil and Petroleum Products” Atlantic Economics Journal 35:97-112 Lewis, Matt (2004), “Asymmetric price adjustment and consumer search: An examination of the retail gasoline market.” Working Paper Number CPC04-47Rev. Completion Policy Center University of California, Berkeley. McManus, Walter. "The Link between Gasoline Prices and Vehicle Sales." Business Economics 42.1 (Jan. 2007): 53-60. EconLit. EBSCO. Bryant University Library, Smithfield, RI. 20 Apr. 2008 <http://search.ebscohost.com/login.aspx?direct=true&db=ecn&AN=0913038&site=ehost -live>.

Appendix

Variable

Description

Data Source

Gender

US Census Bureau

Age

The Age of the immigrant at the time of arrival

US Census Bureau

Education

The education degree the immigrant has: Educ: 0: HS; 1: College and Post grad: Educ1: 0: College and Post and 1: HS. The years the immigrant has spent in the US

US Census Bureau

Ethnicity

Years in the U.S

US Census Bureau

Causality Relationship between Foreign Exchange Rates and Stock Market Close: Evidence from Singapore

Jeffrey Allen Shew, Jr.b

Abstract: This paper attempts to examine whether or not a causal relationship exists between exchange rates and stock market returns. By using the Granger Causality model, causality relationships were determined for four data sets created through the use of significant structural breaks between 1990 and 2006 in Singapore. The results suggest that over the course of 16 years there exists no relationship between exchange rates and stock market returns. However, from 1990 up until the Asian Financial Crisis exchange rates led stock prices as suggested by Granger et al. (2000). The other two time series, between the Asian Financial Crisis and September 11th, 2001 and September 11th to 2006, show the exact opposite results suggesting that stock prices granger cause exchange rates.

JEL Classification: G15, c32 Keywords: Exchange Rates, Stock Market Returns, Granger Causality, Singapore b

Department of Economics, Bryant University,1150 Douglas Pike, Smithfield, RI02917. Phone: 860.305.1180 email:jshew@bryant.edu

1.0 Introduction The year 1973 marks a significant turning point in the world’s global financial market; the creation of floating exchange rates. Floating exchange rates have allowed for investors, businessman, and stock market speculators to determine the currency’s value. Typically economically stable countries and unions have allowed for their currency to float including the United States, Japan, and the European Union. Although some countries have kept quasifloating or pegged exchanges rates, many countries have left the faith of the currency in the hand of the financial markets. In September, 1992 Gregory Soros (investopedia.com), a currency speculator, bet $10 billion on a speculation that the British pound would decline. In nearly a day time he generated a $1 billion dollar profit. Some argue that the overwhelming size of his trade instantaneously changed the mindset of investors indirectly forcing the value of the British pound and the stock market lower. It has also been argued that the effects of countries suffering from economic downturn are amplified by their floating exchange rates. However, the question remains what exactly is the relationship between exchange rates and stock market? There are two explanations for which variable cause the other. The floworiented model approach as described in Dornbusch and Fischer (1980) research show that currency movements directly affect international competitiveness. In turn, currency has an effect on the balance of trade within the country. As a result, it affects the future cash flows or the stock prices of firms. The counter argument suggests that taking a portfolio-balance approach (Dornbusch, 1976), where portfolio holders should diversify to eliminate firm specific risk, requires effective investments allocation including currencies. As with other financial instruments, currencies therefore are under the rules of supply and demand for assets. In order for investors to purchase

new assets they must sell off other less attractive asset in their portfolio. In other words buying and selling of domestic or foreign investments if less attractive. As countries assets become more valuable, interest rates begin to increase creating an appreciation of domestic currency. Although two valid explanations, no consensus has been made between the two. This study looks to validate one of the explanations over the other by looking at Singapore’s exchange rate to its stock market prices between 1990 and 2006. 2.0 Trends The Singapore dollar (SGD) over the past five years has continued to appreciate over the United States Dollar (USD) as displayed in Chart (1). Although this can be the result of prior and current economic downturns in the United States, the Singapore still appreciated over the USD during the period between the internet bulb burst and the current credit crisis. The Singapore Exchange has shown a general upwards trend for the previous five years shown in the Chart (2). However, during the 1990’s the market fluctuated from decade highs to decade lows as well as a plateau between 1993 and 1997. Looking at the both five year graphs, exchange rates have been strengthening as well as the Singapore markets. Chart (1) SGD/USD Exchange Rate 2003-2008 (Chart provided by: Yahoo.com/finance)

Chart (2) Straits Times Index 2001-2008 (Chart provided by: Yahoo.com/Finance)

Chart (3) Straits Times Index 1988-2008 (Chart provided by: Yahoo.com/Finance)

3.0 Literature Review Economic theory states that changes in exchange rates will directly affect a firm’s net income through its foreign operation. This change in profits will impact the underlying value of the firm’s stock. There has been much research in the past 30 years focusing on exchange rates and their relationship to stock prices. In the 1980’s, researchers tested specifically the correlation between both exchange rates and stock prices. Aggarwal (1981) research suggested a

positive correlation between both variables where as Soenen and Hennigar (1988) tests appeared to show a stronger negative relationship. It was not until the early 1990’s that Granger Causality Testing to explain the relationship between equity markets and exchange rates. Bahmani-Oskooee and Sohrabian (1992), one of the first studies to use the GC testing, found bidirectional causality between the returns of the S&P 500 index and the effective exchange rate of the dollar. Since then, research on the topic continues to grow. Different variations have been attempted, including different samples and time series, giving various explanations to whether exchange rates Granger cause stock price or vice versa. Although results diverge, the traditional theory described above tends to explain the majority. In their 2003 study, Smyth and Nadha (2003) found a uni-directional causality from exchange rates to stock price in both India and Sri Lanke, however Bangladesh and Pakistan show that both variables independent from one another. In respect to Brazilian markets, Benjamin Tabak’s work, The Dynamic Relationship between Stock Prices and Exchange Rates: Evidence for Brazil, discovered Granger causality from exchange rates to stock prices (Tabak 2006). Among other countries in Granger et al. (2000) study, Singapore’s data suggested a relationship flowing exchange rates to stock prices prior to the Asian Financial Crisis. Amongst those whom studied the effects of short term or long term relationships between the two variables, Muhammad and Rasheed (2002) discovered in Pakistan and India there was no long run relationship between stock prices and exchange rates. 4.1 Methodologies and Results To measure causality between the Singapore stock market closes and exchange rates, the Granger Causality Model was used. This test is used to determine whether or not one variable is the cause of the other, vice versa, or neither. After eliminating serial correlation, Granger’s

(1969) GC test provides four possible outcomes to a regression of variables x and y: No causality, X granger causes Y only, Y granger causes X only, and bi-directional causality. The following summarizes the step necessary to complete the GC test: i.

Test for Unit Root between stock markets and exchange rates using the Augmented Dickey-Fuller Test(ADF)

ii.

If a Unit Root is present, difference the data and execute a ADF test again

iii.

Estimate co-integration using the same order of integrated variables

iv.

Use a VAR to test Causality

The Granger Causality Test estimates the results of two regressions: 𝑛

𝑛

𝑀𝐾𝑇 = ∑ 𝛼𝑖 𝑋𝑅𝐴𝑇𝐸𝑡−1 + ∑ 𝐵𝑗 𝑀𝐾𝑇𝑡−1 + ε1𝑡 𝑖=1

𝑚

𝑋𝑅𝐴𝑇𝐸 = ∑ 𝜆𝑖 𝑀𝐾𝑇𝑡−1 + ∑ 𝛿𝑗 𝑋𝑅𝐴𝑇𝐸𝑡−1 + ε2𝑡 𝑖=1

(1)

𝑗=1

(2)

𝑗=1

Where: MKT= the Singapore market close for the day XRATE= the Singapore Dollar to USD exchange rate for the day Equation (1) regress the dependant variable, MKT, to how it relates to lagged variables of XRATE and MKT. Equation (2) is similar to Equation (1), it regress XRATE to lagged variables of XRATE and MKT. Both equations assume that the disturbance terms, ε, are uncorrelated. The null hypothesis for equation (1) and (2) state no evidence of causation from MKT to XRATE and no causation form XRATE to MKT respectfully. Based on these results one of the four possible outcomes, as described above, will become evident. The goal of this study was to expand on the Granger et al. (2000) study which observed the time up to the Asian Financial Crisis and shortly after. This study focused primarily on the 38

period after the crisis to the present time period, which to the author’s knowledge has not be studied in depth. Base on the Granger Causality Model procedure, four different time series were tested. One tested the Granger causality between stock market closes and exchange rates for the time period from 1990-2006. The other three use the same test however different data sets. Two structural breaks were used on the overall data set at the July 1st, 1997(time of the Asian Financial Crisis) and September 11th, 2001(Terrorist Attack on the USA). There are a few explanations for this. The structural break at the financial crisis allows for comparison to previous studies using GC test on exchange rates and stock market closes insuring valid results and data selection. It also allows us to examine the larger test (1990-2006) in different sections to see if the is some causation between exchange rates and the stock market in shorter spans of time appose to fifteen years. 4.2 Data The data selected for this research was acquired from internet historical data logs. To measure the changes to the stock market, the day’s closes were used based on the Singapore Index. Exchange rates were based on the daily close of the Singapore dollar to the US dollar exchange rate. Yahoo.com was used to collect Singapore Index historical closes and oanda.com, a provider of foreign exchange data and services, to collect the daily foreign exchange rates. The accumulated data spans over 16 years from January 01, 1990 to December 31, 2006 for a total of 4,167 observations. Although stock markets close during weekends and specific holidays, exchange rates are calculated on a daily basis. To account for this discrepancy, all exchange rate closes during the weekend and these holidays were excluded to allow for one exchange rate and one stock market close on any specific trading day.

5.0 Empirical Analysis The hypothesis in the introduction suggested that there are two possible explanations for the directions of causality between stock market returns and exchange rates. The results suggested by the Augmented Dickey Fuller required that the unit root data be differenced. By differencing the data, it allowed for the time series to become stationary and eliminated misspecification in the causality test. Table 1 and 2 show the results from both the ADF and cointegration tests. The empirical results from the study show over the entire period from 1990 to 2006, there is no relationship between stock prices and exchanges rates in Singapore. This conclusion supports the results found by Muhammad and Rasheed (2002) where they found that there is no long-run relationship between both variables in Pakistan and India. Granger et al. (2000) results suggested that up until the Asian Financial Crisis, Singapore’s exchange rates lead the price of stocks. To insure the correct methodology, the period between 1990 and up until the financial crisis was tested. The results from this study suggest that exchange rates indeed lead the price of stocks. These results were consistent with previous studies as suggested above. The empirical results for both the time periods between the structural break (the Financial Crisis in 1997 to September 11, 2001) and from September 11th until 2006 suggest, with statistical significance, that stock prices lead exchange rates in Singapore. The results from this time series explain the portfolio balance approach, an alternative explanation for the relationship between both variables. These results contradict a majority of previous studies conducted on this topic. A possible reason for the discrepancy could be the economic conditions follow both time series. After 1997, financial markets were experiencing the side-effects from the Financial Crisis and a

combination of the internet bubble bursting and the terrorist attacks on the United States. These events could have forced skewed results for the time series after 1997. 6.0 Conclusion This study looked at the causal relationship between stock market prices and exchange rates in Singapore between 1990 and 2006. By using two structural breaks at the Asian Financial Crisis and September 11th 2001, four different time series were tested for Granger Causality. After making the data stationary through differencing the data and testing for co-integration, the results from the four time series complemented prior research. In the long run, neither variable Granger caused the other. The data before the first structural break, the Asian Financial Crisis, showed that exchange rates lead the prices in equity markets. However this was not the case for both the time periods between the structural break (the Financial Crisis in 1997 to September 11, 2001) and from September 11th until 2006 where stock prices Granger cause exchange rates. The extreme conditions that followed 1997, including the Asian Financial Crisis and the terrorist attacks on the US, may have caused skewed results for both time series. In summary, based on results, the study favors neither explanation for the relationship between both variables. Depending on the time series tested in Singapore, the relationship between stock prices and exchange rates differs.

Table 1: Results of the ADF Test 1/1/9012/31/06 ADF UR(Level)

Fisher Chisquared Choi Z-stat

2.04535

2.58535

0.4612

Fisher Chisquared

36.8414 *** Choi Z-stat 5.25948 *** Denotes Significant at 5% Critical Value

UR(1st Difference)

1/1/067/1/97 ADF

7/1/979/11/01 ADF

9/11/0112/31/06 ADF 0.09612

0.13692

11.7223 *** 1.73176 **

173.798 *** 12.6694 ***

371.522 *** 18.9056 ***

190.557 *** 13.3309 ***

3.08412

*** Denotes Significant at 1% Critical Value

Table 2: Co-integration Results Time Period

λ Trance

λ Max

1/1/90- 12/31/06 1/1/90- 7/1/97 7/1/97-9/11/01 9/11/01-12/31/06

3.158116

2.237945

9.381953

6.993401

15.91005**

14.22739**

10.12588

8.251146

** denotes significant at 5% Critical Value

Table 3: Granger Causality Test Results Stock Market to Exchange Rate

Exchange Rate to Stock Market

January 1, 1990- December 31,2006

January 1, 1990- July 1, 1997

Yes

July 1, 1997- September 11, 2001

Yes

September 11,2001-December 31,2006

Yes

Time Period

Aggarwal, R. (1981), “Exchange Rates and Stock Prices: A Study of the US Capital Markets under FloatingExchange Rates”, Akron Business and Economic Review, 12: 7–12. Bahmani-Oskooee, M. and Sohrabian, A. (1992), “Stock Prices and the Effective Exchange Rate of the Dollar”, Applied Economics, 24: 459–464. Dornbusch, R. (1976) “Expectation and Exchange Rate Dynamic”, Journal of Political Economy, reprinted in Dornbusch,R. (1988 ed.) “Open Economy Macroeconomics” NY: Basic Books Publisher. Dornbusch, R. and Fischer S.,(1992), “Exchange Rates and the Current Account”. Exchange rate economics. 1: 296-307. International Library of Critical Writings in Economics,16 Granger, C., (1969), “Investigating Causal Relations by Economic Models and Cross-Sppectral Methods”, Econometrica , 37: 251-71 Granger, C. W. J., Huang, B. N. and Yang, C. W., (2000), “A bivariate causality between stock prices and exchange rates: evidence from recent Asian flu”, Quarterly Review of Economics and Finance, 40: 337–54. www.investopedia.com/university/greatest/georgesoros.asp Muhammad, N. and Rasheed A., (2002), ”Stock Prices ad Exchange Rates: Are they Related? Evidence from South Asian Countries”, Karachi University Nadha, M. and Smyth, R.,(2003), “Bivariate causality between exchange rates and stock prices in South Asia”, Applied Economic Letters, 10: 699-704 Soenen, L.A. and Hennigar, E.S., (1988), “An Analysis of Exchange Rates and Stock Prices: The US Experience Between 1980 and 1986”, Akron Business and Economic Review,19(4): 71–76. Tabak, B., (2006), “The Dynamic Relationship between Stock Prices and Exchange Rates: Evidence for Brazil”, International Journal of Theoretical and Applied Finance, 9: 1377-1396

Factors Affecting Health Care Spending Jennifer Resende a ABSTRACT With an aging population and an increase in health care spending across many nations, there is a need to determine what is affecting this increase and whether this trend can be expected to continue into the future. This paper aims to investigate the possible differences in health care expenditures in different countries. The study incorporates the use and analysis of several independent variables that are believed to affect health care spending in an array of countries including life expectancies, the increase in aging populations, health care spending on private sectors, and the quality of health care as represented by the number of hospital beds available to patients. This paper seeks to determine if there is in fact a steady correlation among these independent variables among all thirty pre-selected countries, or whether health care expenditure varies by country and is affected by other determinants. The results of this study reveal that, as predicted, female life expectancy, spending on private health care sectors, and the increase of population over the age of 65 does in fact positively affect the health care expenditures among countries. JEL Classification: I10, I11, I12, I18 Keywords: Health Expenditures, Determinants

Bachelor of Science in Business Administration: Marketing & Economics, Bryant University, 1150 Douglas Pike, Box 1516. Smithfield, RI 02917. Phone: (401) 333-1314. Email: jresende@bryant.edu a

_____________________________ The author gratefully acknowledges the help and guidance from Dr. Ramesh Mohan and also thanks reviewers Sahana Zutshi and Kenneth Smith for their assistance.

1.0 INTRODUCTION The topic of comparing public health expenditures among a variety of different countries is important to analyze because there is a general trend of increasing health care costs across the globe. As the general population in many countries continues to age and baby boomers begin to retire, health care costs are expected to rise drastically across multiple countries. This is a concern to many, particularly in lower income countries with a lower GDP where the quality of health care is much lower than in other developing or fully developed countries. Over the next fifty years, the aging population will continue to increase. While not all countries are affected by an overwhelmingly large population of “baby boomers” and adults reaching the age of 65, almost all countries are faced with rising health care costs. Generally speaking, the older population requires more health care and as the elderly population increases, health care costs will most likely be expected to increase with it. A second reason for the increasing health care costs across countries could be explained by the fact that many countries, particularly developing countries, are attempting to provide better quality health care to their citizens. Naturally, it is expected that as the quality of health care provided to citizens increases, so will the cost. This paper aims to investigate and analyze the determinants of health care expenditures among thirty different countries. A regression analysis will be run and results will be compared for a variety of different variables between the years of 2000 and 2005. A selection of about thirty countries ranging from low income countries to high income countries have been chosen, with many of these countries belonging to the Organization for Economic Cooperation and Development (OECD). One of the main empirical questions that this paper aims to answer in researching this topic on health care expenditures includes determining which variables affect health care expenditures the most and why. This paper will address these topics of interest and will be organized in the following order: Section 2.0 addresses trends in the health care market; Section 3.0 provides a literature review of previous research; Section 4.0 explains in detail the empirical model, data, and methodology and is broken into subsections; Section 5.0 provides empirical results and an analysis of the research findings; Section 6.0 will end the paper with a conclusion and reasoning. Finally, additional charts and tables can be found in Appendix A and B, followed by a listing of all the tables used within this paper.

2.0 TRENDS IN THE CURRENT HEALTH CARE MARKET Identifying some of the major factors of rising health care expenditures is a pressing topic in recent years. With the general population aging over the next forty to fifty years, health care is on the rise in countries across the globe.

Figure 1: Current Patient Trends

Figure 1 illustrates trends in patients between the years of 1992 and 2000. The graph shows how there was a large peak in the year 1996 of patients who received health care at home, and that the majority of this peak was a result of a spike in patients over the age of 65. However, after 1996, the number of home health care patients has decreased, which could indicate that more patients are receiving health care in public facilities such as hospitals and private practices and that less are staying at home. This is important in analyzing factors in rising health care costs because if more and more citizens are moving to hospitals for treatment and are staying in these facilities as opposed to returning home, the costs of health care will clearly increase as well. Figure 2: Government Health Care Spending as a Percentage of GDP

Across the globe is a trend of higher spending in terms of a country’s gross domestic product (GDP). Figure 2 shows a comparison of different countries and their spending on health care. We can see how the United States is one of the top countries in spending on health care, with other developed countries following. Moreover, we can see how developing countries are beginning to focus on health care spending. Figure 3: Government Health Care Spending as a Percentage of GDP in 2050

Many countries, even lower income and developing countries are experiencing higher levels of spending on health care due to a demand for better care, more services, and a movement in many countries towards a public health care plan. As a result, many countries are making a move towards a public policy in terms of health care, which enables all tax-paying citizens to receive medical attention and health care as sponsored and provided by the government. This is a new, fast growing trend among many countries, and it is arguable that it provides better, higher health care than private health care providers with lower spending. Figure 3 illustrates how the United States, an extremely developed country, compares to other high-income countries in terms of health care spending and how it is expected to continue in the future. The graph shows health care spending in terms of a country’s GDP in the year 2050.

3.0 LITERATURE REVIEW Previous studies have investigated possible correlations between variables and health care spending among individual countries. Each of these studies has developed different results about variables and health care expenditures. According to Huang (2004), some of the “key determinants are growth in gross domestic product and percentage of government health expenditure to gross domestic product. Other factors, like the ageing population and the number of doctors per thousand populations do not significantly affect the growth of health care expenditure.” This is an important conclusion because while Huang concludes that the aging population and quality of health care based on the number of doctors available does not affect spending on health care for high income countries, this paper will aim to either support or contradict this claim. Another study on this topic performed by Herwartz (2000) makes a similar claim by determining through his research that high GDP growth in countries is directly correlated with health expenditures. This study focused primarily on OECD countries. However, this study by Herwartz “applied methods of statistical inference, which took into account cross-sectional error correlations and heteroskedasticity and which perform particularly well in small samples.” In other words, there were doubts about whether the results would be statistically significant and valid, and the author took precautionary steps to ensure that running tests with a small sample did not affect the validity of his results. One particular working paper (by Dreger and Reimers (2005)) showed results that claims that there are other determinants involved in health care spending. In Dreger and Reimers (2005), 47

a study on health care expenditures resulted in discovering that along with income and GDP growth, “the other driving force is medical progress, which can be observed in the evolution of several variables, like life expectancy, infant mortality and the share of the elderly.” This directly contradicts Huang (2004) who claims that only GDP and health expenditures affect spending. Interestingly, in a study on health care determinants in Portugal, Salvado (2007) claims that there are not only many determinants that affect spending and health care quality, but also that public health care combined with private care can be both beneficial and harmful. “Both private insurance and subsystems ensure additional cover and are widely used by approximately 25% of the Portuguese population. They may constitute an object of study about excess of consumption associated with moral hazard problems.” This is an interesting topic that is discussed further in this working paper by Salvado (2007), and a claim is made that even though it is possible that having both public and private care can be harmful and may cause problems, it can also be beneficial because sometimes, public care is not enough to handle the high demand for medical attention. In a study performed by Or (2000), one of the limitations of his study on health care determinants of OECD countries included analyzing differences between males and females and isolating their effects on health care expenditures. Because this was seen as a limitation on a previous study and these factors are believed to have significant effects on health care spending, these variables were included in this study and their effects on health care will be discussed later on in this paper.

4.0 DATA AND EMPIRICAL METHODOLOGY 4.1 Definition of Variables This model is different from regression models and studies that have been run in the past. While other regression models have focused on only two or three independent variables that are thought to affect health care spending, this model that will be studied and analyzed in this paper takes into account different variables that were used in a variety of papers. Of course, in the real world there are many factors and variables that affect the health care costs in different countries. By analyzing all of these variables in one model, the results will capture the effects of each independent variable on health care costs, taking into account other variables. The regression model will include the dependent variable, which is public health expenditures, and the following independent variables, including age-dependency ratio, population of males and females over the age of 65, percentage of a country’s GDP that is spent on private health expenditures, life expectancies of males and females, the number of hospital beds per 1000 people, and a country’s individual GDP. The basis of these independent variables lies on some preliminary research that has been previously conducted by examining other working papers of similar topics. By indicating that the dependent variable is the total health expenditure of a country allows for an investigation of any relationships between the dependent variable and the independent variables. The dependent variable, which is total health care expenditure (HE) for a country, was given as a total percentage of an individual country’s GDP for a given year. The first independent variable, which is the age-dependency ratio, AD, allows for an investigation of any relationship between the ratio of dependents to the working-age population. In other words, using this variable aims to examine a possible relationship between every worker in the labor force and the number of dependents that they have claimed and whether this affects health care costs. By examining the population of citizens who are 65 and above, POP, which is 48

the second independent variable, the researcher is able to determine if health care costs rise in various countries as citizens’ age and as the average population age increases. The third and fourth independent variables, which are life expectancy for males, LEM, and females, LEF, allow the regression to capture any relationship between gender and the health care costs. Life expectancy provides a general idea of how long males and females are expected to live. This number does range from country to country and may be an indicator of the level of quality of health care in a given country. One would expect that better health care would yield a longer life expectancy, which would in turn result in higher health care costs for a given country. The fifth independent variable, which is private health expenditures, HEP, investigates any relationship between a country’s total health expenditures and a country’s spending on private expenditures. This variable is a percentage of a country’s total GDP. The sixth variable, which is the number of hospital beds available, BED, investigates the relationship between health care expenditures and the quality of health care provided in a given country. This variable is investigated in terms of the number of beds available per one-thousand people in the country’s population. The seventh and final independent variable is each country’s current GDP, GDP, which is a country’s gross domestic product for a given year. This is the total market value of all final goods and services produced within a given country in a given period of time (usually a calendar year) 4.2 Data The study uses annual data from 2000 to 2005. Data were obtained from the World Development Indicators (WDI) website. By conducting a cross-sectional analysis, data was collected for the six year time period from 2000-2005 for thirty countries. One of the advantages to running a cross-sectional analysis is that for the few circumstances in which data was unavailable for a particular independent variable within a certain year, an average of the following year’s figure and the previous year’s figure calculated a figure for the missing data. Summary statistics for this data can be found in Table 1 below.

Variable

Obs.

Table 1: Summary Statistics Mean Std. Dev. Min

6.701600

1.942326

2.260000

9.880000

0.563080

0.114766

0.423000

0.919000

POP

9.347760

5.246896

3.613000

19.12800

LEF

76.38444

5.374724

64.02700

85.20800

LEM

70.45656

5.379637

58.79200

78.26900

HEP

2.709000

1.050476

1.333000

5.222000

BED

3.744960

3.467822

0.500000

14.500000

GDP

4.81E+11

9.06E+11

6.21E+9

4.33E+12

Max

4.3 Research Methodology In a cross-section analysis, a comparison of many countries is performed over a specified time period while a time-series or longitudinal analysis is performed for one specified country over 49

a length of time. For the intent and purpose of this paper, a cross-sectional regression analysis will be used because of the fact that more data was available for a specified time period of 2000-2005 for about 30 countries than was data for one specific country for a time period of about 30 years. A multivariate regression will be run, where the multivariate regression coefficient indicates the change in the dependent variable associated with a one-unit increase in the independent variable in question, holding constant the other independent variables in the equation. The primary goal of running an Ordinary Least Squares is to choose the Beta-hat as to minimize the summed square residuals. Ordinary Least Squares is a very easy regression model to use and is effective. Many times, there is a tendency among economists to accept regression estimates as they come directly from a computer, or as they are published in an article, without thinking about the meaning or validity of those estimates.

5.0 EMPIRICAL RESULTS The primary purpose of this study was to determine whether specific variables had a significant effect on health care expenditures consistently over a variety of different countries. In order to accurately regress the determinants of health care expenditures for thirty countries from 2000-2005, four separate regressions were run, one that captured the results of female life expectancy on health care expenditures, a second that captured the results of male life expectancy on health care expenditures, a third that captured the results of the age-dependency ratio on health care expenditures, and a fourth and final regression that identifies the effects of the aging population about 65 years old on health care expenditures. These four separate regressions were performed in order to prevent any multi co-linearity between the variables of life expectancy of males and females and the age-dependency ratio and the population over the age of 65. Because the these variables are not only related, but similar and also interchangeable, regressions were run, one with each variable, to avoid any possible bias or skewed results and to avoid any multi colinearity that might have altered the regression results and to try and accurately capture the effects of each variable. The results of these regressions are displayed in Table 2 below. Table 2: Regression Results for Factors Influencing Health Care Expenditures HEALTH CARE EXPENDITURES

CONSTANT AD POP LEF

III

-14.21789 ** (5.919648) 1.223227 (3.109296) 0.190826 * (0.095646) 0.218544*** (0.069054)

-11.19949 * (6.544245) 1.545232 (3.491426) 0.217555 ** (0.104658)

-18.44117 *** -12.58643 *** (6.295904) (4.832347) -0.179873 (3.579780) 0.198679 ** (0.094257) 0.412708 ** 0.345056 ** (0.184818) (0.165109) -0.122131 -0.144068 (0.189279) (0.163514) 0.827267 *** 0.731420 *** (0.275075) (0.257415)

LEM HEP

0.768801 *** (0.257296)

0.181254 ** (0.076791) 0.792305 *** (0.283483)

-0.076231 (0.154406) -9.07E-14 (4.37E-13)

0.073382 (0.206813) -4.04E-13 (5.53E-13)

-0.004662 (0.237084) 9.15E-14 (6.54E-13)

-0.236961 (0.210539) 2.71E-13 (5.24E-13)

0.699641

0.643001

0.673014

0.709583

F-statistics

6.988044

5.403377

6.517747

7.329985

Number of obs.

BED GDP

Note: ***, **, and * denotes significance at the 1%, 5%, and 10% respectively. Standard errors in parentheses. The results of these regressions do in fact show that out of seven variables and a constant, that only four variables were statistically significant throughout all four regressions. These significant variables include the constant, the population over the age of 65, the life expectancy of females, and the health expenditures on private health sectors. The variable for the population of age 65 and above was not only statistically significant, but had an expected positive sign that corresponded to the results of the study. This indicates that whether all of the other variables and factors are held constant or not, population will have an effect on health care expenditures across all countries. This corresponds with many studies and information provided that claims that as the population of citizens age 65 and above increases, as will the health care expenditures for all countries. This can be explained simply by taking a look at a simple fact. As people age, they become more likely to encounter health problems and will likely require more medication, treatment, and health care in general. As the population continues to age, health care spending can be expected to rise in order to meet the demand of these people. The second variable that was statistically significant was that of private health care spending. The expected sign for this variable was positive and the results indicate that a country spending on private health care will also positively affect the spending on health care as a whole. This was not surprising, as one would expect overall health care spending to increase as one spends more on private health care. However, thinking about the current trends of increasing quality of health care and spending, it cannot be determined from this study that there is any relationship that proves that private health care spending affects the quality of health care received or not. A sound conclusion cannot be drawn from this information because there is no correlation. The third statistically significant variable is the life expectancy of females. Consistent with the study’s results, it was predicted that this variable would have a positive expected sign, which indicates that as the life expectancy of females increases over time, so will the overall spending on health care. This can be explained by the fact that women have a longer life expectancy than males, therefore resulting in more time for health problems to arise, and consequently, more demand for health care. This regression was run independent of the male life expectancy, again to prevent the possibility of multi co-linearity and to capture the full effect of women’s life spans on health care spending. It is interesting to see that when these two variables are run separately, the life expectancy of males is only significant when it is run separately while the life expectancy of females is consistently significant. This could indicate that the life expectancy of males may not affect health care expenditures for several reasons. First, even though current trends indicate that the male life expectancy is rising more rapidly than female life expectancy ages, females have an overall higher life span than males. 51

It is interesting to note that out of all of the variables that were studied and tested, that only three variables and the constant were significant. Therefore, conclusions can only be drawn on these significant conclusions. The results of this study could be skewed due to some limitations on data. Several studies have been performed in the past that analyze a variety of factors that influence health care expenditures. However, valid and reliable data was not available for all of these variables for the countries analyzed, which resulted in studying the above variables. This may or may not affect the significance of these variables.

6.0 CONCLUSION With an increasingly aging population and an increase in health care spending across many countries, there is a need to determine what is affecting this increase and whether this trend can be expected to continue into the future. The primary purpose of this study was to identify specific factors that contribute to this rapidly increasing spending on health care expenditures among both developing nations and developed nations alike. The results of this study indicate that there are several factors that will consistently affect the spending on health care in all countries, independent of other variables. As expected, as the population of people over the age of 65 continues to increase, spending on health care will also increase. The second variable that affects health care expenditures is the life expectancy of females, which indicates that because females have a higher life expectancy in general than males, that spending will continue to increase as long as this trend prevails. The third factor that affects health care spending is spending on private health care sectors. Naturally, as private spending will increase, overall spending on health care for nations will increase as well. This study indicates that there is a direct correlation between all three variables and health care expenditures, indicating that these three factors may be constant variables across all countries, which contradicts Huang’s (2004) claim in his analysis of Singapore’s factors affecting health expenditures. Of course, there may be other factors that were not analyzed in this study that could affect a country’s spending on health care. However, from the results of this study, there is no solid evidence to support such a claim. Appendix A: Variable Description and Data Source

Acronym

Description

Data source

Total Health Care Expenditure for a given country, taken in terms of a percentage (%) of country’s GDP

World Development Indicators Online

Age-Dependency Ratio, ratio of dependents to the working-age population per country.

World Development Indicators Online

POP

Country’s population over the age of 65, taken as a percentage (%) of the total population

World Development Indicators Online

LEF

Life Expectancy of Females, average life expectancy of all females for a given country, in years

World Development Indicators Online 52

Life Expectancy of Males, average life expectancy of all males for a given country, in years

World Development Indicators Online

Private Health Expenditures, government spending on private health care costs, in terms of percentage (%) of country’s total GDP)

World Development Indicators Online

BED

Number of Hospital Beds available per 1,000 people

World Development Indicators Online

GDP

Gross Domestic Product (GDP) per country, in current US dollars ($)

World Development Indicators Online

LEM

HEP

Appendix B: Variables and Expected Signs

Acronym

Variable Description

What it Captures

Expected Sign

How many dependents are claimed for each Age-Dependency Ratio person in the labor force

POP

Percentage of Population over the age population that is over of 65 the age of 65

LEF

Life Expectancy of Females

Average life span for females

LEM

Life Expectancy of Males

Average life span for males

HEP

Private Health Expenditures

How much is spent on private health care in terms of the country’s GDP

BED

GDP

Hospital Beds

Gross Domestic Product

Quality of health care, number of beds available The total market value of all goods and services produced in given country

BIBLIOGRAPHY Dreger, Christian. Reimers, Hans-Eggert. 2005. “Health Care Expenditures in OECD Countries: A Panel Unit Root and Cointegration Analysis.” Discussion Paper Series. Paper # 1469. Available at: < http://ssrn.com/abstract=651985 >. Herwartz, Helmut. 2000. “The Determinants of Health Care Expenditure: Testing Pooling Restrictions in Small Samples. Health Economics, Vol. 12 No. 2: 113-124. Hodges, Michael. 2005. “Grandfather Economic Health Care Report.” Financial Sense Editorials. http://www.financialsense.com/editorials/hodges/2005/1218.html Huang, Seng Lee. 2004. “Factors Influencing Healthcare Spending in Singapore: A Regression Model.” International Journal of the Computer, the Internet and Management. Vol.12 No.3: 51-62. Or, Zeynop. 2000. “Determinants of Health Outcomes in Industrialized Countries: A Pooled, CrossCountry, Time-Series Analysis.” OECD Economic Studies, No. 30. Salvado, Joao. 2007. “The Determinants of Health Care Utilization in Portugal: An Approach with Count Data Models.” Social Science Research Network. Available at: < http://ssrn.com/abstract=1096568 >. World Development Indicators Online [online data file] <http://go.worldbank.org/IW6ZUUHUZ0 >.

The Importance of Governance to a Stable and Developed Economy

Peter Litvinenko1 Abstract:

This paper investigates the assumption that proficient governance is essential in securing a stable economy. It studies and analyzes many relevant variables that one would find controlled by the government in its economy. My results show that these established nations all have differences in terms of debt and balance of trades but all seem to agree on the importance of maintaining a morally stable and efficient government structure that allows them to compete in the world economy.

JEL Classification: G18, G21 Keywords: Governance, Regulation, Developed Countries

Department of Economics, Bryant University, 1150 Douglas Pike, Smithfield, RI 02917.

Phone: (917) 254-7563. Email: plitvine@bryant.edu.

1.0 INTRODUCTION Governance is defined as “rules, processes and behavior that affect the way in which powers are exercised…. particularly regarding openness, participation, accountability, effectiveness and coherence (Durlauf and Blume, 2006). Many theoretical and empirical papers have explored the issue of what certain macroeconomic policies adopted by governments have the most profound implications for their country’s economic growth and stability. Governance expresses its theme as the study of good order and workable arrangements (Dixit, 2006). This includes the institutions and organizations that underpin economic transactions by collecting property rights, enforcing contracts, and organizing collective action to provide the infrastructure of rules, regulations, and information. All of those factors are needed to lend feasibility or workability to the interactions among different economic factors, individual and corporate (Dixit, 2006). While the rapid progress of many countries is often attributed to the freeing up of markets, governance remains an important factor for economic and social development. Over time the topics of focus have shifted. Traditionally, academic writings have argued that democracy promotes economic growth. Economists have sought within democracy the preservation of property rights which provides the necessary causal factor for promoting economic growth (Gupta et al. 1998). Those writings further argued that non-democratic regimes are skewed in favor of the ruling elites; which hinders the free flow of capital. However, since the majority of the developed nations were themselves democratic it would have been pointless to use that variable. This research did lead to the realization that democracy’s economic strength has been said to come from its emphasis on literacy, education, and communication. That was a big reason for why the author felt it important to measure the level of provided education in my regression. It was also noted that an exchange economy cannot be most efficient until business morality made contractual behavior sustainable and inexpensive. Effective governance has also been instrumental in driving potential entrepreneurs away from productive operations to the higher returns offered by shady deals (Lee, 2006). Figure 1 shows the recent association between corruption and economic freedom. However, the regulation of corruption needs to start from the very top; from the government itself. The author used a measure of government effectiveness under the assumption that it would show to be very relevant for developed countries.

(Figure 1)

The last assumption this paper made was that increased involvement in the global marketplace as well as increased military spending all correlates with a stable economy. As businesses move to engage in more multi-national commerce it should be important to look at the amount of global economic participation by countries. Figure 2 shows how greater sums of FDI inflows parallel an increase in GDP growth. In order to quantify the level of global participation, foreign direct investment was looked at as a variable. Military spending is also taken into account with the belief that military spending helps the economy by offering opportunities to gain education and certain skills for later employment (Blume and Darlauf, 2006).

(Figure 2)

This paper should contribute significantly to the general field of governance and economics by focusing on already developed nations specifically and by collecting data samples from a time period not previously observed. This study will look at the weight of a wide range of factors in a cross sectional, multi-national, analysis.

Some of these variables, such as foreign direct

investment, have never been tested in this setting. The data on some of the other variables has simply just not been updated in more than 11 years. The rest of the paper is organized as follows: Section 2 reviews issues in the literature pertinent to the debate on the role of governance in economic growth. Section 3 outlines the empirical model used and then provides a descriptive analysis of the data and estimation methodology. Finally, section 4 presents and discusses the empirical results. This is all wrapped up with a conclusion in section 6. 2.0 LITERATURE REVIEW There have been many arguments offered and researched on the relationship between governance and a stable economy. A search on EconLit for the relevant categories shows just how much the topic has exploded through the last three decades. There were just five matches of papers completed from 1970 to 1979. The number then jumps to 112 for the 1980’s, 3,825 for the 1990’s, and all the way up to 7,948 today. Khan (2006) argues that if states can ensure 59

efficient markets, by enforcing property rights, rule of law, reducing corruption and committing not to expropriate, then private investors will drive economic development. He claims that there are two types of governance: market enhancing and growth enhancing. Growth enhancing governance is to be used more for underdeveloped or developing nations. Khan did not offer much hard evidence for his theories but they were taken into account when deciding what variables to use for the research. Blume and Durlauf (2006) examine the topic on a more microeconomic level. They compare the following institutions: state politico-legal institutions, private ordering within the law, for-profit governance, and social networks and norms. They argue that private institutions are limited in size and as economic activity expands a transition towards formal institutions is usually observed. Their variables (crime, credit rating agencies) were a too specific for the research but allowed me to determine what aspects of a government had the most weight on its economic stability. Sen (2006) echoes the author’s thesis that an “exchange economy could not be very efficient until business morality made contractual behavior sustainable and inexpensive.” He proposed that governance is also instrumental in curbing corruption, which drives potential entrepreneurs away from productive operations to the higher returns offered by shady deals. This would later lead to the consideration of various political variables and result in the use of the variable of government effectiveness. Garrison and Lee (1995) found evidence supporting one of this studies’ hypotheses yet predicted other data to come out contrary to some of the author’s expected signs. They tried to prove that countries that pursue macroeconomic policies that result in high inflation, large budget deficits, and high levels of government consumption spending show no signs of suffering from low rates of growth per capita. They were able to show weak evidence for a negative effect of marginal tax rates and strong evidence for a strong economic growth due to an expansion in foreign trade. A lot of ideas for variables were drawn from their work but the scope of this research was narrowed to just developed nations. Jalilian et al. (2006) also did a study on the impact of regulation on economic growth; however they focused on specifically developing countries. They were able to find a strong casual link between regulatory quality and economic performance. Gupta et al. (1998) provided a study focusing more on the relationship between democracy and economic growth. They did however intertwine that with political stability. Their studies supported the decision to remove the dummy variables for democracy from this study but they further justified the importance of political stability for any nation’s growth. The paper had a very low R2 which 60

showed that the variables did not fit the model or explain the dependent variables that well. The author tried to refine their regression models to have a more accurate regression. 3.0 DATA AND EMPIRICAL METHODOLOGY 3.1 Definition of Variables CGDP=β0 + β1GOVEXP + β2LFDII + β3LGOVEFF + β4LEDUGOV +β5MILEXP + β6LIINF + ε CGDP is the real GDP per capita; used as an endogenous variable. It is defined as the gross domestic product (GDP) at purchasing power parity (PPP) per capita, the value of all final goods and services produced within a nation in a given year divided by the average population that same year. GDP dollar estimates here are derived from purchasing power parity calculations; using the long term equilibrium exchange rate of two currencies to equalize their purchasing power. The data and definition from this paper is consistent with those given from the World Bank Group. The independent variables will consist of seven variables obtained from the World Bank Group and the International Monetary Fund. Figure 3 provides information on the data source, acronyms, descriptions, expected signs, and justifications for using the variables. The first provided variable is government consumption expenditure as a percentage of GDP, GOVEXP. After that the paper shows net inflows of foreign direct investment as a logged variable, LFDII. The next used variable was the logged government effectiveness index, LGOVEFF. This was followed by LEDUGOV , the logged version of a variable for government public education expenditure as a percentage of GDP. MILEXP was government military expenditure as a percentage of GDP. Lastly we have LIINF, the log for the inflation percentage since the year 2000. (Figure 3)

Acronym

Variable Description

What it captures

Expected sign

Data source

CGDP

GDP Per Capita

Country’s

Dependent Var.

International Monetary

GDP/Population GOVEXP

Government

Government Fiscal

Consumption +

Policy and Management

Government

of Money

Fund +

International Monetary Fund

Investment+ Transfer of

Payments as a % of GDP

LFDII

Logged value of the

Global Participation

World Bank Group

Logged, Rating on

Efficiency, Corruption,

World Bank Group

management

Political Stability

Value of human capital

World Bank Group

Government Military

Whether necessary for

World Bank Group

spending as % of GDP

economic growth and

World Bank Group

amount other country’s invest in you LGOVEFF

effectiveness of given government. LEDUGOV

Logged, Government Spending on Public Education as % of GDP

MILEXP

stability LIINF

Logged, change in

Government

inflation

management of money supply

3.2 Data The study used annual data from the years 1982 to 2000, compiled as an average. It is based on a time series cross-section data. Data was obtained from the International Monetary Fund and the World Bank Group. The countries used were those classified in the CIA World Factbook as “developed countries.” These mainly democratic, market oriented countries are generally regarded as first world, high income, and industrial and have a per capita GDP in excess of $10,000. After that, the author narrowed the list of used countries to those that were able to provide all of the relevant data to make an accurate study. 3.0 EMPIRICAL RESULTS The results showed the three variables of the net inflows of foreign direct investment, the government efficiency indicator, and the government education as a percent of GDP are all highly significant for the data. They all hold 3 star ratings and have t-statistics greater than 3.5. This allows them to reject the null hypothesis on a 99% confidence level. The regression showed to have a good R2 indicating that the author has a good proportion of data showing variability in my model. The adjusted R2 shows a penalization for extra variables included in the model. The shown drop in this studies’ adjusted R2 is probably because of how general the variable government expenditure is and how it might relate with government spending on both education and military and even possibly with foreign investment. In the future one should 62

probably try to subtract the impact of those variables out of the government expenditure data but the exact data for that was unavailable and the author’s intentions with that variable was to broaden the scope to better view the effects that a good fiscal policy have on the dependent variable. None the less, the adjusted R2 still showed good variability in my model. The next Variable

Obs.

Mean

Std. Dev.

Min

Max

CGDP

24772.54

6969.266

13244.16

50334.58

GOVEXP

.3999

3.5850

-6.6128

11.94

LFDII

3.7576

.0865

1.3079

5.1874

LGOVEFF

.0539

.4069

-2

.3747

LEDUGOV

1.1270

.1076

.9047

29.24

MILEXP

6.9672

6.4907

29.24

LIINF

.0350

.2378

-.5086

.8319

statistic to be looked at was the Durbin-Watson stat. The Durbin-Watson statistic is a statistic used to detect the presence of autocorrelation, with a value of 2 showing no apparent autocorrelation, the paper’s value of 2.19 showed a small, but not significant, sign of negative serial correlation. The paper was able to maintain low autocorrelation by logging the variables that had positive data. A summary of the results is presented in Tables 4, 5, and 6. Table 4: Summary Statistics

Variable

Obs.

Mean

Std. Dev.

Min

Max

CGDP

24772.54

6969.266

13244.16

50334.58

GOVEXP

.3999

3.5850

-6.6128

11.94

LFDII

3.7576

.0865

1.3079

5.1874

LGOVEFF

.0539

.4069

-2

.3747

LEDUGOV

1.1270

.1076

.9047

29.24

MILEXP

6.9672

6.4907

29.24

LIINF

.0350

.2378

-.5086

.8319

Table 5: Regression Results GOVEXP

LFDII

LGOVEFF

LEDUGOV

MILEXP

LIINF

R2:

Obs.

Adjusted:

215.95

2985.93

19619.54

9130.406

177.3183

-3318.127

.7254

(1.02)

(4.00***

(3.79)***

(3.58)***

(0.92)

(-0.89)

(.6629)

Table 6 : Covariance Analysis Covariance Analysis: Ordinary Covariance Correlation

CGDP

32274526

GOVEXP

LFDII

LGOVEFF

LIINF

LEDUGOV

MILEXP

1.000000

GOVEXP

LFDII

LGOVEFF

LIINF

LEDUGOV

6759.813

11.20118

0.355527

1.000000

2742.375

-0.113351

0.715665

0.570613

-0.040035

1.000000

622.3335

0.190137

0.035382

0.024556

0.699066

0.362543

0.266902

1.000000

-248.3580

-0.068750

-0.017790

-0.003154

0.029574

-0.254210

-0.119449

-0.122280

-0.117055

1.000000

60.46457

0.135519

-0.033717

0.002695

-0.001135

0.008245

0.117213

0.445934

-0.438939

0.189403

-0.072654

1.000000

MILEXP

1048.360

-2.912573

1.049281

-0.232058

-0.055600

0.002275

16.32172

0.045677

-0.215408

0.307011

-0.366554

-0.080027

0.006202

1.000000

This paper had predicted that government expenditure could be positively or negatively correlated depending on just how involved the government would become. Government spending is good for the economy because it would provide public goods such as infrastructure. On the other hand government could hinder economic growth by transferring additional sources from the productive sector of the economy to the government, which would use them less efficiently. The research proves that economic growth has been achieved and that governments can choose to intervene and try to improve the nations through services such as public work projects (Jalilian et al. 2007). This paper found very high significance between the inflow of cash from foreign direct investment and economic growth and stability. Previous literature on the subject has not shown as much of a correlation. The rise in the significance indicated how the global market in itself is growing and how now it is essential to be an active competitor. The author hypothesized a positive correlation and was hoping to prove that the significance would be higher than the current literature out on the subject. That is because direct investment creates new production capacity and jobs. Foreign investment can help to transfer technology and know-how and can strengthen linkages to the global marketplace. Research has linked increases in foreign investment to higher national wages (Jalilian et al. 2007). Criticisms of foreign investment have citied a loss of market share for competing domestic firms and flow straight back to the multinational’s economy. The research shows that the pros outweigh the cons considerably. The next variable is the government effectiveness index as given by the World Bank Group. On the group’s website they state that “the authors draw 194 different measures from 17 different sources of subjective governance data constructed by 15 different organizations.” The sources include international organizations, political and business risk agencies, think tanks, and non-government organizations. The index combines perceptions of the quality of public service provision, the quality of bureaucracy, the competence of civil servants, the independence of the civil service from political pressures, and the credibility of the government’s commitment to policies into a single grouping (World Bank, 2007). The main focus of the index is on the inputs required for the government to be able to produce and implement good policies and deliver 65

public goods. The website adds “the component indicators are aggregated using an unobserved data in each cluster as a linear function of the unobserved common component of governance, plus a disturbance term capturing perception errors and/or sampling variation in each indicator. Higher or positive values indicate government effectiveness.” The regression indicates that government effectiveness is very significant through both a three star probability and a high t-statistic. The data also met the initial predictions of the data having a positive correlation. The author hypothesize that this is due to the fact that once a country has a developed economy that next important step is to manage it efficiently. The research has shown that countries with a troubled economy have had a correlation of a tumultuous political system as well; therefore it is fair to assume the opposite for countries with a strong economy. The government spending on public education was the next variable and it also had a very high, three-star, significance and a high t-statistic of 3.57. The research has shown that basic education is crucial because it improves worker production. The author was correct with my prediction for a positive correlation. The ability to educate your population allows you to better train future employment and should be reflected in your economy, as it was in the sample data for my regression. The study also enjoyed a three star significance, along with a t-statistic of 3.57 for the variable of government expenditure on its military. It has been generally noted through observation of today’s media that the developed countries of the world all hold a strong military so it was assumed that the study would find a positive correlation between the two. This follows the principles of Military Keynesianism. The theory states that increased military demand for goods and services is generated directly by government spending and that this direct spending generates a multiplier effect of general consumer spending. On the supply side maintenance of a standing army removes many workers, usually young males with less skills and education, from the work force; giving this demographic a high level of unemployment. Enlistment in the Unites States has even been touted as offering opportunities to gain education and certain skills for later employment (Blume and Darlauf, 2006). The final used variable was inflation percentage. The variable was logged and was observed on the basis of growth from the year 2000. The author predicted that inflation would have a negative correlation with the GDP per capita. The variable was indeed negative but did 66

not have a enough significance to reject the null hypothesis. The prediction was based on research which showed that inflation would take away from a country’s incentive to save. A higher inflation tends to also cause investors to take more systematic risks in order to compensate for their inflations. Wages also have a hard time catching up with inflation. All of these factors will help to cause debt in the individual and corporate sectors. 5.0 CONCLUSION In summary, Foreign Direct Investment inflows, effective governance, and increases to human capital have shown to be essential in order for a developed country to continue to prosper. High inflation hinders growth by adversely affecting the exchange mechanism and by distorting the taxation of capital (Garrison and Lee, 1995). This paper hypothesizes that high inflation and Government effectiveness both are good indicators to how well a government is managed. It is just as important for a developed economy to compete with foreign rivals in the export and investment sector. This allows you to take advantage in the economies of scale. In the future, to further improve the shown research, one should be sure to remove defense and education spending since they contribute to private sector productivity and property rights, which in turn are reflected in private investment. This study adds to the already available information on the subjects by confirming and updating outdated studies on the said theories. It focuses the scope on just developed countries; something that has not yet been done. Lastly, the study tries to highlight the increased importance of FDI. I conclude that a regulatory regime that promotes economic growth is an important part of good governance. The ability of the state to provide effective regulatory institutions can determine how well markets and the economy will perform. The impact will depend on the efficiency of the policies and the quality of the governance.

BIBLIOGRAPHY Durlauf, Steven, and Lawrence Blume. "Economic Governance." New Palgrave Dictionary of Economics (2006). Garrison, Charles B., and Feng Yao Lee. "The Effect of Macroeconomic Variables on Economic Growth Rates: a Cross-Country Study." Journal of Macroeconomics 17 (1996): 303-317. Gupta, Dipak K., M.c. Madhaven, and Andrew Blee. "Democracy, Economic Growth and Political Instability; an Intergrated Perspective." Journal of Socio-Economics 27 (1998): 587-611. "International Monetary Fund." <www.IMF.org>. Jalilian, Hossein, Colin Kirkpatrick, and David Parker. "The Impact of Regulation on Economic Growth in Developing Countries: a Cross-Country Analysis." World Development 35 (2006): 87-103. Khan, Mushtaq H. Governacne, Economic Growth, and Development Since the 1960's: Background Paper for World Economic and Social Survey. 2006. Lee, Bryan. "Good Governance Key to Progress,." The Straits Times 15 Sept. 2006. World Bank Group. <www.worldbank.org>.

An Analysis of the Female Labor Force Participation Rate in the U.S. 1980-2004

Larry Martinsa

Abstract: This research uses census data from the Bureau of Labor Statistics to examine the female labor force participation from 1980 to 2004. These statistics are used to find the determinants of women’s decisions to enter the job market. The purpose of studying the female involvement in the labor force is to illustrate if females are still having trouble in the market. This article also reviews historical labor force statistics to determine how the labor force has changed and which factors have affected its changes between 1980 and 2004. The model, estimated with U.S. data, has provided empirical support for the underlying theoretical predictions and variables. Analysis of twenty-five years of data suggests that having children, an education, a husband, and a contribution to family income determine women’s labor force participation in the United States.

JEL Classification: J12, J16 Keywords: Economics of Gender, Marriage.

Department of Economics, Bryant University, 1150 Douglas Pike, Smithfield, RI 02917. Phone: (401) 726-0281. Email: lmartins@bryant.edu.

1.0 INTRODUCTION One of the most striking issues of recent times has been the extent to which women have increased their share of the labor force; the increasing participation of women in paid work has been driving employment trends and the gender gaps in labor force participation rates have been shrinking. The increase in women's participation in the U.S. labor force is one of the most important social, economic, and cultural trends of the past century. The growing proportion of women doing paid work has transformed gender relations, changed patterns of marriage and childbearing, and is often viewed as a key indicator of women’s progress toward gender equality in the labor force. This study tries to explain why the women’s labor force participation rates have increased over time, yet leveled off since 1990. The reasons for this are not well understood. It may be that unprecedented economic growth during the 1990s raised men’s incomes to the point that some married women opted out of the labor force. Another possibility is that women’s ability to balance work responsibilities inside and outside of the home may finally have reached a limit. As women’s labor force participation rates have increased, the time available for raising children and doing household chores has been compressed, creating stress for families and particularly for working mothers. Some people believe that an increasing proportion of women are now choosing to stay at home to avoid this work-family conflict. This paper was guided by two research objectives that differ from other studies: First, it investigates the possibility of male unemployment as a factor in female participation rates. Second, it investigates children as a factor. The rest of the paper is organized as follows: Section 2 illustrates the trends in labor force participation. Section 3 gives a brief literature review and section 4 examines the data and

estimation methodology. Finally, section 5 discusses the empirical results, which is followed by a conclusion in section 6. 2.0 TRENDS Women’s labor force participation rate in the United States grew from 33 percent in 1948 to 59 percent in 1995. Over the same period, the rate for men fell gradually from 87 percent to 75 percent. As a result, the gap between women’s and men’s participation shrank from 54 percentage points in 1948 to 16 points in 1995. The most rapid growth in women’s participation occurred from 1975-85. Subsequently, growth slowed and, since 1990, the labor force participation rate for women has been at a virtual standstill. Factors underlying this leveling-off included declines in participation among young women under age 25, the long-term slowing of participation rate growth among women 25 to 44 years old, and an unusually slow employment rebound from the 1990-91 recession. In 1970, about 43 percent of women ages 16 and older were in the labor force. By 2000, 61 percent of adult women were in the labor force. Over the same period, men’s labor force participation rates declined from 78 percent to 74 percent. These trends are part of the rapid increase in women’s labor force participation, combined with the simultaneous decline in men’s participation. This has closed much of the gender gap in the labor force. In 2000, about 47 percent of people in the labor force were women. If current trends continue, women will soon make up the majority of the U.S. work force.

Figure 1: Labor Force Participation Rates for Men, Women, and Women with Children Under 6, 1970-2000

Source: U.S. Census Bureau

In 1950, about one out of three women participated in the labor force. By 1998, nearly three out of every five women of working age were in the labor force. Among women age 16 and over, the labor force participation rate was 33.9 percent in 1950, compared to 59.8 percent in 1998. Figure 2: Labor Force Participation Rates Of Women by Age, 1950 and 1998

Source: Bureau of Labor Statistics

Changes in labor force participation varied by age group. The biggest increase in labor force participation was among those ages 25 to 34: their rate more than doubled, from a level of 34.0 percent in 1950 to 76.3 percent in 1998. Also, in 1950, women age 16 to 24 had the highest 72

labor force participation rate (43.9 percent); in 1998 women age 35 to 44 had the highest rate (77.1 percent), followed closely by those age 25 to 34 (76.3 percent) and those age 45 to 54 (76.1 percent). The only age group to experience a decline in labor force participation between 1950 and 1998 was those age 65 and over. The rate for women in this age group dropped from 9.7 percent to 8.6 percent. As more women are added to the labor force, their share will approach that of men. In 2008, women will make up about 48 percent of the labor force and men 52 percent. In 1988, the respective shares were 45 and 55 percent. In the near future, for the first time in United States history, the female participation rate will equal, and possibly surpass, the male participation rate. Figure 3: Shares of the Labor Force by Sex, 1988-2008

Source: Bureau of Labor Statistics

3.0 LITERATURE REVIEW Semyonov (1980) focused on women’s labor force participation as a characteristic of the social structure. He utilizes data from sixty-one societies. First, the analysis demonstrates that participation is positively related to economic development and divorce rate and negatively related to fertility and income inequality. It illustrates that the most significant effect on female labor 73

force participation is that of income inequality. In societies where inequality is high, women are less likely to join the labor force. The analysis goes on to demonstrate that female labor force participation has consequences for job discrimination. The odds that women can achieve high status and well-paid occupations decrease with the proportion of women in the labor force. The findings reported here suggest that the integration of females into the labor force is determined by the shape of the stratification system. Such integration, however, results in job discrimination. Fullerton (1999) illustrates how women’s labor force participation rates have increased significantly over the past 50 years, narrowing the gap between rates for women and men. Between 1950 and 1998, most of the increase in the labor force participation rate occurred between 1970 and 1990. During this 20 year period, the participation rate jumped from 60.4 percent to 66.4 percent. This labor force increase occurred because of the baby boom generation. There was also a 14.2 percent increase in the aggregate labor force participation rate for women. This article reviews historical labor force statistics to determine how the labor force has changed and which factors have affected its changes between 1950 and 1998. It focuses on labor force trends of men and women and also discusses the projected changes in the labor force from 2015 to 2025. Lingle (1978) examines the unemployment rates of females. One of the conclusions was that female rates have tended to exceed male rates since World War II. In this paper, the author uses cross-section data from the Censuses of Population of 1960 and 1970. He tries to examine whether the relationship between female and male unemployment rates shifted during the decade. He presents a model for examining the structure of female unemployment rates using cross-section data and then tests for a parameter shift between 1960 and 1970 in the relationship between male and female unemployment.

Hill (1984) observes the female labor force participation in Japan, using aggregate crosssectional data. The empirical results reinforce her previously observed similarities between the behavioral responses of women in the U.S. regarding the decision to enter the labor force and those of Japanese women regarding that same decision. However, there are a few shortcomings in this analysis. For example, the wage rate reported by those women who are working may not appropriately measure the wage that a woman out of the labor force would receive if working. In the U.S., labor force participation usually means that a woman leaves home to work for someone else. However, in Japan and in other countries characterized by labor markets, this may not be the case. Nam (1991) investigates the determinants of labor force participation of women living in male-headed households in Seoul, South Korea. Analysis of data from the 1970 and 1980 Korean Population Censuses suggests that both women’s educational level and the family economic status determine their labor force participation in Seoul. Women with middle school education or above are more economically active than those with no education. Women from lower economic backgrounds are almost two to three times more likely to be employed than those in high-status families, controlling for age, number of children under the age of 6, and marital status. Cotter, Hermsen, and Vanneman (2001) investigate how the demand for female labor is a central explanatory component of economic theories of gender stratification. The study analyzes how the structural demand for female labor affects gender differences in labor force participation. They develop a measure of the gendered demand for labor by indexing the degree to which the occupational structure is skewed toward female occupations. Using census data from 1910 through 1990 and National Longitudinal Sample of Youth (NUY) data from 261 contemporary

U.S. labor markets, the authors shows the importance of the topic of gender differences in labor force participation. 4.0 DATA AND EMPIRICAL METHODOLOGY 4.1 Definition of Variables The basic model follows the regressions of Moshe Semyonov and Christopher Lingle. In the early 1980’s, they both carried out regressions relating to the female job market. After using some of their variables, the model chosen for this analysis became:

FPLF = β0 + β1CHILD + β2COLLEGE+ β3MARRIED + β4CONTRIB + β5MULTI + β6MALEUN + β7SIXFIVE + ε

FPLF is the female participation rate in the labor force in the United States. It is characterized as the total employment status of the noninstitutional population of 16 years or older, represented in thousands. Independent variables consist of seven variables obtained from various sources. Appendix A and B provide data source, acronyms, descriptions, expected signs, and justifications for using the variables. First, CHILD represents the total number, in thousands, of employed females with children under the age of 18. Second, COLLEGE illustrates the percent distribution of civilian labor force 25-64 years old with 4 or more years of college. Third, MARRIED corresponds to the total employment status, in millions, of women that are married. Fourth, CONTRIB signifies the total percentage contribution of wives’ earnings to family income. Fifth, MULTI stands for the total number of women, in thousands, who hold multiple jobs. Sixth, MALEUN is the total male unemployment represented in thousands. And finally, SIXFIVE was the female labor participation rate with 65 or more years of age. 76

4.2 Data The study uses annual data from 1980 to 2004. Data was obtained from the Bureau of Labor Statistics (BLS) website as well as the U.S. Census Bureau website. Summary statistics for the data are provided in Table 1. Table 1: Summary Statistics Variable

Obs.

Mean

Std. Dev.

Min

Max

EMP

57953.52

7206.25

45487.00

68421

CHILD

21276.96

2921.75

16526.00

25030

COLLEGE

25.57

4.22

18.60

32.6

MARRIED

30.34

3.70

23.60

34.8

CONTRIB

31228.00

2.53

26.70

35.6

MULTI

3009.72

710.30

1549.00

3800

MALEUN

4339.00

850.79

2975.00

6260

SIXFIVE

11984.00

1.75

9.60

15.5

5.0 EMPIRICAL RESULTS The primary objective of this study was to find the determinants of female participation in the United States job market. In total, out of 7 variables, 4 were significant. Table 2: Regression Results Variable

Coefficient

Child

0.602127***

College

533.0086***

Married

561.3638***

Contrib

599.5685**

Multi

-0.635705

Maleun

-0.178778

t-Statistic 3.0086 4.8764 3.4293 2.1933 -1.4187 -0.8299

Expected Sign + + + + + 77

Sixfive

60.40417

-2273.326

1.1166

+ -0.7824 Note: ***, **, and * denotes significance at the 1%, 5%, and 10% respectively. For women with children, there is a positive influence on their labor force participation rate. However, the positive effect is not a great number. It may be explained by mothers wanting to stay home and take care of their kids, while some mothers want to work so they can provide for their children.

This differs from the results produced in Semyonov’s (1980) regression.

Semyonov’s regression generated a negative fertility variable. He explained that when fertility is high, women are more likely to be occupied with household responsibilities and therefore less likely to be able to join the labor force. When women get an education and receive a college degree, there is also a positive influence on their participation rate.

Education gives people a positive incentive to seek

employment, since education is an investment that is positively correlated with earning potential. Consequently, it raises the opportunity cost of economic inactivity. Also, more-educated women have higher income aspirations over their less-educated counterparts and tend to be more active in the labor market. The more education females get, the better chance they have at obtaining a job. Employers look for a highly educated individual to fulfill the company’s demands, whether it be a male or female. This variable agrees with Lingle’s (1978) female unemployment variable. He presented data that showed the level of female unemployment falls with increasing levels of education. After females decide to get married, their labor force participation rate is significantly increased. Perhaps, this occurs because of the need for a higher family income. The economy

forces families to earn more money due to the high standard of living. There is also pressure from the husband to increase family earnings, if possible. For the contribution of wives’ earnings to family income, there is a positive effect. When a wife can play a role in family salaries, they decide to partake in those roles. The more money they have, the better life the family can live. When the opportunity cost to be a housewife is too high, they will become an active worker in society. This can happen for numerous reasons, all depending on the situation at hand. When women participate in multiple jobs, it should increase their labor force participation rate. It was not significant, however, probably because of the multicollinearity with the dependent variable. It is assumed that when women partake in numerous jobs, their participation rate in the labor force has increased. The male unemployment variable has a negative impact on the female participation rate in the labor force. The fluctuations of unemployment with the business cycle are not useful as an explanation of the long-term growth in women’s labor force participation. Moreover, the unemployment rate does not adequately capture the concept of employment opportunities as it affects labor supply and labor demand factors. The female labor force participation rate with 65+ years of age was not significant. This is possibly due to a high correlation with the dependent variable. This variable already contained the labor force participation rates, except it was for the elderly and not for the entire female population. 6.0 CONCLUSION In summary, the empirical results suggest that certain variables do affect female labor force participation rates in the United States. Having a child, does affect a women’s decision to

enter the labor market; primarily, because it costs the mother an incredible amount of money to nurture the child. Going to college also significantly increases their participation rate. The more educated women become, the smaller the gap becomes between the male labor force participation rate and the rate of females. When women are married, it has an important positive influence on their decision to enter the job market. Being married involves more expenses that need to be paid for. Their contribution to the family income is a big factor because it is difficult for just the husband to support the entire family. All of these factors influence women in deciding on whether or not it is beneficial to enter the labor force. With this regression, it is evident that having children, going to college, being married, and contributing to family income all affect that decision. It will be interesting to see where the United States will be in the near future, with women surpassing males in labor force participation rates. Appendix A: Variable Description and Data Source

Acronym

Description

Data source

FPLF

employment status of the female noninstitutional population 16 years and over (thousands)

US Bureau of Labor Statistics

CHILD

total number of females employed with children under 18 (thousands)

US Bureau of Labor Statistics

COLLEGE

distribution of female labor force 25-64yrs with 4+years of college (percentage)

US Bureau of Labor Statistics

MARRIED

total employment status of women that are married (millions)

US Bureau of Labor Statistics

CONTRIB

contribution of wives’ earnings to family income (percentage)

US Bureau of Labor Statistics

MULTI

total number of women who hold multiple jobs (thousands)

US Bureau of Labor Statistics

MALEUN

total male unemployment (thousands)

US Bureau of Labor Statistics

SIXFIVE

female labor force participation rate with 65+ years of age (percentage)

US Census Bureau

BIBLIOGRAPHY Bureau of Labor Statistics. 5 Mar. 2008 <www.bls.gov>. Cotter, David, Joan Hermsen, and Reeve Vanneman. "Women's Work and Working Women: the Demand for Female Labor." Gender and Society 15 (2001): 429-452. 5 Mar. 2008. Fullerton, Howard. "Labor Force Participation: 75 Years of Change, 1950-98 and 1998-2025." Bureau of Labor Statistics Monthly Labor Review (1999). 5 Mar. 2008. Hill, Anne. "Female Labor Force Participation in Japan: an Aggregate Model." The Journal of Human Resources 19 (1984): 280-287. 5 Mar. 2008. Lingle, Christopher, and Ethel Jones. "Women’s Increasing Unemployment: a Cross-Sectional Analysis." The American Economic Review 68 (1978): 84-89. 5 Mar. 2008. Nam, Sunghee. "Determinants of Female Labor Force Participation: a Study of Seoul, South Korea, 1970-1980." Sociological Forum 6 (1991): 641-659. 5 Mar. 2008. Semyonov, Moshe. "The Social Context of Women's Labor Force Participation: a Comparative Analysis." The American Journal of Sociology 86 (1980): 534-550. 5 Mar. 2008. U.S. Census Bureau. 5 Mar. 2008 <www.census.gov>.

Economic and Social Factors that Influence Life Expectancy and Infant Mortality

Kenneth Smith

Abstract: This paper investigates life expectancy and infant mortality rates in randomly selected countries from around the world. This study takes into consideration many economic and social factors that could potentially have an effect on life expectancy and infant mortality in different countries. Factors are modeled using a least-squared regression model and are determined as significant based on their probability factor. Results show both positive and negative effects depending on the economic and social factor.

JEL Classification: I10, I30 Keywords: Life Expectancy, Infant Mortality

Department of Economics, Bryant University, 1150 Douglas Pike Smithfield, RI 02917. Phone: (774) 254-1241. Email: ksmith5@bryant.edu.

______________________ The author thanks Ramesh Mohan for his help and guidance throughout the writing of this paper. The author would also like to thank Lindsay Kahler and Jonathan Stachelek for their reviews of this paper. 82

1.0 INTRODUCTION Life expectancy and infant mortality are two statistics that vary in great measure depending on which country around the world you are talking about. Thirty-nine randomly selected countries (See Appendix F) were taken in consideration overall, and the range from the highest life expectancy to the lowest one was nearly 45 years (Angola = 37.63 years - Japan = 82.02 years). The same two countries also had the highest and lowest infant mortality rates with Japan at about .28% while Angola’s is about 18.44%. These are staggering numbers and one can only help but wonder what factors influence these two statistics. Do economic and social factors have an impact on life expectancy and infant mortality? This paper will seek to answer this question and show the empirical results of models that were developed to answer this question. The paper will also look at which factors do and do not affect life expectancy and infant mortality and whether these effects have a positive or negative correlation. Finally, this paper will explain why these factors have an impact on life expectancy and infant mortality. The recent trend in life expectancy over the past 20-30 years has been an increasing one. On the other hand, infant mortality has had the opposite trend as it has been decreasing. These two are correlated because as infant mortality decreases, there are less people (infants) dying at a young age which increases the overall life expectancy age. Life expectancy has also been increasing due to medical advancements and people being more aware of their overall health. As medical care becomes more affordable people can have regular doctor’s appointments to regulate and watch their health. This leads to a longer life and a higher overall life expectancy age. Figure 1: Modern World Trend in Life Expectancy

www.childtrendsdatabank.org

In terms of infant mortality, medical advancements have also had a big impact. There have been increased technologies in the medical field that help track the health of babies throughout the 83

birth process and this can lead to treatment if necessary. Increased medical treatment and attention leads to healthier infants being born and lessens the overall infant mortality rate. Figure 2: Modern World Trend in Infant Mortality

www.childtrendsdatabank.org

Despite the increased medical care and technologies, there are still hindrances on life expectancy and infant mortality. The HIV/AIDS virus continues to grow in countries all over the world and especially on the continent of Africa. The lack of a solution for this virus is causing an increasing rate in deaths all over the world. As this virus continues to grow and spread, more people will continue to die and life expectancy will grow at a slower rate. When medical advances can be used to cure this virus and countries can learn all of the factors that are affecting life expectancy and infant mortality, then life expectancy will grow more rapidly and infant mortality will continue to decline around the world. The purpose of this paper is to increase the understanding of what economic and social factors have an impact on infant mortality and life expectancy, and how countries can use this information to better help the people of their country. A number of these factors can be addressed in certain ways to help lower infant mortality and increase life expectancy, but most countries either do not know what these factors are or do not know how to fix them. This study will increase their knowledge on this subject and hopefully the information can be used to determine the most influential factors, and then, take the corrective steps to fix these problems. The rest of the paper is organized as follows: Section 2 gives a literature review, Section 3 outlines the Data and Empirical Methodology, Section 4 presents and discusses the empirical results, and finally this is followed by a conclusion in section 5.

2.0 LITERATURE REVIEW Five papers in total were used to build the model for this paper and help determine the results of the two models. The five papers have been combined to build one large model that takes into consideration various economic and social influences on life expectancy and infant mortality. The base model being used for this experiment comes from the paper by Kennelly et al (2002) and then other variables were added from the other papers whose models contained different variables. A few variables such as Unemployment and HIV/AIDS Deaths, that were not a part of any other models, were added to get a broader range of ideas for what could have an effect on life expectancy and infant mortality. Kennelly et al (2002) use their model to determine the relationship between social capital and public health. In terms of public health they meant life expectancy and infant mortality as those were the two dependent variables used in their paper. In the basic model for their paper Kennelly et al (2002) used determinants such as GINI, GDP per capita, public health expenditure, doctors, fruit and vegetables, alcohol, and tobacco. In this paper these authors conclude that social capital had no effect on public health, but that GDP and public health expenditure do have a positive and negative effect on life expectancy and infant mortality respectively. In another paper by Houweling et al (2001), they compared infant mortality to inequality in various countries. What was found was that literacy, GDP per capita, and inequality all influenced child mortality while healthcare expenditure had no effect. These factors are congruent with factors that are being run in the regression of this paper. In a similar paper by Flegg (1982) we see a comparison of income, literacy, and medical care to infant mortality in various countries. In Flegg’s (1982) model he uses GDP, GINI, literacy, and physicians as determinants of infant mortality along with a few other variables. These four factors are variables used in this paper to determine the causes of both life expectancy and infant mortality. The conclusion in Flegg’s paper was that GINI, illiteracy, nurses, and physicians all had an impact on infant mortality with nurses and physicians having a negative impact and GINI and illiteracy having a positive impact. In Gortmaker’s (1979) comparison of poverty and infant mortality we see a strong comparison between the two. Although Gortmaker took into consideration many other variables,

poverty was the only one used in this paper. From his research he concludes that if poverty is higher in one country than another, infant mortality will also be higher in that country. Last, we have a paper that shows the effect of the HIV/AIDS virus on life expectancy all over the world. In Neumayer’s (2004) account on life expectancy we see that having the HIV/AIDS virus at birth has a significant impact on a person’s life expectancy. This demonstrates a major problem occurring in certain countries around the world because most people do not know if they have the virus or not. The conclusion of this paper shows the virus has had a negative impact on life expectancy up until recently where new technologies, medicines, and better healthcare have changed that. The determinants of all the preceding papers have been taken into account and combined to make models for life expectancy and infant mortality. These models will determine which factors do and do not impact life expectancy and infant mortality. These various economic and social factors plus the addition of a few extra determinants will be used in a basic regression model and observed to determine their impact on life expectancy and infant mortality in countries across the world. 3.0 DATA AND EMPIRICAL METHODOLOGY 3.1 Definition of Variables Life Expectancy 𝑌𝑖 𝐿𝑖𝑓𝑒𝐸𝑥𝑝𝑒𝑐𝑡 = 𝛽0 + 𝛽1 𝐼𝑛𝑓𝑎𝑛𝑡 + 𝛽2 𝐻𝐼𝑉/𝐴𝐼𝐷𝑆 𝑃𝑒𝑜𝑝𝑙𝑒 + 𝛽3 𝑃𝑜𝑣𝑒𝑟𝑡𝑦 + 𝛽4 𝑃𝑢𝑏𝐻𝑒𝑎𝑙𝑡ℎ + 𝛽5 𝑆𝑚𝑜𝑘𝑖𝑛𝑔 + 𝛽6 𝑈𝑛𝑒𝑚𝑝𝑙𝑜𝑦 + 𝑢𝑖

Life expectancy is the amount of years that a certain person is expected to live. This statistic varies from country to country with certain factors causing it to increase and others causing it to decrease. Life expectancy in this model is the dependent variable and it is represented by 𝑌𝑖 𝐿𝑖𝑓𝑒𝐸𝑥𝑝𝑒𝑐𝑡 in this model which shows what social and economic factors have an influence on Life Expectancy. The Independent variables in this equation consist of six variables that were obtained from various sources. First, 𝛽1 𝐼𝑛𝑓𝑎𝑛𝑡 shows the infant mortality rate of the different countries and explains its effect on life expectancy around the world. Second, 𝛽2 𝐻𝐼𝑉/𝐴𝐼𝐷𝑆 𝑃𝑒𝑜𝑝𝑙𝑒 shows the amount of people in each country that contain the HIV or AIDS virus. Third, 𝛽3 𝑃𝑜𝑣𝑒𝑟𝑡𝑦 shows the amount of people in each country that live below the poverty line. Fourth,

𝛽4 𝑃𝑢𝑏𝐻𝑒𝑎𝑙𝑡ℎ shows the public health expenditure per capita in each country. Fifth, 𝛽5 𝑆𝑚𝑜𝑘𝑖𝑛𝑔

shows the amount of cigarettes consumed in a country per capita. Sixth and final, 𝛽6 𝑈𝑛𝑒𝑚𝑝𝑙𝑜𝑦 is the unemployment rate in each country for the most recent year. Infant Mortality 𝑌𝑖 𝐼𝑛𝑓𝑎𝑛𝑡 = 𝛽0 + 𝛽1 𝐷𝑜𝑐𝑡𝑜𝑟𝑠 + 𝛽2 𝐺𝐷𝑃 + 𝛽3 𝐻𝐼𝑉/𝐴𝐼𝐷𝑆𝐷𝑒𝑎𝑡ℎ𝑠 + 𝛽4 𝐻𝐼𝑉/𝐴𝐼𝐷𝑆𝑃𝑒𝑜𝑝𝑙𝑒 + 𝛽5 𝑃𝑜𝑣𝑒𝑟𝑡𝑦 + 𝛽6 𝑃𝑈𝐵𝐻𝑒𝑎𝑙𝑡ℎ + 𝛽7 𝑆𝑚𝑜𝑘𝑖𝑛𝑔 + 𝛽8 𝑈𝑛𝑒𝑚𝑝𝑙𝑜𝑦 + 𝑢𝑖

Infant Mortality is the percent of babies that pass away from birth up to an age of one year old. In this paper the rate is formed by using the amount of infant deaths per 1000 births. Infant Mortality in this model is represented by 𝑌𝑖 𝐼𝑛𝑓𝑎𝑛𝑡 and it is the dependent variable. There are eight variables in this model and this regression will show which factors do and do not have an impact on Infant Mortality. The dependent variables in this model consist of eight variables that were obtained from various sources. First, 𝛽1 𝐷𝑜𝑐𝑡𝑜𝑟𝑠 is the amount of doctors per 100,000 people in a certain country. Second, 𝛽2 𝐺𝐷𝑃 is the Gross Domestic Product in country relative to PPP. Third, 𝛽3 𝐻𝐼𝑉/𝐴𝐼𝐷𝑆𝐷𝑒𝑎𝑡ℎ𝑠 shows the amount of people that have died from HIV/AIDS in a country in

the past year. Fourth, 𝛽4 𝐻𝐼𝑉/𝐴𝐼𝐷𝑆𝑃𝑒𝑜𝑝𝑙𝑒 is the amount of people in a country that have the HIV or AIDS virus. Fifth, 𝛽5 𝑃𝑜𝑣𝑒𝑟𝑡𝑦 is the amount of people that live under the poverty line a certain country. Sixth, 𝛽6 𝑃𝑢𝑏𝐻𝑒𝑎𝑙𝑡ℎ is the amount of money per capita spent on public healthcare in a given country. Seventh, 𝛽7 𝑆𝑚𝑜𝑘𝑖𝑛𝑔 is the amount of cigarettes smoked per capita in a given year for each country. Eighth and final, 𝛽8 𝑈𝑛𝑒𝑚𝑝𝑙𝑜𝑦 shows the unemployment rate for each country in the past year. 3.2 Data This study uses data that was most recently available, but most of the data comes from the past few years (2005-2007). The data for this topic came from various sources, but the majority of the information has come from the World Fact Book which can be found on the website of the Central Intelligence Agency (www.cia.gov). The data of countries that was not available through the World Fact Book was either not found or retrieved from other sources. The research data for this topic can be found in Appendix A.

Variable Description and Data Source Economic/Social Factor

Description of Factor

Source

Doctors

Number of doctors for every 100,000 people in a country

http://earthtrends.wri.org/searchable_db/index.php? theme=4&variable_ID=1297&action=select_countries

GDP (PPP)

Overall GDP of a country

https://www.cia.gov/library/publications/the-worldfactbook/index.html

HIV/AIDS (# of People)

Number of people in a country that have HIV or AIDS

http://hivinsite.ucsf.edu/global?page=cr06-mo-00

Number of people in a country who dies from AIDS in the past year

http://hivinsite.ucsf.edu/global?page=cr06-mo-00

Infant Mortality

Number of infant deaths per 1,000 births

https://www.cia.gov/library/publications/the-worldfactbook/index.html

Life Expectancy

Average age a person lives to in a given country

https://www.cia.gov/library/publications/the-worldfactbook/index.html

Poverty

Percent of the population that lives under the poverty line

https://www.cia.gov/library/publications/the-worldfactbook/index.html

Public Health

Public health expenditure as a percent of GDP

http://www.infoplease.com/ipa/A0934556.html

Smoking

Number of cigarettes consumed per capita in a given year

http://www.nationmaster.com/graph/hea_tob_cig_conhealth-tobacco-cigarette-consumption

HIV/AIDS (Deaths)

https://www.cia.gov/library/publications/the-worldfactbook/index.html

http://www1.worldbank.org/tobacco/database.asp Unemployment

Unemployment rate

https://www.cia.gov/library/publications/the-worldfactbook/index.html

Economic/Social Factor

Description of Factor

Source

Doctors

Number of doctors for every 100,000 people in a country

http://earthtrends.wri.org/searchable_db/index.php? theme=4&variable_ID=1297&action=select_countries

GDP (PPP)

Overall GDP of a country

https://www.cia.gov/library/publications/the-worldfactbook/index.html

HIV/AIDS (# of People)

Number of people in a country that have HIV or AIDS

http://hivinsite.ucsf.edu/global?page=cr06-mo-00

Number of people in a country who dies from AIDS in the past year

http://hivinsite.ucsf.edu/global?page=cr06-mo-00

Infant Mortality

Number of infant deaths per 1,000 births

https://www.cia.gov/library/publications/the-worldfactbook/index.html

Life Expectancy

Average age a person lives to in a given country

https://www.cia.gov/library/publications/the-worldfactbook/index.html

Poverty

Percent of the population that lives under the poverty line

https://www.cia.gov/library/publications/the-worldfactbook/index.html

Public Health

Public health expenditure as a percent of GDP

http://www.infoplease.com/ipa/A0934556.html

Smoking

Number of cigarettes consumed per capita in a given year

http://www.nationmaster.com/graph/hea_tob_cig_conhealth-tobacco-cigarette-consumption

HIV/AIDS (Deaths)

https://www.cia.gov/library/publications/the-worldfactbook/index.html

http://www1.worldbank.org/tobacco/database.asp Unemployment

Unemployment rate

https://www.cia.gov/library/publications/the-worldfactbook/index.html

4.0 EMPIRICAL RESULTS The purpose of this paper is to determine which economic and social factors impact life expectancy and infant mortality around the world. Using a simple least-squares regression model, data of different factors was entered to obtain results. Thirty-nine countries were taken into consideration, but not all the data could be found for each country. Each regression takes into consideration a different number of observations depending on the data that was available for each factor in the different countries. For both Infant Mortality and Life Expectancy there were five variables that were statistically significant. The results of these regressions can be found in Appendix D for Infant Mortality and in Appendix E for Life Expectancy. 89

Infant Mortality In terms of Infant Mortality there were eight variables that were tested, but only five proved to be statistically significant. HIV/AIDS Deaths, Poverty, Unemployment, Doctors, and HIV/AIDS People were all found to be significant while Public Health, Smoking, and GDP were found not to have any significance. Of these five variables, Doctors was the only one found to be significant at the 1% level, while HIV/AIDS Deaths, HIV/AIDS People, and Unemployment were all found to be significant at the 5% level. Also, Poverty was the only variable found to be significant at 10% level. These five variables all proved to have some type of influence on Infant Mortality in countries around the world. The results of the infant mortality regression can be found in Appendix D. Results – Infant Mortality Variable

Observations

Probability

Coefficient

Min

Max

HIV/AIDS DEATHS POVERTY

0.0239**

-8.14E-07

100

310,000

0.0507*

0.069411

86%

UNEMPLOYMENT

0.0318**

0.104357

2.5%

80%

DOCTORS

0.0073***

-0.000186

425

HIV/AIDS OF PEOPLE PUBLIC HEALTH

0.0356**

4.95E-08

500

5,100,000

0.4781

3.89E-06

$26

$6,096

SMOKING

0.3457

8.36E-06

3023

GDP PPP

0.2291

-3.27E-15

$6.186 Billion

$13.86 Trillion

Note: ***, **, * denote significance at 1%, 5%, and 10% respectively.

Doctors was the only variable that demonstrated a significance at the 1% level, and this variable shows how many doctors a country has per 100,000 people. This shows that the Doctors variable is extremely significant in its effect on infant mortality. Doctors also showed a negative coefficient which means that as the amount of doctors per 100,000 people increases, infant mortality in that country will decrease. The negative coefficient correlates to two other papers, where the amount of doctors increased and the infant mortality fell [Flegg (1982) and Kennelly et al (2002)]. From this, it is proven that the more doctors available for births will lessen the chance of an improper birth, and therefore lower infant mortality. HIV/AIDS Deaths shows the amount of people in each country who have died from having this virus. This factor was one of the factors used to expand the model and show how it 90

also has an impact infant mortality. HIV/AIDS Deaths showed significance at the 5% level which means this factor is moderately significant in its impact on infant mortality. HIV/AIDS Deaths also demonstrated a negative coefficient, similar to Doctors, which means that as more people die from HIV/AIDS there will be a decrease in infant mortality. This occurs because as more people die from HIV/AIDS there are less people that have the virus and less of a chance of passing it on to their new-bourns who will most likely die from the virus. This once again, will lower infant mortality. HIV/AIDS People is the variable that shows the amount of people in a country that have the HIV/AIDS virus. This factor also showed significance at the 5% level, but instead of a negative coefficient, this factor showed a positive coefficient. The coefficient for this factor showed a positive sign because the more people that contain the HIV/AIDS virus means that there are more people in that country to pass the virus onto their offspring. This in turn would kill these infants and increase the infant mortality rate in that country. Unemployment was the third factor that showed significance at the 5% level. Unemployment is the unemployment rate for each country, and this significance level shows that unemployment has a moderate effect on infant mortality overall. Unemployment was another factor that was added to this model to show its effect on infant mortality and build upon other models. Unemployment showed a positive coefficient in this model meaning that as unemployment increases, infant mortality increases. This happens because when people are unemployed they have less money to spend on medical treatment and for doctors which leads to a higher risk of infant mortality. Poverty, the last significant factor, shows the percentage of people in a given country that are living under the poverty line. Poverty has a significance level of 5% when it comes to infant mortality which means that is has a moderate influence on the number of infant deaths. Poverty, like unemployment, showed a positive coefficient which correlates to the paper “Poverty and Infant Mortality in the United States” by Gortmaker (1979) where poverty has a positive significance on infant mortality. This can be explained by the fact that if people are poor and cannot afford necessities like food, water, and medical care then their baby will not be born with proper nutrition and medical care because it is not affordable. This in turn will lead to higher infant mortality.

Life Expectancy In the Life Expectancy model there were six variables that were used, and five of them proved to be significant. Public Health, Poverty, Infant Mortality, HIV/AIDS People, and Unemployment all proved to be significant while Smoking proved to be insignificant. Of the five significant variables, Unemployment and Infant Mortality both had a significance level of 1%, Public Health and Poverty had a 5% significance level, and HIV/AIDS people had a 10% significance level. These five variables all demonstrated some sort of significance on life expectancy in countries around the world. The results of the life expectancy regression can be found in Appendix E. Variable

Results - Life Expectancy Observations Probability Coefficient

Min

Max

PUBLIC HEALTH

0.0204**

0.001629

$26

$6096

POVERTY

0.0192**

-14.91305

86%

INFANT MORTALITY HIV/AIDS OF PEOPLE SMOKING

0.0000***

-135.3439

100

310,000

0.0774*

-1.51E-06

500

5,100,000

0.1216

-0.002025

3023

UNEMPLOY

0.0006***

-25.03279

2.5%

80%

Note: ***, **, * denote significance at 1%, 5%, and 10% respectively.

The unemployment factor, as with infant mortality, also proved to be significant when determining life expectancy. Unemployment is once again the unemployment rate in each of the countries and it showed a 1% significance level when determining life expectancy. Unemployment proved to have a negative coefficient which implies that when unemployment rises, life expectancy falls. This occurs because when there is more unemployment, people do not have as much money to spend on necessities such as food, water, and shelter. Also people cannot afford to pay to go to the doctors and have routine checkups. This once again causes higher health risk and therefore will lower life expectancy. Infant mortality proved to have a huge influence on life expectancy. Infant mortality shows the number of infant deaths per 1,000 births. Infant Mortality had a significance factor at the 1% level which means that it is extremely significant when determining life expectancy. Infant mortality also proved to have a negative coefficient which means that the more infant 92

deaths there are, the lower the life expectancy age will be. This can be explained by the fact that as more infants are dying, their age of death is going to lower the overall life expectancy age and therefore affect it negatively. Public health shows the amount of money spent per capita for public health care. This proved to have a moderate influence on life expectancy as it was at the 5% significance level. Public health expenditure demonstrated a positive coefficient when it came to life expectancy which means that as public health expenditure increased so did life expectancy. This is congruent with another model [Kennelly et al (2002)], where we see that public health expenditure has a positive impact on life expectancy. This can be justified by the fact that more money spent on healthcare reduces health risks and chances of developing sicknesses. If this occurs then people will live for longer and increase the average life expectancy age. In terms of poverty, there is a moderate effect on life expectancy due to 5% significance level that is demonstrated. Poverty represents the amount of people that live under the poverty line in a given country for that year. Poverty proved to have a negative coefficient when it came to life expectancy which shows that people who live in poverty will die at a younger age. This can be elucidated because the people who live in poverty have less money to purchase necessities and pay for health care. This in turn shows that these people are at a greater risk of dying at a younger age then people who live above the poverty line. Without everyday necessities and good health you have a better chance of dying at a younger age, and henceforth the average life expectancy age is lowered. HIV/AIDS People tells the amount of people in each country that have the virus in that given year. The number of people with HIV/AIDS proved to have a minor influence on life expectancy being in the 10% significance interval. HIV/AIDS people showed a negative coefficient in terms of life expectancy meaning as the number of people with the virus increases, the average life expectancy age falls. This can also be seen in Eric Neumayer’s (2004) paper “HIV/AIDS and CrossNational Convergence in Life Expectancy,” where life expectancy is affected negatively when more people obtain the virus. This theory is vindicated by the fact that the more people that have the virus, have a better chance of dying at a younger age, which in turn lowers their life expectancy.

5.0 CONCLUSION Infant mortality and life expectancy are just statistics when it comes down to it, but they are important statistics. As we have seen from the results, there are many social and economic factors 93

that influence these two statistics, both negatively and positively. Awareness of these factors and their impact on these statistics are what change a person’s life. Hopefully with more effort and research, countries can see that infant mortality is a huge problem (especially in parts of the world like Africa) that can be fixed. Hopefully, countries can also see that life expectancy is not just a number, but a centerpiece for how much time people spend on this planet. With more dedication to the cure of HIV/AIDS, ending poverty, and lowering unemployment, countries can make a huge impact on how long people live for. Better allocation of resources to healthcare and health awareness around the world can help lead to lower infant mortality and a higher life expectancy. It is not realistic to believe these problems will ever fully be resolved, but with more effort, time, and better allocation of resources to these different factors, there can be an enormous impact made. The HIV/AIDS virus is a huge problem in Africa, and if more resources and time are devoted to this problem for testing and awareness, then the overall amount of people who have the virus could decrease greatly. This study should lead to better awareness of this subject and the breakdown of these factors should help increase the understanding of what can be done in countries around the world. Eventually with more time and research, these problems can receive devotion from countries to help people live a longer, healthier, more fulfilling life.

Appendix A: Variable Description and Data Source Economic/Social Factor

Description of Factor

Source

Doctors

Number of doctors for every 100,000 people in a country

http://earthtrends.wri.org/searchable_db/inde x.php?theme=4&variable_ID=1297&action =select_countries

GDP (PPP)

Overall GDP of a country Number of people in a country that have HIV or AIDS

https://www.cia.gov/library/publications/the -world-factbook/index.html http://hivinsite.ucsf.edu/global?page=cr06mo-00

HIV/AIDS (# of People)

https://www.cia.gov/library/publications/the -world-factbook/index.html HIV/AIDS (Deaths)

Number of people in a country who dies from AIDS in the past year

http://hivinsite.ucsf.edu/global?page=cr06mo-00 https://www.cia.gov/library/publications/the -world-factbook/index.html

Infant Mortality Life Expectancy

Poverty

Public Health

Smoking

Number of infant deaths per 1,000 births Average age a person lives to in a given country Percent of the population that lives under the poverty line Public health expenditure as a percent of GDP Number of cigarettes consumed per capita in a given year

https://www.cia.gov/library/publications/the -world-factbook/index.html https://www.cia.gov/library/publications/the -world-factbook/index.html https://www.cia.gov/library/publications/the -world-factbook/index.html http://www.infoplease.com/ipa/A0934556.ht ml http://www.nationmaster.com/graph/hea_tob _cig_con-health-tobacco-cigaretteconsumption http://www1.worldbank.org/tobacco/databas e.asp

Unemployment

Unemployment rate

https://www.cia.gov/library/publications/the -world-factbook/index.html

Appendix B: Variables and Expected Signs – Life Expectancy Economic/Social Factor HIV/AIDS (# of People)

Infant Mortality

Poverty

Public Health

Smoking

Unemployment

Description of Factor Number of people in a country that have HIV or AIDS.

Number of infant deaths per 1,000 births. Percent of the population that lives under the poverty line.

Public health expenditure as a percent of GDP. Number of cigarettes consumed per capita in a given year.

Expected Sign

Rationale

More people with HIV/AIDS means that more people will die at a younger age, therefore lowering life expectancy.

Unemployment rate. _

More infant deaths lead to a decrease in the average age length, thus lower life expectancy. Higher poverty rates mean more people that cannot afford necessities to live which results in a lower life expectancy. Higher health expenditure leads to better health services and a higher life expectancy. High cigarette consumption leads to higher health risk and a lower life expectancy. Higher unemployment means less money per family for necessities and therefore a lower life expectancy.

Appendix C: Variables and Expected Signs – Infant Mortality Economic/Social Factor Doctors

Description of Factor Number of doctors per 100,000 people in each country.

Expected Sign

Rationale More doctors for people to go to when they have children means better individual medical attention and lower infant mortality.

GDP

Gross Domestic Product relative to PPP in each country.

HIV/AIDS (# of Deaths)

HIV/AIDS (# of People)

Poverty

Number of deaths in a given country from the HIV and AIDS viruses.

Number of people in a country that have HIV or AIDS.

Percent of the population that lives under the poverty line.

Public Health

Public health expenditure as a percent of GDP.

Smoking

Number of cigarettes consumed per capita in a given year.

Unemployment

Unemployment rate. +

Higher GDP means more money to spend on general necessities and medical expenses which would lead to less infant mortality. The more people that die from the HIV/AIDS viruses means that there are less people to pass on the viruses to their children and less infant mortality. More people with HIV/AIDS mean that there are more people to have kids and pass the viruses on to them which would lead to higher infant mortality. Higher poverty rates mean more people that cannot afford necessities to live or medical expenses which results in a higher infant mortality rate. Higher health expenditure would lead to better healthcare and lower infant mortality. High cigarette consumption leads to higher health risk and a higher infant mortality rate. Higher unemployment means less money per family for medical care and therefore an increase in infant mortality.

Appendix D: Results – Infant Mortality Variable

Observations

Probability

Coefficient

Min

Max

HIV AIDS DEATHS

0.0239**

-8.14E-07

100

310,000

POVERTY

0.0507*

0.069411

86%

UNEMPLOYMENT

0.0318**

0.104357

2.5%

80%

DOCTORS

0.0073***

-0.000186

425

HIV/AIDS OF PEOPLE

0.0356**

4.95E-08

500

5,100,000

PUBLIC HEALTH

0.4781

3.89E-06

$26

$6,096

SMOKING

0.3457

8.36E-06

3023

GDP PPP

0.2291

-3.27E-15

$6.186 Billion

$13.86 Trillion

Note: ***, **, * denote significance at 1%, 5%, and 10% respectively.

Appendix E: Results - Life Expectancy Variable

Observations

Probability

Coefficient

Min

Max

PUBLIC HEALTH

0.0204**

0.001629

$26

$6096

POVERTY

0.0192**

-14.91305

86%

INFANT MORTALITY

0.0000***

-135.3439

100

310,000

HIV/AIDS OF PEOPLE

0.0774*

-1.51E-06

500

5,100,000

SMOKING

0.1216

-0.002025

3023

UNEMPLOY

0.0006***

-25.03279

2.5%

80%

Note: ***, **, * denote significance at 1%, 5%, and 10% respectively.

Appendix F: Countries Afghanistan Algeria Angola Argentina Australia Bolivia Brazil Canada Chad China Denmark Egypt France Germany Guatemala India Iran Iraq Ireland Ivory Coast Japan Kenya Malaysia Mexico Mongolia Morocco Nicaragua 99

Niger Russia Saudi Arabia Spain Sudan U.K. U.S.A. Ukraine Vietnam Yemen Zambia Zimbabwe BIBLIOGRAPHY

Flegg, A.T., (1982), “Inequality of Income, Illiteracy and Medical Care as Determinants of Infant Mortality in Underdeveloped Countries”, Population Studies, 36 (3): 441-458. Available at <http://www.jstor.org/stable/2174055?seq=1>

Gortmaker, Steven L., (1979), “Poverty and Infant Mortality in the United States”, American Sociological Review, 44 (2): 280-197. Available at <http://www.jstor.org/stable/2094510?seq=1>

Houweling, T., Kunst, A., and Mackenbach, J., (2001), “World Health Report 2000: inequality index and socioeconomic inequalities in mortality”, The Lancet, 357: 1633. Available at <http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6T1B439MNHK33&_user=545676&_coverDate=05%2F26%2F2001&_rdoc=1&_fmt=high&_orig=brow se&_cdi=4886&_sort=d&_docanchor=&view=c&_ct=1&_refLink=Y&_acct=C0000279 38&_version=1&_urlVersion=0&_userid=545676&md5=5db92e0a77d8850b75700ee06 04a80c7>

100

Kennelly, B., O’Shea E., and Garvey E., (2002), “Social capital, life expectancy and mortality: a cross-national examination”, Social Science and Medicine, 56: 2367-2377. Available at <http://www.sciencedirect.com/science?_ob=ArticleURL&_udi=B6VBF47DM64V1&_user=657938&_rdoc=1&_fmt=&_orig=search&_sort=d&view=c&_acct=C0000356 79&_version=1&_urlVersion=0&_userid=657938&md5=c08b9d2ae49788547f3672e112 3ea188#appa>

Neumayer, Eric, (2004), “HIV/AIDS and Cross-National Convergence in Life Expectancy”, Population and Development Review, 30 (4): 727-742. Available at <http://www.jstor.org/stable/3657336?seq=1>

The World Fact Book, [online data file], https://www.cia.gov/library/publications/theworld-factbook/

101

The Interaction between the Stock Market, Monetary Policy and Inflation in Singapore and Malaysia

Lindsey Kahler a

Abstract: This paper investigates and compares the interactions among the stock market, monetary policy, and inflation in both Singapore and Malaysia from 2005 through 2007 using bivariate and multivariate vector autoregressive cointegrating specifications. The Granger-causality test shows that for Malaysia there significant unidirectional relationships of inter-bank loan rates to inflation, and inflation to Kuala Lumpur stock returns. For Singapore there is only one significant marginal unidirectional Granger-causality relationship of inflation to Straits Times stock returns. There are no reciprocal relationships in either country. Based on changes in the stock market, the multivariate results show negative changes on the interbank interest rates in Malaysia and positive changes on inflation except for during the third lag, where changes are negative. In Singapore, multivariate results show that changes in the stock market lead to negative changes in interbank interest rates, and negative changes in inflation for two lags before the changes turn positive during the third lag. Changes in interbank interest rates and inflation have no significant effect on any of the three variables.

JEL Classifications: E52, E44, G10 Keywords: stock market returns, monetary policy, inflation, Granger-causality, VAR, impulse response function, Malaysia, Singapore

Department of Economics, Bryant University, 1150 Douglas Pike, Smithfield, RI 02917. Phone: (774) 259-3157. Email: lkahler@bryant.edu.

102

1.0 INTRODUCTION The performance of the stock market corresponds with a perception of the overall state of an economy. Equity prices are watched closely because they are considered to be very sensitive to economic conditions. One of the most important challenges among economists is to understand how monetary policy affects the stock market and other economic issues. Monetary policy is looked at as one of the main tools for stabilizing the economy. In the United States the Federal Reserve has the ability to set the discount rate in order to influence the country’s money supply. Monetary authorities in other countries have different levels of control concerning economywide interest rates. A country’s rate of inflation is also expected to have an effect upon the performance of the stock market. Theories have been developed which look to explain how monetary policy and inflation interact with the stock market. The traditional view is that expansionary monetary policy will increase the demand for assets and will stimulate the economy. Expansionary monetary policy is used to increase the size of the money supply in a nation, which is achieved indirectly by decreasing interest rates. When there is a favorable interest rate in a country, asset prices will increase, leading inflation to increase. Once this occurs the central bank must contain the inflationary pressure through monetary policy. This study aims to enhance the understanding about how changes in interbank interest rates and inflation affect the stock market of Malaysia and Singapore, and the differences between the two countries. It will look at whether any significant interdependencies exist among the three variables in each country. Malaysia is considered to be a newly industrialized country which has recently been experiencing growth rates of 5 to 7% a year. At various times the KLSE was the most active exchange in the world, with trading volume exceeding even the NYSE. During the Asian Financial Crisis, however, the Kuala Lumpur Stock Exchange fell from 1300 to only 400 points in just a few weeks. Malaysia refused help from the IMF, which resulted in the country not being affected nearly as much as others in the area. In 1998 the country’s GDP contracted 7.5%, but rebounded again in 1999. Malaysia began a plan of increased government spending in order to stimulate the economy. For years after the crisis they recorded budget deficits which were led by increased exports. Pressure of inflation has remained low since, and because of this the Bank Negara Malaysia has been able to implement low interest rates. 103

Singapore’s economy is highly developed capital mixed economy which exhibits open business practices, stable prices, and one of the highest per capita GDPs in the world. Between 1960 and 1999 Singapore experienced average real growth of 8%. Different factors, including the Asian Financial Crisis, the worldwide electronics slump, and the SARS outbreak had in the past led to reductions in economic growth, however the country continuously managed to make significant recoveries following each crisis. Singapore has also been able to keep inflationary pressures low, allowing the central bank to be able to implement low interest rates The decision to test for the interactions between the stock market, monetary policy, and inflation in these countries is because they are two quickly emerging and developing Asian economies which have both experienced significant growth in wake of the financial crisis. During the years that this paper examines, these two countries have both experienced their highest growth rates. The goal of this research is to see if monetary policy in Malaysia has had different effects on Malaysia’s stock market than monetary policy implemented in Singapore has had there. This paper examines whether or not changes in the stock markets of Malaysia and Singapore are caused by interest rates and inflation or vice versa, whether the changes are positive or negative, how long these changes remain significant, and how long after the change it takes the variable to return to normal. A multivariate cointegration test is used to determine the number of cointegrating vectors, which is then used to determine the general VEC model. After running a unit root test, the Granger-causality test is run to examine the relationships between variables in each country. Before running a multivariate test, bivariate VAR/VEC estimates are first attained between stock returns and inflation, and stock returns and interbank rates. These results examine how inflation and interbank rates interact upon the stock market independent of each other. Impulse response graphs are used for both bivariate and multivariate tests to determine how long shocks in the economy have an effect upon the different variables. The paper is organized as follows: Section 2 is the literature review which examines the results of past studies dealing with the interactions of the same variables; Section 3 discusses data and empirical methodology which includes construction of the data, the multivariate cointegration test, definition of the variables, the unit root test, the Granger-causality test, and bivariate VAR/VEC estimates; Section 4 presents and discusses the empirical results of the tests,

104

consisting of the multivariate VAR/VEC estimates; Section 5 offers a conclusion, and is followed by Appendix A. 2.0 LITERATURE REVIEW Studies have been conducted examining the interactions between the stock market, interbank rates, and inflation however none have tested the linkages of these three variables for Malaysia or Singapore. In regards to studies conducted examining the interactions in the United States, Lee (1992) has found using the VAR/VEC estimate that stock returns explain little variation in inflation while interest rates explain a large portion of the variation in inflation. Titman and Warga (1989), however, did find a statistically significant positive relation between stock returns and interest rate changes. Laopodis (2006) reports finding no relationship between monetary policy and stock prices. Negative relationships between inflation and stock market returns have been found by several studies which include Fama (1981), Geske and Roll (1983), and Kaul (1987). Cozier and Rahman (1988) find that in both the United States and Canada there is an inverse relationship between stock returns and inflation, however the Granger-causality test shows that inflation does not cause stock returns. These studies have shown that there is no conclusive relationship between the stock market, monetary policy, and inflation. The aim of this research is to develop a model to explain the impact changes in interbank interest rates and inflation have on the stock markets of Malaysia and Singapore, and to compare the findings between the two countries. 3.0 DATA AND EMPIRICAL METHODOLOGY 3.1 Data Construction and Preliminary Statistical Investigation Monthly data for 2005 to 2007 on the returns of the Kuala Lumpur and Straits Times stock exchanges comes from Econstats.com. Interbank overnight interest rates for Malaysia and Singapore were retrieved from each country’s central bank website (www.bnm.gov.my and www.mas.gov.sg, respectively). Inflation rates based on the CPI come from the IMF database for the close of 2004. Monthly changes were calculated from data collected from the Reuters database showing the monthly percentage change from the previous month for January 2005 through 2007. Table 1 summarizes all variable acronyms, their descriptions, and their sources.

105

Table 1: Variable Description and Data Source Acronym

Description

Data Source

KLSM

Close for the Kuala Lumpur stock exchange on the last day of each month

Econstats.com

STSM

Close for the Straits Times stock exchange on the last day of each month

Econstats.com

KLIB

Singapore’s interbank overnight interest rates on the last day of each month

Malaysia Central Bank Website www.bnm.gov.my

STIB

Malaysia’s interbank overnight interest rates on the last day of each month

Singapore Central Bank Website www.mas.gov.sg

KLINF

Inflation based on Singapore’s CPI; base = 100 in 2002

Reuters Database

STINF

Inflation based on Malaysia’s CPI; base = 100 in 2002

Reuters Database

Table 2 reports descriptive statistics for the sample period for both Malaysia and Singapore. Table 2: Descriptive statistics of monthly stock returns, interest rates, and inflation in Malaysia and Singapore, 2005-2007 Variable

Mean

Skewness

Kurtosis

Min

Max

KLSM KLIB KLINF

1063.666 3.21111 111.0955

198.105 .368338 2.742826

.725946 -0.593078 -0.378262

1.851828 1.479094 1.894236

860.73 2.67 106.27

1445.03 3.56 115.4149

STSM STIB STINF

2746.449 2.348056 103.7077

547.6147 .762491 1.776262

.539957 -0.425137 1.282517

1.787898 2.369497 4.209011

2096.32 .5 101.1841

3805.7 3.5 108.6326

3.1.1 Cointegration Tests and Results Table 3: Multivariate Cointegration Tests Hypothesized number of CE(s) Malaysia None* At most one At most two Singapore None*

Trace statistics

5% Critical Value

Max-Eigenvalue statistic

5% Critical Value

32.56104* 9.080347 0.605318

24.27596 12.32090 4.129906

23.47169* 8.484029 0.605318

17.79730 11.22480 4.129906

29.53352*

24.27596

10.70479

17.79730

106

At most one 9.828724 12.32090 6.626326 11.22480 At most two 3.202398 4.129906 3.202398 4.129906 *Statistical significance at the 0.05 level CE(s) indicate cointegrating equations; ‘none’ means there is one cointegrating relationship binding all variables

Multivariate cointegration was checked for using the Johansen Cointegration Test, which produces statistics for the number of cointegrating vectors. The lag interval as determined by a unit root test was in second differences. Table two presents the results of this procedure. It is reported that by looking at both the trace statistic and max-eigenvalue statistic in both Malaysia and Singapore, that one statistically significant trend exists that binds stock market returns, interbank interest rates, and inflation. For Malaysia this trend was significant under both the trace statistic and max-eigenvalue statistic, while for Singapore it was significant only under the trace statistic. 3.1.2 Definition of Variables With the observed existence of one cointegrating relationship among the variables in Xt, the casual relationship among the three variables can be determined by estimating the following general VEC model: ∆𝑋𝑡 = ∝ + 𝛾𝛽𝑋𝑡−1 + ∑𝑘𝑗=1 𝜏𝑗 ∆𝑋𝑡−𝑗 + 𝜀𝑡

(1)

where α is a constant vector representing a linear trend, and γ and β represent the speed of adjustment and the cointegration vector, respectively. The vector Xt consists of stock returns, KLSRt and STSRt, overnight interbank interest rates, KLIBt and STSRt, and inflation, KLINFt and STINFt. The three cointegrated variables have the following joint VEC integration under a single cointegrating relationship: n2 n3 ∆𝐼𝑅𝑖,𝑡 = ∝1 + γ1 ε𝑡−1 + ∑n1 𝑖=1 β1,𝑖 ∆𝐼𝑅𝑖,𝑡 + ∑𝑖=1 β2,𝑖 ∆𝑆𝑀𝑗,𝑡 + ∑𝑖=1 β3,𝑖 ∆𝐼𝑁𝐹i,t + e1,𝑡

(2)

m2 m3 ∆𝑆𝑀𝑗,𝑡 = 𝛼2 + 𝛾2 𝜀t−1 + ∑m1 𝑖=1 𝛿1,𝑖 ∆ 𝐼𝑅𝑖,𝑡 + ∑𝑖=1 𝛿2,𝑖 ∆𝑆𝑀𝑗,𝑡 + ∑𝑖=1 𝛿3,𝑖 ∆𝐼𝑁𝐹𝑖,𝑡 + e2,𝑡 l2 l3 ∆𝐼𝑁𝐹𝑖,𝑡 = ∝3 + γ3 ε𝑡−1 + ∑l1 𝑖=1 ∅1,𝑖 ∆𝐼𝑅𝑖,𝑡 + ∑𝑖=1 ∅2,𝑖 ∆𝑆𝑀𝑗,𝑡 + ∑𝑖=1 ∅3,𝑖 ∆𝐼𝑁𝐹i,t + e3,𝑡

(3) (4)

where ∆ is the first difference operator, and e1, e2, and e3 are stationary random error terms. The n’s m’s and l’s are the orders of the autoregressive process for a given variable. The ε𝑡−1 magnitudes are the EC

terms from the cointegrating equations so that changes in the variables are partly based on past values of ε𝑡 . Equations (2) – (4) are used to evaluate the short- and long-run interactions among the three 107

variables. The short-run interactions between two variables are exhibited by the β1,i, 𝛿2,𝑖 , or ∅2,𝑖 terms. If one or more of these coefficients is nonzero and significant, then there is a short-run effect on the variable. The existence of a long-run relationship between variables depends on the significance of the γ1 and γ2 coefficients.

3.2 Data 3.2.1 Unit Root Tests The ADF Fisher Unit Root test uses an autoregressive model to test whether a time series is non-stationary by determining whether a unit root is present. In the case of this test, for both countries the second difference must be taken before the data is considered to be stationary. Stationary data must be achieved before the Granger-causality test can be run and prior to determining VAR/VEC estimates. Table 4: ADF Fisher Unit Root Test Method ADF – Fisher Chi-Square ADF – Choi Z-stat

Malaysia Statistic 87.9737 -8.4082

Probability 0.0000 0.0000

Intermediate ADF Test Results: Second Difference Series Probability KLIB 0.0001 KLINF 0.0000 KLSM 0.0000

Singapore Method ADF – Fisher Chi-Square ADF – Choi Z-stat

Statistic 89.4968 -8.51665

Probability 0.0000 0.0000

Intermediate ADF Test Results: Second Difference Series Probability STIB 0.0000 STINF 0.0000 STSM 0.0001

3.2.2 Granger-Causality Test Results Before estimating the multivariate VAR/VEC models among the three variables, the bivariate estimates for pairs of variables are first examined. The Granger-causality estimates appear in table four. 108

Table 5: Granger-Causality Test Arrows show the direction of Granger causality tested; t-statistics are in parentheses Country Malaysia

Granger Causality Test KLINF→ KLIB F-stat = 0.54161 KLIB → KLINF F-stat = 3.13569** KLSM → KLIB F-stat = 1.18253 KLIB → KLSM F-stat = 1.88898 KLSM → KLINF F-stat = 0.79871 KLINF → KLSM F-stat = 2.60849** Singapore STINF → STIB F-stat = 1.15285 STIB → STINF F-stat = 0.91186 STSM → STIB F-stat = 1.32596 STIB → STSM F-stat = 1.14381 STSM → STINF F-stat = 2.36701 STINF → STSM F-stat = 3.61680** **Statistical significance at the 0.01 level

(0.8032) (0.0414) (0.3880) (0.1622) (0.6163) (0.0714) (0.3298) (0.4130) (0.2812) (0.3326) (0.1116) (0.0396)

The optimal lag length for Malaysia is two months, and for Singapore it is three months. For Malaysia the significant unidirectional relationships are KLIB to KLINF, and KLINF to KLSM. For Singapore there is only one significant marginal unidirectional Granger-causality relationship of STINF to STSM. There are no reciprocal relationships in either country. 3.2.3 Bivariate VAR/VEC Estimates Table 6: Bivariate VAR/VEC Estimates Estimates of two-equation systems based on Equations (1) through (4). t-statistics are in parentheses Panel A: Stock returns and inflation Malaysia Singapore ∆KLSMt ∆KLINFt ∆STSMt ∆STINFt Constant 44.16361 0.321324 58.91450 .504142 (2.12439) (1.84593) (1.26839) (2.06045) ∆SMt-1 -0.125850 0.000492* 0.073909 -0.001749* (-.60122) (0.28074) (0.27895) (-1.25278) ∆SMt-2 -0.006719 0.001652* -0.50349 -0.000883* (-0.03296) (0.96785) (-0.19102) (-0.63612) ∆SMt-3 0.173933 -0.002095* -0.295712 -0.00815* (0.86548) (-1.24521) (-1.10707) (-0.57945) ∆SMt-4 -0.259187 -0.002038* 0.259939 -0.002077* (-1.20404) (-1.13053) (0.94383) (-1.43151) ∆INFt-1 -26.74194 -0.219347 -55.08434 -0.241093 (-1.10332) (-1.08080) (-1.43700) (-1.19397) ∆INFt-2 -20.36029 -0.015764 65.33357 -0.180160 (-0.82461) (-0.07625) (1.52849) (-0.80014) ∆INFt-3 -27.66862 -0.020617 -42.39851 0.339398 (-1.14932) (-0.10228) (-0.97115) (1.47535) ∆INFt-4 -11.55920 0.080268 -52.99225 -0.031884 (-0.48646) (0.40343) (-1.22999) (-0.14049) Adjusted R2 -0.092601 -0.82997 0.146380 0.167805 *Statistical significance at the .05 level

Panels A and B of Table 5 display the bivariate VAR/VEC estimates between the stock market paired with interbank interest rates and inflation.

109

The results of the bivariate VAR/VEC test shows that for four lags the changes in the Kuala Lumpur stock returns predict changes in Malaysia’s inflation. The changes start out small after the first lag (which has a coefficient of 0.000492) and then becomes stronger during the second lag. However, by the third and fourth lag the changes become negative, with a stronger coefficient during the third lag (-0.0002095) than the fourth lag (-0.002038). The estimates for stock returns do not reveal an impact of either variable’s past changes on the stock returns, nor does past changes in inflation have any impact on present inflation. The bivariate estimates for Singapore show that for four lags the changes in the Straits Times stock returns predict changes in the country’s inflation. Unlike the results for Malaysia, however, the changes begin negative (-0.001749) during the first lag and then become less strongly negative through the second and third lags before turning more strongly negative again during the fourth lag. Similarly to the results for Malaysia, the estimates for stock returns do not reveal an impact of either variable’s past changes on stock returns, nor do past changes in inflation have any impact on present inflation.

Constant ∆SMt-1 ∆SMt-2 ∆SMt-3 ∆SMt-4 ∆IBt-1 ∆IBt-2 ∆IBt-3 ∆IBt-4 Adjusted R2

Panel B: Stock returns and interbank rates Malaysia Singapore ∆KLSMt ∆KLIBt ∆STSMt ∆STIBt 34.26658 0.017819 41.88649 0.156971 (2.92015) (.79511) (1.05455) (0.82192) -0.17560 -0.000106* -0.52613 -0.000300* (-0.86472) (-0.28104) (-0.23343) (-0.27654) -0.061249 -0.000120* -.280344 -0.002050* (-0.30916) (-0.31647) (-1.03752) (-1.57822) 0.124295 -4.90E-05* 0.044281 -0.001031* (0.64025) (-0.13218) (0.15250) (-0.73815) -0.278187 -0.000286* 0.306559 0.000278* (-1.38381) (-0.74449) (1.06320) (0.20035) -37.10213 -0.144267 -20.34380 -0.461835 (-0.34737) (-0.70722) (-0.43845) (-2.07011) -99.52809 0.512994 36.24685 -0.496089 (-0.94479) (2.54976) (0.71227) (-2.02745) -165.7910 0.144945 20.12884 -0.377604) (-1.41014) (0.64551) (0.40008) (-1.56092) -86.03998 0.124406 31.51636 -0.272138 (-0.73900) (0.55948) (0.72058) (-1.29405) -0.013118 0.201362 -0.216179 0.240118

*Statistical significance at the .05 level

The results of the bivariate VAR/VEC test for Malaysia show that for four lags the mutual, short-run relationships between changes in the stock market affecting interbank rates are statistically significant and negative for Malaysia. From the first lag to the second lag the changes become more strongly negative (-0.000106 to -0.000120), the third lag becomes less

110

strongly negative (-4.90E-5), and during the fourth lag the changes become the most strongly negative (-0.000286). However, the estimates for changes in stock returns do not have a significant impact on interbank interest rates or on the Kuala Lumpur’s stock returns. Past changes in interbank rates also have no impact on present rates. The bivariate estimates between the Straits Times stock return and interbank interest rates in Singapore show negative, statistically significant results for four lags for the changes in the stock market predicting changes in the interbank rates. The changes become more strongly negative from the first to the second lag (-0.000300 to -0.002050), and then become gradually less strongly negative during the third and fourth lags. The test shows no statistically significant impact of stock returns on interbank rates or on the stock market itself, and also there is no impact of changes in interbank rates upon present rates.

111

3.2.4 Impulse Response Graphs Figure 1: Impulse Response Graphs Panel A: Malaysia

Response of KLSM to Cholesky One S.D. Innovations

50 40

40 30

10 10

0 0

-10 -20 1

KLSM

-10 1

KLINF

KLSM

Response of KLINF to Cholesky One S.D. Innovations

KLIB

Response of KLIB to Cholesky One S.D. Innovations

.12 .10

.08

.06

.1 .04

.02

-.1

.00

5 KLSM

7 KLINF

5 KLSM

KLIB

The impulse response graphs show the response of the Kuala Lumpur stock returns to a one standard deviation shock in inflation. The result is positive, and although after ten months it has still not died out, it slowly has less of a response over time. The response of the Kuala Lumpur stock returns to a shock in interbank interest rates is also positive and slowly becomes less positive over the ten month period. However, this response also has not died out by the end of the observed period.

112

Panel B: Singapore Response of STSM to Cholesky One S.D. Innovations

Response of STSM to Cholesky One S.D. Innovations

160

150

120 100

80 50

-40

-50 1

STSM

STINF

STIB

Response of STIB to Cholesky One S.D. Innovations

Response of STINF to Cholesky One S.D. Innovations .8

.6 .4

.4 .2

.2 .0

-.2

-.2 -.4

-.4 1

5 STSM

STINF

5 STSM

STIB

The response of the Straits Times stock index to a one standard deviation shock in inflation is strongly positive. The impact varies over an eight month time, but continues to stay positive. After the eighth month the response then dies out. The response of the stock index to a one standard deviation shock in the interbank interest rates is also initially strongly positive. Over the ten month period the impact varies, however by the end of the period the response still has not died out. 4.0 EMPIRICAL RESULTS 4.1 Multivariate VAR/VEC Estimates Table 6 presents the multivariate estimates from the VEC model for both Malaysia and Singapore. Kuala Lumpur returns in Malaysia for the first three lags return negative impacts on the country’s interbank loan rates. The changes grow more strongly negative from the first to second lag, and then again become less negative during the third lag. The stock market returns positively affect inflation rates in Malaysia for the first and second lags, but during the third lag the impact becomes negative. Kuala Lumpur returns have no impact upon themselves at any lag. 113

Also, changes in neither interbank interest rates nor inflation have any statistically significant impact on the stock returns, interbank interest rates, or inflation. Table 7: Multivariate VAR/VEC Estimates

Constant ∆SMt-1 ∆SMt-2 ∆SMt-3 ∆IBt-1 ∆IBt-2 ∆IBt-3 ∆INFt-1 ∆INFt-2 ∆INFt-3 Adjusted R2

∆KLSMt 33.67378 (1.98304) -0.167981 (-0.81927) -0.006814 (-0.03526) 0.171475 (0.85050) -13.94252 (-0.11780) -77.43331 (-0.62705) -130.7026 (-1.13288) -13.93420 (-0.52836) -4.903865 (-0.17997) -19.31979 (-0.68075) -0.094489

Malaysia ∆KLIBt 0.017970 (0.58555) -0.000157* (-0.42416) -0.000256* (-0.73163) -0.000187* (-0.51348) -0.079759 (-0.37288) 0.526864 (2.36077) 0.065006 (0.31177) 0.065008 (1.36393) -0.014708 (-0.29866) -0.040840 (-0.79626) 0.461450

∆KLINFt 0.276943 (2.74989) 0.000762* (0.62651) 0.001227* (1.07072) -0.001975* (-1.65142) 1.247146 (1.77671) -0.247367 (-0.33776) 2.496058 (3.64788) -0.045051 (-0.28803) -0.427785 (-2.64710) 0.100230 (0.59548) 0.658274

∆STSMt 74.50307 (1.87644) 0.113677 (0.46059) -0.150496 (-0.56271) -.351034 (-1.40660) -29.87748 (-0.90639) 38.41194 (1.03000) -36.69107 (-0.90046) -86.05422 (-2.40554) 65.76075 (1.57235) -39.83637 (-0.93500) 0.188638

Singapore ∆STIBt 0.289865 (1.10219) -0.001401* (-0.85696) -0.002941* (-1.66034) -0.000471* (-0.28490) -0.608116 (-2.78523) -0.529476 (-2.14347) -0.236640 (-0.95176) -0.053618 (-0.22628) -0.224283 (-0.80962) -0.046439 (-0.16456) 0.040417

∆STINFt 0.286607 (1.32940) -0.001522* (-1.13591) -0.00308* (-0.21197) 0.000262* (0.19348) 0.176650 (0.98695) 0.003761 (0.01857) 0.254190 (1.24712) -0.229921 (-1.18366) -0.198700 (-0.87496) 0.372451 (1.60995) 0.168883

*Statistical significance at the .05 level

Results for Singapore show that Straits Times returns negatively affect interbank loan rates for three lags. The changes grow more strongly negative from the first to third lag. Also, returns on the stock market have a statistically significant impact on inflation rates. During the first lag the impact is strongly negative, and becomes less strongly negative during the second lag. However, during the third lag the impact is positive. These estimates also show that changes in Singapore’s interbank loan rates and inflation both have no impact on Straits Times returns, interest rates, or inflation. Straits Times returns have no impact upon themselves at any lag. 4.2 Impulse Response Graphs In response to a one standard deviation shock in both interbank rates and inflation, the Kuala Lumpur stock index has a positive reaction initially of about 40 points. Over an eight month period the reaction becomes less positive, and dies out around the ninth month. The response to the same situation on the Straits Times index is an initial positive response of about

114

100 points. Over the observed ten months period the response varies, always staying positive, and at the end of the period the response still has not died out. Response of STSM to Cholesky One S.D. Innovations

Response of KLSM to Cholesky One S.D. Innovations 150

100

50 20

0 0

-50 -100

-20 1

KLSM

KLIB

STSM

KLINF

STIB

STINF

Response of STIB to Cholesky One S.D. Innovations

Response of KLIB to Cholesky One S.D. Innovations .8

.16

.12

.4 .08

.2 .04

.0 .00

-.2

-.04

-.4 1

KLSM

KLIB

STSM

KLINF

Response of KLINF to Cholesky One S.D. Innovations

STIB

STINF

Response of STINF to Cholesky One S.D. Innovations

.25

.20

.4 .15 .10

.05

.0 .00 -.05

-.2 1

KLSM

6 KLIB

KLINF

STSM

6 STIB

STINF

5.0 CONCLUSION This paper examines the interactions among the stock market, monetary policy, and inflation in Malaysia and Singapore. The analysis uses bivariate and multivariate VAR/VEC models to test for statistical significant changes among variables. The bivariate results for the Kuala Lumpur stock index returns and inflation show initial positive changes which become negative during the third lag, and between the Straits Times stock index returns and inflation the changes are negative through four lags. In Malaysia there is a negative change on interest rates based on changes in the Kuala Lumpur stock index. However, in Singapore there is a negative change in interest rates based on changes in the Straits Times stock index through three lags, and then a positive change during the fourth lag. Based on changes in the stock market, the multivariate results show negative changes on the interbank interest rates in Malaysia and positive changes on inflation except for during the third lag where changes are negative. In 115

Singapore, multivariate results show that changes in the stock market lead to negative changes in interbank interest rates, and negative changes in inflation for two lags before the changes turn positive during the third lag. Changes in interbank interest rates and inflation have no significant effect on any of the three variables. It is not possible to say that there is a consistent relationship between monetary policy and the stock markets in Malaysia and Singapore since changes in variables have different effects in one country than in the other. According to the Granger-causality test, there are two significant unidirectional relationships (KLIB to KLINF, and KLINF to KSM) in Malaysia, while there is only one (STINF to STSM) in Singapore. References Cozier, B.V. and Rahman, A.H., 1988. Stock Returns, Inflation, and Real Activity in Canada. The Canadian Journal of Economics 21(4), 759 – 774. Fama, E.F., 1981. Stock Returns, Real Activity, Inflation and Money, American Economic Review 71(4), 545 – 565. Geske, R. and Roll, R, 1983. The Fiscal and Monetary Linkage between Stock Returns and Inflation, Journal of Finance 38(1), 1 – 33. Kaul, G., 1987. Stock Returns and Inflation: The Role of Monetary Sector, Journal of Financial Economics 18, 253 – 176. Laopodis, N.T, 2006. Dynamic Interactions among the Stock Market, Federal Funds Rate, Inflation, and Economic Activity, The Financial Review 41, 513 – 545. Lee, B.S., 1992. Casual Relations among Stock Returns, Interest Rates, Real Activity, and Inflation, Journal of Finance 47(4), 1591 – 1603. Titman, S. and Warga, A, 1989. Stock Returns as Predictors of Interest Rates and Inflation, The Journal of Financial and Quantitative Analysis 24(1), 47 – 58.

116

The Impact of Sectoral Performance on the Stock Market: Does Volatility Equal Explanatory Power?

Jeffrey Haydocka

Abstract: This paper investigates the real impact of sector performance on the overall stock market and the possibility that the primary culprit of negative or positive performance isn’t necessarily the one that significantly drives overall market performance. The study compares average gain or loss by sector with regression analysis to determine if the largest gainer or loser is also reflected in regression analysis after beta calculation. The results show that there is very limited correlation between the two measures suggesting that major volatility in one sector doesn’t necessarily have the greatest impact on total market movement.

JEL Classification: E44, G11, G14 Keywords: Sector Performance, Market Movement, Volatility. a

Economics Student, Bryant University, 1150 Douglas Pike, Smithfield, RI 02917. Phone: (401) 719-7710. Email: jhaydock@bryant.edu.

_____________________________ The author gratefully acknowledges the help/guidance from Dr. Ramesh Mohan.

117

1.0 INTRODUCTION Individual sector performance on the overall stock market has not really been analyzed. One could argue that this is because logically, the market reacts to news, news which affects any individual sector, so the relationship should seem quite simple. It is this relationship that I believe requires further attention. Is logical thought really accurate in this situation? Does Volatility have Explanatory Power? This is the question that will be addressed. This study aims to enhance the understanding of the intricate relationship between sector volatility and actual explanatory power in overall market performance. From a policy perspective, this analysis is important because if the results show that there is a disconnect in the two measures, then it could have implications for the accuracy of corrective measures in the market and even monetary policy. The relevance of this study is that it impacts the investing strategies of countless investors, as well as the impact the results could have on policy makers. If there is no disconnect between volatility and explanatory power, then logical thought is correct and we haven’t really learned much new information. On the other hand, if that disconnect does in fact exist, then this information immediately becomes valuable. This paper was guided by two research objectives that differ from other studies: First it investigates the possibility of a disconnect between sector volatility and explanatory power on overall market performance. Second, it looks at which sectors tend to have, on average, the highest levels of volatility and explanatory power as calculated through regression analysis. There tends to be an absence of research on this specific topic. This paper successfully fills that void. The rest of the paper is organized as follows: Section 2 analyzes stock market trends over the last eight years. Section 3 provides a brief literature review on other research papers written on related

118

subjects. Section 4 outlines the empirical model, data and estimation methodology. Finally, section 5 presents and discusses the empirical results. This is followed by a conclusion in section 6. 2.0 STOCK MARKET TRENDS While the scope of this paper covers a time period of the last eight years, volatility in the stock market has been a characteristic since the establishment of stock exchanges. Since the year 2000, however, we have seen plenty of volatility both in well defined crises and in less aggressive price swings. The bursting of the tech bubble, along with the current financial crisis are two of the major occurrences in sell-pressure volatility. The bulls took over at the beginning of 2003, resulting in constant price appreciation until the recent credit crisis began in August of 2007. Below are three charts displaying the trends in performance since 2000. Figure 1:

Source: Yahoo! Finance

119

Figure 2:

Source: Yahoo! Finance Figure 3:

Source: Yahoo! Finance Based on what the above charts show, it becomes very obvious that this eight year time range is comprised ultimately of one phase of buy volume sandwiched between two phases of sell volume. While it is not important in the scope of this paper whether there is buy or sell volume, it is important that there is volatility and movement in the market, which is established here. It generally seems to be widely accepted without much of a question as to what determines this volatility. When the tech stocks tanked, we assumed the market was down because of the 120

technology sector. Now, in the midst of a financial crisis, we are assuming that the market is down because of the sell pressure on financial stocks. But is that really the truth? That question is ultimately what we are looking to answer in this research paper. When the market has a significant price swing, what is actually the cause? An eight year performance chart of the S&P 500 (SPY) has been broken down into twenty-three “significant price swings” or as they will be referred to moving forward, “ranges.” Embedded in the chart as a black line is a sixty week moving average which helps determine these ranges. Each time the moving average intersects the performance line, the high and low point on either side is noted to determine a range. These ranges are marked out in the chart below with vertical red lines. Figure 4:

Source: Author Compilation with data from Yahoo! Finance The ranges determined statistically significant after regression analysis are numbered in the above graph. These are the ranges that will receive additional analytical attention in the remainder of this research paper. Included as Appendix B is a table outlining the exact dates of 121

each data range. Outlined in Table 1 is the breakdown of statistically significant and insignificant ranges: Table 1: Corresponding time range breakdown to Figure 4. Statistically: Significant 4 6 9 Insignificant 1 2 3

10 5

11 7

12 8

Range: 13 16 14 15

19 17

20 18

3.0 LITERATURE REVIEW Stock market volatility is a widely researched concept, with many different attack angles. According to Is There a Positive Relationship between Stock Market Volatility and the Equity Premium?(Kim, Morley & Nelson, 2004), volatility feedback is the idea that an exogenous change in the level of market volatility initially generates additional volatility as stock prices adjust in response to this new information about future discounted expected returns. Ultimately this means that premature concerns about weak earnings reports results in high sell volume and downward pressure on stock prices. Another offered possibility in Daily Stock Market Volatility: 1928-1989 (Turner & Weigel, 1992) suggests that most volatility is attributable to trading in the derivative markets. This makes sense presently, as there has been high volume trading in the options markets, as well as short selling of financial stocks. This explanation fits the bill as the financial sector has been the most volatile in recent months. One could also argue that volatility is due to movement in international stock markets, with the idea that markets need to “decouple.” However, in Volatility and Links between National Stock Markets (King, Sentana & Wadhwani, 1994) it is found that international markets are not quite integrated and have low correlation. While that paper is a bit outdated especially

122

given recent developments in the concept of decoupling, their findings hold significance in the idea of volatility. Unfortunately, there is a lack of literature regarding stock market sector performance, but that is the benefit of this paper, as we will analyze sectoral performance on the stock market. 4.0 DATA AND EMPIRICAL METHODOLOGY 4.1 Definition of Variables

SPYr = β0 + β1XLFr + β2XLKr + β3XLIr + β4XLVr + β5XLYr + β6XLPr + β7XLEr + β8XLUr +β9XLBr + ε SPYr represents the Standard & Poor’s 500 index for time period, or range “r.” This is the dependent variable, and is the benchmark used in this study to approximate overall stock market performance. All of the independent variables are representative of each sector of the stock market over different ranges, or “r.” XLF is the benchmark for the financial sector, XLK = Technology sector, XLI = Industrials, XLV = Healthcare, XLY = Consumer Discretionary, XLP = Consumer Staples, XLE = Energy, XLU = Utilities, and XLB = Materials. To sum it up, Table 2 displays each variable and what it represents. Table 2 Variable SPY XLF XLK XLI XLV XLY XLP XLE XLB XLU

Description S&P 500 Financial Sector Technology Sector Industrial Sector Healthcare Sector Consumer Discretionary Sector Consumer Staples Sector Energy Sector Materials Sector Utilities Sector

123

These benchmarks are all designed to track performance of each individual sector, and are actual Exchange Traded Funds (ETFs) available for trading on the stock market. The acronyms used as variables are actually ticker symbols for each ETF and can be reviewed on any financial website or the website of their creator, www.sectorspdr.com. They are constructed with approximately 50 companies that are considered major players in their respective sectors. 4.2 Data The study uses weekly closing price data from January 10th, 2000 through January 14th, 2008. Data were obtained from the Yahoo! Finance website. This weekly data for the SPY was then plotted and graphed, and a 60 week moving average was plotted along the line. Ranges of time were identified by looking for time periods where the moving average crossed the SPY line. These time ranges are considered significant price swings in the SPY, and thus qualifying these ranges for analysis. A 60 week moving average was used for two reasons; first, because betas for individual stocks are typically calculated on a 60 month basis, but since 60 months in this case represents over half of the total time range, it was reduced to 60 weeks. Second, a 60 week moving average eliminates most of the smaller and less important price swings in the SPY. Ultimately, twenty-three ranges were identified (see Appendix B for exact breakdowns), and the weekly closing prices downloaded from Yahoo! Finance were divided into those ranges for the SPY as well as all of the independent variables. Week over week percentage price changes were calculated and these numbers were used as inputs for the X and Y variables in the regression equation. A regression was run for each individual time range, totaling twenty-three regression outputs. The regressions without enough observations or statistically insignificant were discounted and removed from analysis, leaving thirteen of the original twenty-three outputs.

124

Once the Regression analysis was completed, the average gain or loss for each sector (X variables) in each range was calculated. This is necessary because we are comparing the largest gaining or losing sector with the highest correlation coefficient in each range to determine if there is consistency between the two measures. Gain or Loss is our measure of volatility, while correlation coefficients are the measure of explanatory power. Remembering that the key question we are trying to answer here is if volatility equals explanatory power. If it does, then we should see the most volatile sector in each range also have the highest correlation coefficient for that range. 5.0 EMPIRICAL RESULTS The results of the data analysis are astounding. Of the thirteen data ranges, only two showed consistency in variables from % Gain or Loss (Volatility) to Highest Correlation Coefficient (Explanatory Power). This suggests that sector volatility in this sample has very little to do with explaining overall market performance, and thus, completely negating what I would have expected. The major concern and implication of this in monitoring and aiding financial markets is that we may be implementing the wrong solutions, simply because we have not correctly identified the problem. This is particularly troublesome in range twenty-three. This range covers October 8th, 2007 through January 14th, 2008, the primary period of losses due to subprime writedowns and the credit crunch. While the XLF shows the greatest loss (volatility) of any sector during that period, regression analysis suggests otherwise. Regression shows the XLY, or consumer discretionary stocks, as the most explanatory in accounting for market performance over the aforementioned time period. This does make sense, as economists had been speaking of recessionary fears during that time, scaring consumers from spending, ultimately causing those companies to suffer the most.

125

However, my concern here is that the initial problem was thought to have been clearly identified as financial companies experiencing turmoil. That turmoil, being felt immediately, should logically be most explanatory for market performance, but instead, the expected troubles in the future earnings reports for consumer discretionary companies is what best explains market performance during this time. What could this tell us? Perhaps brokers, hedge funds, mutual funds, and institutional investors are more afraid of what could happen in the future than they are of what’s happening right now. Or maybe they recognize that there’s more than just one leak in the pipeline. Either way, it’s concerning that there isn’t more consistency from Volatility to Explanatory power. Table 3 outlines the results: Table 3 Column 1

Range Four Six Nine Ten Eleven Twelve Thirteen Sixteen Nineteen Twenty Twenty-One Twenty-Two Twenty-Three

Column 2

Column 3

Volatility

Explanatory Power

Variable with Greatest % Gain or Loss Financials Cons. Discretionary Technology Cons. Discretionary Technology Technology Technology Energy Cons. Discretionary Materials Technology Technology Financials

% Gain or Loss 0.84% 1.89% -1.88% 1.54% -2.11% 0.77% -0.49% 1.07% -0.29% 0.89% -1.19% 0.52% -2.17%

Variable with Highest Correlation Coefficient Technology Cons. Staples Financials Financials Cons. Staples Technology Financials Industrials Financials Financials Materials Technology Cons. Discretionary

Column 4 Correlation Coefficient 0.399 0.452 0.263 0.229 0.227 0.184 0.313 0.303 0.258 0.28 1.509 0.203 0.794

Sector Consistency? No No No No No Yes No No No No No Yes No

Column 1 identifies each data range, as these were the ranges resulting in statistical significance. Column 2 is our Volatility column, identifying which sector was the most volatile during the given time range. Volatility is measured in % Gain or Loss for the given range. Column 126

3 is the measure of Explanatory Power, identifying the sector with the most explanatory power as found through regression analysis. Explanatory Power is assigned by the correlation coefficient of the sector with the market, and is listed in the corresponding column. Column 4 simply determines if there is consistency in our measures across sectors. In other words, “Yes” means the most volatile sector is also the one with the most explanatory power, and “No” is vice versa. The two highlighted rows are the only data ranges with sector consistency. Based on the results, we can also compare a volatility index with an explanatory index to show the relationship between the two and find if there is any consistency in variables and their overall contribution to either measure. Table 4 outlines which sectors (variables) were most frequently calculated as the leader in volatility in Column 2 of Table 3. Table 5 outlines which sectors (variables) most frequently produced the highest correlation coefficients in Column 3 of Table 3, and thus, have the most explanatory power. Tables 4 and 5 are below: Table 4

Table 5 Volatility Breakdown

Volatility Ranking 1 2 3 4 4

Variable Tech C.D. Fin’ls Energy Mat’ls

Volatility Leader % 46.15% 23.08% 15.38% 7.69% 7.69%

Explanatory Breakdown Explanatory Ranking Variable Explanatory Leader % 1 Fin’ls 38.46% 2 Tech 23.08% 3 C.S. 15.38% 4 Ind’ls 7.69% 4 Mat’ls 7.69% 4 C.D. 7.69%

The “Leader %” columns show the frequency with which any one variable appears as the most volatile or as having the most explanatory power. This shows that in this sampling, Technology and Financial stocks are major market movers while Consumer Discretionary stocks are secondary. This information is important in that day traders and investors using technical analysis, option investors, and short sellers can expect to make the most money, and have the most opportunities 127

to profit by focusing on tech and financial stocks. This is because all of these types of investors thrive on volatility and movement. When the market moves sideways with no vertical movement, these investors lose money. One perfect example of the policy implications that have been discovered by this study is illustrated in Range 23. This time period is the bulk of the credit crunch that began in the summer of 2007 through the beginning of 2008. Our volatility leader was found to be financials, which logically makes sense. On the other hand, the sector found to have the most explanatory power is actually consumer discretionary. This, also makes sense, but the problem is that there is no sector consistency here. Consumer Discretionary stocks are generally hit the hardest during recessionary times, which explains why they had the most explanatory power in this recent time range. The issue is that these companies are affected most by consumer sentiment, and there were no preventive or corrective measures taken to ease and comfort consumers. Instead, we saw rate cut after rate cut to bail out banks and financial institutions. This study proves that that course of action was an error in judgment. The Fed believed that financial companies were driving the market, when in reality, consumer discretionary companies such as Best Buy or Home Depot were the actual root of the market movement. Instead of rate cuts, we should have taken measures to control inflation, and maintain a favorable exchange rate. Instead, we saw the complete opposite. 6.0 CONCLUSION Ultimately, we have determined that just because one sector moves the most in a given time range, chances are that it does not actually have the greatest impact or explanatory power on overall market performance. My expectations were proven to be inaccurate; however, this information can certainly create profit opportunities or avoid losses for investors with the perspective to realize these opportunities. By knowing and understanding that volatility is only the

128

surface of the movement and that it does not identify which sector is the root of the market movement, we become investors with a deeper and more skeptical vision of investing. These results also imply that taking corrective measures in periods of market turmoil may not address the actual rooted problem. Therefore, taking drastic action without sufficient information and analytics could further deteriorate markets, or delay recovery. If we look specifically at the most recent data range, which covers the credit crisis of the last few months, we can see that consumer discretionary stocks are actually the most impactful on the market. So in terms of policy implications, one must consider the possibility that rate cuts to bail out financial firms may not have been the most prudent decision. If consumer spending is what really drove the market, then the Fed should have focused more on inflation and the value of the dollar. Hindsight is 20/20, however, and perhaps this new information could be of some assistance in future policy decisions. While this paper doesn’t offer a step-by-step solution to what the real determinants of market movement are, it does prove that there is more analysis required to create and implement an accurate and efficient solution. We have found that volatility does not equal explanatory power, and thus, finding the most volatile sector and attempting to ease this volatility is not the needed solution.

129

Appendix A: Variable Description and Data Source

Acronym

Description

SPY

S&P 500 ETF, tracking the performance of the S&P 500

XLF

ETF designed to accurately track the Financial Sector

XLK

ETF designed to accurately track the Technology Sector

XLI

ETF designed to accurately track the Industrial Sector

XLV

ETF designed to accurately track the Healthcare Sector

XLY

ETF designed to accurately track the Consumer Discretionary Sector

XLP

ETF designed to accurately track the Consumer Staples Sector

XLE

ETF designed to accurately track the Energy Sector

XLU

ETF designed to accurately track the Utilities Sector

XLB

ETF designed to accurately track the Materials Sector

Source Yahoo! Finance Yahoo! Finance Yahoo! Finance Yahoo! Finance Yahoo! Finance Yahoo! Finance Yahoo! Finance Yahoo! Finance Yahoo! Finance Yahoo! Finance

130

Appendix B – Range Breakdowns Range # From To 1 1/10/2000 2/22/2000 2 2/22/2000 3/20/2000 3 3/20/2000 4/10/2000 4 4/10/2000 8/28/2000 5 8/28/2000 10/9/2000 6 10/9/2000 1/29/2001 7 1/29/2001 3/19/2001 8 3/19/2001 5/14/2001 9 5/14/2001 9/24/2001 10 9/24/2001 3/11/2002 11 3/11/2002 9/30/2002 12 9/30/2002 3/1/2004 13 3/1/2004 8/2/2004 14 8/2/2004 9/13/2004 15 9/13/2004 10/11/2004 16 10/11/2004 2/28/2005 17 2/28/2005 4/11/2005 18 4/11/2005 6/13/2005 19 6/13/2005 10/17/2005 20 10/17/2005 5/1/2006 21 5/1/2006 7/10/2006 22 7/10/2006 10/8/2007 23 10/8/2007 1/14/2008 Green dictates the ranges with valid regression outputs

131

BIBLIOGRAPHY Kim, Chang-Jin; Morley, James C.; Nelson, Charles R.; (2004), “Is There a Positive Relationship between Stock Market Volatility and the Equity Premium?” Journal of Money, Credit and Banking, Vol. 36, No. 3, Part 1, pp. 339-360. Turner, Andrew L.; Weigel, Eric J.; (1992), “Daily Stock Market Volatility: 1928-1989” Management Science, Vol. 38, No. 11, Focused Issue on Financial Modeling, pp. 15861609. King, Mervyn; Sentana, Enrique; Wadhwani, Sushil; (1994), “Volatility and Links between National Stock Markets” Econometrica, Vol. 62, No. 4, pp. 901-933. Yahoo! Finance. Mar. 2008 <http://finance.yahoo.com>.

132

United States Homeownership Rates: The Effect of Macroeconomic Factors on the Domestic Real Estate Market

Alexander N. Grande1

Abstract: This paper examines the correlation between U.S. economic indicators and the domestic real estate market. The analysis of the study’s findings and results show that some indicators adversely affect homeownership rates in conjunction to the overall state of the market during the time frame depicted. The regression is from a model used on an international level and it is taken and applied here to the domestic market of one country; the United States. The results from the research and tests performed highlight the economic indicators that are closely correlated to the rate of homeownership.

JEL Classification: R21

Keywords: Homeownership Rates, Factors that influence

Department of Economics, Bryant University, 1150 Douglas Pike, Smithfield, RI 02917. Phone: (508) 728-2523. Email: agrande@bryant.edu

The author thanks Anthony Delmonico and Joshua Lopes for providing data and gratefully acknowledges the guidance from Professor Ramesh Mohan.

133

1.0 INTRODUCTION The focus of this study is that of a prevalent macroeconomic issue, the housing market. The test is to portray the effects that independent economic factors have on the percentage of residential homes in the United States which are owned (not rented or leased). The percentage of homes owned is a sound indicator of the economy and one that analysts frequently refer to and use as a benchmark to measure performance. Since the housing market is dependent on individuals, indicators chosen were broad independent factors that affect each and every individual therefore influencing their decision to own or rent/lease a home. Basic economic theory supports that the housing market is pro-cyclical to the performance of the economy. That is if the economy is prospering the housing market will be on the upswing because all individuals will be reaping the benefits and purchasing new homes. The opposite is also true, if the economy is experiencing hardship it will be evident in the figures of the housing market. David A. Lareah, the senior vice president and chief economist of the National Association of Realtors was published stating “The housing sector directly and indirectly accounts for about 15 to 20 percent of our nation’s Gross Domestic Product (GDP) every year. Moreover, most studies indicate that households spend about 30 to 40 percent of their disposable income on housing-related expenses. Those expenditures help to support other sectors of the economy.”4 The study’s mission is to use national economic data to test and further reinforce the economic theory that these factors have an effect, either positive or negative, on the housing market as a whole. “NAR shared its findings with the Federal Reserve Board in mid-October, in response to a meeting between the association’s leadership and Chairman Greenspan earlier this year. “At the time, Mr.Greenspan theorized that the wealth effect of homeownership was offsetting some of the losses on Wall Street in the overall economy – this survey shows Mr. Greenspan is right.”5 Home ownership has tremendous social benefits, stabilizing neighborhoods and making people more willing to invest in their communities. And it has economic benefits, too, serving as a forcedsavings program that allows people to leverage their incomes and build wealth. Thus is the correlation that exists with the homeownership rates and the stability and state of the economy. The equity that one can gain from homeownership makes up for approximately 45% of the households wealth. This study aims to enhance the overall perspective of the U.S. domestic homeownership rate and its impact on the economy along with the factors that contribute explaining the homeownership rate. This analysis is important because the housing market is a key component in the economy’s well being. To

4 2

http://www.realtor.org/sg3.nsf/Pages/americashousing?OpenDocument http://www.realtor.org/PublicAffairsWeb.nsf/pages/WealthEffect?OpenDocument

134

discover and elaborate on the economic indicators that closely describe or that explain homeownership rates will help down the road in predicting housing market turbulence and simultaneously offer various reasons to why the troubles came about and how they can be aided. This study looks to offer a substantial analysis into these factors. The economic model which this empirical analysis is based on is that of Fisher and Jaffe (2003). Their model however looked at homeownership rates through the span of 106 countries throughout the world, whereas this analysis will remain domestic and look at the homeownership rates in the U.S. The comparative differences and the contribution of this study do not simply lie in the contrast of regions taken into consideration but the variables in which the model uses to attempt to describe the homeownership rate. This examination has added numerous variables not taken into consideration by Fisher and Jaffe such as average household income, the DJIA, real GDP, spending on residential construction, the civilian unemployment rate, and the interest rate. 2.0 HOMEOWNERSHIP TRENDS IN THE U.S. “The gains of the last ten years have lifted homeownership growth to a higher trajectory. Remaining on this path depends on whether the recent conditions that have strongly favored homeownership can continue. A major reason for the recent climb in homeownership is that house price appreciation has been unusually strong over the past five years. In addition, long-term interest rates have remained at historic lows even as short-term rates have returned to more normal levels. If the economy picks up steam, interest rates are likely to increase and the growing share of households with adjustable-rate mortgages will find themselves with rising payments. Interest-only borrowers who do not sell their homes or refinance before principal payments come due will also get hit by much higher payments. Already though, an increasing number of borrowers have refinanced their adjustable loans.”6 If the economy instead stumbles and job growth falters, a larger number of subprime borrowers will be at greater risk. At the same time, however, the lower interest rates that usually accompany such slowdowns would help adjustable-rate borrowers and create opportunities for other homeowners to refinance their loans on more favorable terms. These trends in homeownership also come from promotion of homeownership through various organizations and programs aimed to making owning homes affordable and discovering different ways to finance more middle class Americans into owning homes.

http://www.jchs.harvard.edu/publications/markets/son2006/son2006_homeownership_trends.pdf

135

The housing trends are best looked at throughout the country broken up into four regions; Northeast, Midwest, South, and West. Rates vary from region to region due to the factors of population and industrialization which both are directly correlated with the amount of homes occupied and owned in the country. The four markets vary in there stability as well where the Northeast has been in turmoil where housing prices are increasing and it is becoming more and more expensive to own a home, yet this is somehwhat offset by the higher income per capita in the region. This is in comparison to they West perse where housing is cheap comparitvely to the Northeast, but the income is not what it is in the Northeast. In looking into trends in homeownership rates throughout this country, regions must be taken into account due to the various lifestyles explored by each region and the lending markets status in each respective region. An interesting sceneriao is what takes place in the years 2001-2003, where the homeownership increased in the midst of the economy entering a recession. The practical belief would be homeownership would take a hit when the economy is in a recession, yet the graphs and data shows otherwise. Even through the recession homeownership rates continued to climb and it was the attributes of the recession which led to the climb. This paper will take a look into how a high unemployment rate and a high inflation rate actually boost home ownership and there is empirical evidence and resaearch to credit the validity of their positive influence. A factor also taken into consideration when describing homeownership rates is that of race. Through the different races the U.S. has to offer, homeownership rates fluctuate considerably from one to another as the graph shows.

136

Minorities seem to be less likely to own a home in the U.S. and it boils down to the income inequality among each race. Race however was not included or described in the research to be considered a significant economic indicator. Its implications have no direct impact on the economy therefore it was not taken into account as a variable in the regression. However, race and income are correlated and income indeed was one of the factors chosen to be a variable, so there is an intuitive link with race in the regression. As the graph previously showed the average homeownership rates among different races, this graph shows the average income between the races in the U.S.

The graphs together show the correlation between race and income and how the effect of race can be captured in the data that supports average household income in this model. Overall trends in homeownership are attributed to the aging of the population as well. “To explore the role of changing demographics in the increase in the U.S. homeownership rate, we first look at broad trends between 1994 and 2004 using data from the Current Population Survey, which is conducted by the Bureau of Labor Statistics. Figure 2 breaks the population into three age groups, and the results confirm the well-known fact that the age distribution has shifted as the baby boom

137

generation has moved up the age scale. Figure 3 shows the changes in homeownership rates within these three age groups, and, as expected, it shows that households headed by older people are more likely to be homeowners; it also shows that homeownership rates increased between 1994 and 2004 for young, middle-aged, and older Americans. Given these data, it seems natural to explore whether the aging of the population or the greater propensity for households within each age category to be homeowners accounts for most of the increase in the overall homeownership rate.”7

The trends in homeownership rates have been affected by many different characteristics of this country’s demographics including region, race, and age breakdowns of the population. Yet, these factors are not conclusive to the point where they give an idea of how the economy and the rate itself are directly correlated. To do so this paper will examine economic indicators that take into account the demographics discussed along with factors that directly measure economic growth and the overall stability of the economy; macroeconomic factors. 3.0 LITERATURE REVIEW In the determining of factors that affect the domestic homeownership rates, different analysts and economists come from varying schools of thought. Chevan (1989) discusses that there is economic incentive for owning a home rather than renting because of the very high return on equity, due in part to high rates of inflation. The model that Chevan uses in his study is in conjunction with the financial trend side of the homeownership and focuses on various trends this country has seen in accordance with owning 7

http://www.frbsf.org/publications/economics/letter/2006/el2006-30.html#sub1

138

homes. The ideals brought up in his paper that are really scrutinized in this study is the in the likes of inflation and its effect on homeownership. Fisher and Jaffe (2003) believe that variation in homeownership rates deal with in the availability of inputs to the housing sector and the overall supply of housing. This analysis was that of international scope and took into account 106 countries. Their variable model will be the model in which this paper has based its model on. What was taken from Fisher and Jaffe analysis was the theory of availability of inputs and overall supply. Essentially this paper was the inspiration of the empirical study at hand simply to apply its international analysis on a domestic level, particularly the U.S. The research of Green and Hendershott (1999) has found that there is a positive relationship between homeownership rates and unemployment due to the psyche of the public through their research of a British economist A.J. Oswald. Green and Hendershott provided a much specified look into the relationship between the homeownership rate and unemployment in the U.S. The look at strictly unemployment and its effects offers great intuition into how the two variables are intertwined. Yabaccio, Rubens, and Ketcham (1995) took an in depth look into how real estate can be used to be a partial hedge against inflation measures. Their study showed this through the observation of REITs or Real Estate Investment Trusts, which is a vehicle for investment in the financial markets. This is where the study at hand varies REITs are not taken into consideration due to the level of economics in which it applies. It is more so a microeconomic level where this paper takes into account mainly macroeconomic factors that focus on nationwide indicators that have an effect on homeownership rates. What this study adds to the overall research of homeownership rates is the U.S. perspective and a look at a country’s macroeconomic factors that have potential to affect the rates. It will attempt to explain the unexpected contribution of factors such as unemployment and inflation that actually, contrary to popular belief and economic theory application, are positively correlated to the homeownership rate. Regression analysis will provide intuition into the factors that actually should be pegged to the homeownership rates. All research, variables, and analysis sustained within the literature taken into account, will indefinitely be used to provide a sound foundation for the study at hand. 4.0 DATA AND EMPIRICAL METHODOLOGY 4.1 Definition of Variables

OWN =β0 + β1INT + β2DOW + β3UNEMP + β4INC + β5INF + β6GDP + β7HHSIZE + β8POP + β9CONSTSPEND+ 𝜀 OWN is the percentage of the country that has ownership in residential real estate. OWN is used here as the dependent or endogenous variable. The variables that are considered the independent variables

139

in this model were researched and developed to assure that they will explain the dependent variable and prepare a model that would seek to give a competent report of the factors that can describe the dependent. The independent variables are that of macro proportions as they take a look into economic indicators of the country in aggregate. INT represents the interest rate and captures the rate at which a consumer can borrow capital to invest in various ventures including buying a home. DOW measures the monthly Dow Jones average which is measure of the stock market’s performance and indirectly the state of the economy. UNEMP depicts the civilian unemployment rate contributing what percentage of the work force is unemployed. INC is household income which combines the gross income of all the members of a household who are 15 years and older. Individuals do not have to be related in any way to be considered members of the same household. INF is the inflation rate and it captures the rate at which the general level of prices for goods and services is rising, and subsequently, purchasing power is falling. GDP is gross domestic product which reflects the value of all goods and services produced in a given year, expressed in base-year prices. HHSIZE is household size and is shows the average size of the families in the US. The number of normally resident members of a house is its size. It will include temporary stay ways but exclude temporary visitors and guests. POP represents the population of the country in each year. CONSTSPEND is the monetary spending on residential construction which majority is private homes in the country. 4.2 Data The study uses monthly real data from 2000-2005. Data was obtained from the Bureau of Labor Statistics (BLS), the Bureau of Economic Analysis (BEA) and the U.S. Census Bureau. The summary of the statistics is exhibited in Table 2. 5.0 EMPIRICAL RESULTS AND ANALYSIS The analysis is based on 72 observations for each independent variable. The data collected spans over a 6 year period (2000 – 2005) and was reported in monthly intervals. Each variable was logged to rid the equation of too much serial correlation and therefore strengthen its accuracy. The summary of the regression is given in Table 3 and can be interpreted as follows. As the nationwide civilian unemployment rate (UNEMP) increases by 1% the percentage of homes owned will increase by .03918%, this varies from the negative expected sign that was assumed. As the average monthly inflation rate (INF) increases by 1% the percentage of homes owned will increase by .0084%. These concepts are a little harder to grasp than the other variable analysis. The regression runs into the problem that two of the variables in the model’s output have the opposite effect of what initially was expected. There was an expectation that the nationwide civilian unemployment rate and the average monthly inflation rate would have a negative impact on the percentage of homes owned. More simply stated when unemployment and inflation increased the percentage of homes owned would have 140

decreased. According to this study’s data this initial hypothesis is not true, both the unemployment rate and inflation rate have a positive effect on the percentage of homes sold or as unemployment and inflation increase so does the percentage of homes owned. After this discrepancy much research found evidence to support the regression output and disregard the initial hypothesis. Unemployment is likely to have a positive impact on the percentage of homes owned because when the unemployment rates rise, homeowners are less likely to move because of the increased cost of moving out while unemployed, compared to moving from an apartment. This concept is detailed in an academic account written by Richard Green and Patric Hendershott and specifically states, “Oswald (1999) emphasizes a number of "indirect" effects. For example, areas with high home ownership rates have greater planning laws and restrictions on land development, discouraging business start-ups, and have greater congestion owing to owners commuting further than renters, increasing the cost of having a job. Of course, the primary reason for the ownership-unemployment relation is simply the larger costs of vacating a home (selling costs) versus moving out of an apartment.”8 The positive relationship between homeownership and inflation is a concept that requires a great deal of intuition. One would tend to believe that if the inflation rate increases the home ownership rate would decrease however upon further research it was found that the two are positively related. The home ownership rate and the inflation rate are said to be positively correlated, but the relationship is minimal at best. There has been no conclusive research to date that has explained why home ownership rates and inflation are positively related. The following account from an academic essay written by Elizabeth Yobaccio, Jack Rubens, and David Ketcham explains the positive relationship between the inflation rate and home ownership more precisely: “This study examined the inflation-hedging effectiveness of REIT returns using a model that posits real estate returns are a function of expected inflation, unexpected inflation, and the real return to a market index. Four types of REIT return measures (equity, mortgage, hybrid, and a composite index) were used, as were four expected inflation forecasts across an extended time period (1972:02 through 1992:12). Results indicate that REITs act as poor hedges against any measure of inflation (actual, expected or unexpected) with the poorest performance relative to unexpected inflation. In this respect, REIT returns mirrored results involving equity returns in general and would seem not to be proxies for direct investment in real estate. Studies that have shown the real estate's ability to act as at least a partial inflation hedge may be the result of the well-documented appraisal basis in such returns, rather than real estate's innate ability to act as an effective hedge. Evidence on REITs indicates that real estate, at best, acts as a partial hedge against expected inflation and a perverse hedge against unexpected inflation.”9 8 9

http://www.nmhc.org/Content/ServeFile.cfm?FileID=165 Yobaccio, Rubens, and Ketcham, “The Inflation-Hedging Properties of Risk Assets: The Case of REITs”

141

Coinciding with the belief that inflation and homeownership are positively correlated, a study done by Albert Chevan states, “Although not usually viewed as a protection against inflation, homeownership has served this function because house values have generally kept stride with or exceeded the pace of inflatin........Peiser and Brueggeman (1982) found a decided economic incentive for owning a home rather than renting because of the very high return on equity, due in part to high rates of inflation.”10 The findings of Chevan only strengthen the case of the positive relationship of inflation to homeownership rates, coming from an earlier study and being the basis of a newer study done by Yobaccio, Rubens, and Ketcham. As the fixed monthly mortgage rate (INT) increases by 1% percentage of homes owned will decrease by .0092% making the interest rate negatively correlated with the homeownership rate. This is what was hypothesized as (INT) had a negative expected sign showing that when the interest rate is high it negatively affects the homeownership rate. In theory this is correct because as it becomes more expensive to lend from creditors and banks more and more people steer away from such methods that are imperative for majority of the country to own a home. For every 1 dollar the Down Jones Industrial Average (DOW) increases the percentage of homes owned will increase by .0149%. With the DJIA increasing homeownership is increasing which is practical because the DJIA is one measure of how well the economy is doing and is it positively increases so do the benefits for society, especially that of homeownership. For every 1 dollar average monthly household income (INC) increases the percentage of homes owned will increase by .21434%. This relationship between income and homeownership is one that was expected and easily explained. As income for a household rises, they tend to steer away from renting or leasing and begin owning homes with an increase in purchasing power. As the GDP increases by 1% the homeownership rate will increase by .393%. Along with the DJIA, GDP is a measure how a country’s overall well-being and stability but it’s more of an aggregate look at the country’s goods and services as whole, all industries included. One would expect that if GDP were to increase so would homeownership, so the expectations of the variable were correct. As CONSTSPEND increased by 1% the homeownership rate decreases by .0197%. The expected sign of this variable was initially positive but when one thinks about the rationale overview of the spending on residential construction, the more money and capital that go into the construction of homes, the higher the prices of homes hindering homeownership. As HHSIZE increases by 1% the homeownership rate also increases, by .838%. With the household size higher, there are more occupants within the household with the chance of increasing the overall average household income, making homes more affordable. Lastly, as POP goes up by 1% the homeownership rate goes

Chevan, “The Growth of Home Ownership:1940-1980”

142

down by 1.33%. Population also varied from its positive expected sign and seemed to have a negative impact on homeownership. This relationship can be loosely interpreted by the thought process that as more people enter this country the demand for housing rises and in doing so, prices rise. There was no empirical evidence or papers found to properly describe the relationship between population and homeownership. Since the data is financial and therefore captured in a time series fashion it is almost inevitable that a certain degree of multicollinearity and serial correlation will be present. This is not as worrisome as if the data wasn’t time series because all financial data tends to move together because it is a reflection of the economy as a whole. That is all of these independent variables have some effect on each other. It is imperative to know that when conducting this analysis the coefficient of each independent variable was measured while holding all other independent variables constant. Other important summary statistics of the regression output are as follows. T-statistics are the next portion of summary statistics to examine after the coefficients of the independent variables are interpreted. T-statistics are a measurement used in hypothesis testing and are calculated for each individual independent variable by dividing the variables’ coefficient by its standard error. The rule of thumb for these is if the t-statistic is greater than 2 at 95% confidence the null hypothesis can be rejected. By rejecting the null hypothesis you can conclude that the effects of the individual independent variable on the dependent variable are significant in the expected direction. The R-squared coefficient places a significant amount of confidence in the variable choice of the analysis because approximately 88% of the movement of the dependent variable is explained by the independent variables. This means that variables that were researched and hypothesized fit the model or explained the dependent variable, homeownership rate, sufficiently well. A high R-squared gives credibility to the model and therefore the research and ideals that it claims. 6.0 CONCLUSION In summary the variables researched and chosen to depict and describe the U.S. domestic homeownership rate were all relevant and fit the model. Though there were some discrepancies with a few variables and their expected correlation to the dependent variable, empirical studies were found to support the theories that were created and noted in this regression. The study finds that high unemployment and inflation rates are essentially positively correlated with homeownership rates defying the common practical belief of their negative relationship. The examination also has findings that the interest rate has very little relevance to the homeownership rate along with the result of population having a negative impact. These two flaws in the model however were not explained by and empirical data and leave room for more explanation.

143

Overall the variables that were put together in this model serve their main purpose and in doing so have created a model that is sufficient in relating the macroeconomic factors to the percentage of homes owned in the U.S. With the given information and analysis one can propose various programs aimed to stimulate these factors with the ultimate goal of raising the homeownership rate. Some programs have been in the making such as the immigration laws that look to limit the amount of people in the use which would help the population which, from the empirical results, has a negative impact on homeownership. Also minimizing overhead costs of construction in dealing with residential real estate, with lower costs, homes become cheaper and more attractive and affordable to different income brackets boosting the homeownership rate. What is important about this study is that it offers a macroeconomic view making the initiation of changes and programs a lot easier say if it were the other way around coming from the micro view. In addition, this country needs to see its homeownership rate grow establishing more wealth and capital throughout its communities and much can be done by catalyzing change through the watch and stimulation of various macroeconomic indicators.

144

Table 1: Variables, Descriptions, Sources, and Expected Signs Acronym

Description

Data Source

INT

The interest rate captures that rate at which consumer borrow capital to invest in buying a home. The Dow Jones average measures the stock market’s performance and indirectly the state of the economy. The civilian unemployment rate depicts the rate at which a percentage of the work force is unemployed. Household income is the combined gross income of all the members of a household who are 15 years old and older. Individuals do not have to be related in any way to be considered members of the same household. Inflation rate captures the rate at which the general level of prices for goods and services is rising, and, subsequently, purchasing power is falling. Gross Domestic Product reflects the value of all goods and services produced in a given year, expressed in base-year prices. The average size of the families in the US. The number of normally resident members of a household is its size. It will include temporary stayaways but exclude temporary visitors and guests. Population is the number of people occupying the country at a given time. Construction spending is the amount of money and capital spent on the building of residential homes in the US.

BEA

Expected Sign (-)

BEA

(+)

BLS

(-)

Census Bureau

(+)

BLS

(-)

BEA

(+)

Census Bureau

(+)

Census Bureau

(+)

Census Bureau

(+)

DOW UNEMP INC

INF

GDP HHSIZE

POP CONSTSPEND

145

Table 2: Summary of Statistics Variable

Obs.

Mean

Std. Dev.

Min

Max

OWN

68.21

.628826

67.10

69.20

INT

6.52

.863313564

5.23

8.52

DOW

9980.32

860.118816

7591.93

11215.10

UNEMP

5.18

.730854

3.80

6.30

INC

4902.61

161.0441

4767.33

5213.00

INF

2.69

.81675

1.07

4.69

GDP

38.27

1.983311

36.20

41.60

HHSIZE

2.59

.011132999

2.58

2.61

POP

287197500.00

4916805.619

281421000.00

295507000.00

CONSTSPEND

6316601.00

1237520.425

4653416.00

7804216.00

Table 3: Regression Results of Homeownership Rates

Variables LOGINT LOGDOW LOGUNEMP LOGINC LOGINF LOGGDP LOGHHSIZE LOGPOP LOGCONSTSPEND

Coefficient -0.009164 (.3874) 0.014852 (.0740) 0.039175 (.0005) 0.214337 (.0486) 0.008382 (.0009) 0.392706 (.0001) 0.837873 (.0223) -1.327252 (.0003) -0.019732 (.0167)

Constant

11.35168

.884430

F-Statistic

52.71886

Observations

146

Table 4: Variable Correlation Table

int dow unemp inc inf gdp constspend pop hhsize

int 1.00 0.37 -0.77 0.84 0.24 -0.71 0.10 -0.78 -0.56

dow unemp 0.37 -0.77 1.00 -0.63 -0.63 1.00 0.45 -0.89 0.48 -0.67 0.17 0.33 0.52 -0.42 0.08 0.41 -0.07 0.35

inc 0.84 0.45 -0.89 1.00 0.49 -0.55 0.22 -0.60 -0.45

inf 0.24 0.48 -0.67 0.49 1.00 0.16 0.56 0.13 0.15

gdp constspend -0.71 0.10 0.17 0.52 0.33 -0.42 -0.55 0.22 0.16 0.56 1.00 0.48 0.48 1.00 0.98 0.42 0.53 0.45

pop hhsize -0.78 -0.56 0.08 -0.07 0.41 0.35 -0.60 -0.45 0.13 0.15 0.98 0.53 0.42 0.45 1.00 0.63 0.63 1.00

This table looks to explain the correlation amongst the variables in the equation. Too much serial correlation would render some variables redundant and therefore serve no purpose in the equation. The highlighted red box indicates when the correlation between variables became a little too high.

147

BIBLIOGRAPHY 1. Chevan, Albert. "The Growth of Home Ownership: 1940-1980." Population Association of America: Demography 26(1989): 249-266. 2. Fisher, Lynn M., and Austin J. Jaffe. "Determinants of International Home Ownership Rates." Housing Finance International (2003): 34-42 3. Green , Richard K., and Patric H. Hendershott. "Home Ownership and Unemployment in the U.S." National Multi Housing Council (1999): 1-15 4. Yobaccio, Elizabeth, Jack H. Rubens, and David C. Ketcham. "The Inflation-Hedging Properties of Risk Assets: The Case of REITs." The Journal of Real Estate Research 10(1995): 279-296 5. "Eye on the Economy." Nation's Building News 2006 http://www.hgtvpro.com/hpro/cda/article 6. "About Smart Growth: Successful Growth Begins with Five Principles ." National Association of Realtors 2000 http://www.realtor.org/smart_growth.nsf/pages/aboutsmartgrowth?opendocument

7. "The Rise in Homeownership." FRBSF Economic Letter 2006 http://www.frbsf.org/publications/economics/letter/2006/el2006-30.html#sub1

8. Simmons, Patrick A.. "A Coast to Coast Exapnsion: Geographic Patterns of U.S. Homeownership Gains During the 1990s." Fannie Mae Foundation Census Note 05 2001 http://www.fanniemaefoundation.org/programs/census_notes_5.html#factors

9. Surowiecki, James. "Home Economics." The New Yorker: The Financial Page March 10, 2008 http://www.newyorker.com/talk/financial/2008/03/10/080310ta_talk_surowiecki

10. Garriga, Carlos and William T. Gavin, Don Schlagenhauf. "Recent Trends in Homeownerhsip." Federfal Reserve Bank of St. Louis Review 2006 397-411. http://research.stlouisfed.org/publications/review/06/09/Garriga.pdf 148

11. Myers, Dowell. "Advances in Homeownership Across the States and Generations: Continued Gains for the Elderly and Stagnation Among the Young ." Fannie Mae Foundation Census Note 08 2001 http://www.fanniemaefoundation.org/programs/census_notes_8.html

12. All Data was found through U.S. government databases including the Bureau of Labor Statistics, the Bureau of Economic Analysis, and the U.S. Census Bureau.

149

Causality between Defense Spending GDP and Economic Growth Ryan P. Daleya Abstract This paper addresses whether or not the government members’ of the “Coalition of the Willing” military expenditure as a participant in the war in Iraq will help to generate domestic economic growth on the eve of an impeding recession. This paper analyzes the findings during the time period 1989-2006 regarding military expenditure as a percentage of GDP, and its effect on GDP growth, through comparing that relationship between studies and in different political, socioeconomic circumstances. It also performs the same tests on the most recent and complete set of data available for the 31 member nations of the Coalition, to see the result of past spending activities and whether or not there is a causal relationship. The paper concludes that there is no Granger-causality between military expenditure as a percentage of GDP and economic growth in any of 24 the countries for which regression analysis could be performed. JEL Classification: H59, O41

Keywords: military expenditure, economic growth

Department of Economics, Bryant University, 1150 Douglas Pike, Smithfield, RI 02917. Phone:

(774) 245-0034. Email: ryan_p_daley@yahoo.com.

The author gratefully acknowledges the guidance of Dr. Ramesh Mohan in the completion of this paper. 150

"Every gun that is made, every warship launched, every rocket fired signifies, in the final sense, a theft from those who hunger and are not fed, those who are cold and not clothed. This world in arms is not spending money alone. It is spending the sweat of its laborers, the genius of its scientists, the hopes of its children. This is not a way of life at all in any true sense. Under the cloud of threatening war, it is humanity hanging from a cross of iron." ~President Dwight D. Eisenhower, April 16, 1953 1.0 INTRODUCTION The most recent estimates of the cost of the U.S. led war in Iraq range from $1.2 trillion (Leonhardt, 2007) to $3 trillion (Stiglitz and Bilmes, 2008), each from some of the most highly regarded authorities on the topic. The Pentagon’s proposed budget for annual military spending, if fully approved, will have reached a level unseen since WW II after adjusting for inflation (Shanker, 2008). President Bush told the Today Show, “I think actually the spending in the war might help with jobs…because we’re buying equipment, and people are working. I think this economy is down because we built too many houses and the economy’s adjusting,” as many experts predict an impeding recession: former Treasury head, Larry Summers, prominent global bank, Goldman Sachs, and president and chief executive of the National Bureau of Economic Research, Martin Feldstein, among others (Reuters, 2007; Reuters, 2008; Wingfield, 2008). A study by Global Insight (2008) attributes close to one-third of U.S. economic growth in 2003 to the war in Iraq, however not many other economic analyses have made conclusions about the war and it’s effect on economic growth. Using Granger-causality (GC) testing on the variables, military expenditure as a percentage of GDP, and economic growth in the United States, this study concludes that neither variable Granger-causes the other in any of the countries involved in the war, and so therefore, claims that the current high level of national defense spending in the U.S. will help contribute to growth are unsubstantiated in economic fact. Most of the previous studies of this nature were performed in the 1970s and 80s with earlier data, and utilized cross-sectional evidence across different countries, and employed ordinary least squares (OLS) equations based on the assumption that defense spending causes economic growth, without testing for whether the reverse assumption could be true: economic growth causes defense spending. Joerding (1986) concluded that it is equally plausible for the latter to occur. This study

151

is different from others because it uses GC testing and not the OLS estimation, in an attempt to compare the directions of causality between military expenditure and economic growth in the U.S. Furthermore, other studies take nominal military expenditure into consideration, and do not look at the changes in military expenditure as a percentage of GDP. As some countries’ GDP increases, so too does their military expenditure, however when we observe this as a percentage of GDP we receive a different estimation than other studies have shown by not taking into account how GDP growth may be affected by the percentage of military expenditure as a percentage of GDP instead of simply a nominal amount that may or may not go up at a rate proportional with GDP growth. Accordingly, this study takes the log of military expenditure as a percentage of economic growth as a variable instead of nominal military expenditure. All of the countries included in this study are a part of the “Coalition of the Willing,” the group of 30 nations in addition to the United States, who according to the U.S. State Department participated in initial invasion of Iraq: Afghanistan, Albania, Australia, Azerbaijan, Bulgaria, Colombia, the Czech Republic, Denmark, El Salvador, Eritrea, Estonia, Ethiopia, Georgia, Hungary, Italy, Japan, South Korea, Latvia, Lithuania, Macedonia, the Netherlands, Nicaragua, the Philippines, Poland, Romania, Slovakia, Spain, Turkey, United Kingdom and Uzbekistan (Schifferes). First, this paper discusses the trends regarding GDP growth rate, military expenditure as a percentage of GDP, and exhibits graphical analysis of each plotted against an many countries at once and simply one country’s respective rates alone. The trends section suggests that GDP growth rate is extremely volatile, however military expenditure as a percentage of GDP has been decreasing since 1989. Next, the literature review discusses previous papers on the topic and emphasizes that many studies have shown different results depending on which countries were tested, where those countries were in their development stage, what time periods were included, what political state was implemented, among others. The data and empirical methodology section explains the steps in used to test the data using Granger Causality testing, but first the data must be tested with the Augmented Dickey Fuller test and the Johansen test. Finally, the empirical analysis and conclusion conclude that the countries tested show no Granger Causality between the two variables.

152

2.0 TRENDS Most countries have an average GDP growth rate of between .5% and 4.5%, however most have at one point or another, experienced growth outside of that spectrum. Particularly, Romania, Turkey, Japan, Hungary, and South Korea, and Italy have experienced some of the most volatile growth. There are many factors that contribute to GDP growth rate volatility including political instability, openness and susceptibility of smaller countries to foreign shocks, recessions, level of technology, among others (Canning, et al., 1998). The causes of each country’s specific GDP growth rate and if applicable, its volatility, is beyond the scope of this paper, however they are graphed in Figure 1 to exhibit the trend and average of a sample of the countries tested.

Figure 1: GDP Growth Australia Bulgaria Denmark

Ethiopia Hungary Italy Japan Korea, Rep. Netherlands Philippines Romania Spain Turkey United Kingdom United States

153

Most countries tend to keep military expenditure as a percentage of GDP between the 1%3% range, but there also appears to be a trend since 1989 of military expenditure as a percentage of GDP decreasing slightly. This is depicted in Figure 2.

Figure 2: Military Expenditure (as a % of GDP)

Australia

Bulgaria Denmark Ethiopia Hungary Italy Japan Korea, Rep. Netherlands Philippines Romania Spain Turkey United Kingdom United States

154

A sample of single countries’ military expenditure as a percentage of GDP and the respective country’s GDP growth rate are plotted on the same graphs in Figure 3, to exhibit that no strong visual relationship or correlation exists.

Figure 3A: Albania

Albania GDP growth (annual %)

Figure 3B: Denmark

Denmark GDP growth (annual %)

Albania Military expenditur e (% of GDP)

Denmark Military expenditur e (% of GDP)

Figure 3D: Hungary

Figure 3C: Ethiopia

Hungary GDP growth (annual %)

Ethiopia Military expenditur e (% of GDP)

Figure 3E: Korea

Axis Title

Ethiopia GDP growth (annual %)

Hungary Military expenditur e (% of GDP)

Figure 3F: U.S. United

Korea, Rep. GDP growth (annual %)

States GDP growth (annual %)

Korea, Rep. Military expenditur e (% of GDP)

United States Military expenditur e (% of GDP)

155

3.0 LITERATURE REVIEW In 1973, Benoit became the first economist to address through causal analysis, the relationship between defense spending and economic growth. These early studies, and most popularly by Benoit (1973, 1978), suggested a causal relationship between defense spending and economic growth. This started when Benoit (1973) suggested that higher defense spending was more likely to not be the effect of economic growth, but actually the cause. Benoit (1973, 1978) also exhibits how defense spending stimulates growth by increasing aggregate demand, which ostensibly leads to a higher utility of the capital stock, reduced resource costs, and a higher level of employment. Other studies followed Benoit’s (1973, 1978) work but found different results. Deger and Sen (1983), Faini, Annez, and Taylor (1984), and Leontief and Dutchin (1983) found evidence to reject Benoit’s (1973, 1978) suggestion that defense spending stimulates economic growth. Smith and Smith (1980) found no strong and systematic relationship between defense spending and economic growth, and Biswas and Ram (1986) found no statistically significant relationship between military expenditure and economic growth in middle-income and low-income countries. Some studies have even suggested a negative relationship between economic growth and defense spending. Deger and Smith (1983) and Fredericksen and Looney (1983) each exhibit this type of relationship in a majority of all of their tests of developed countries. Deger and Smith (1983) claim that military expenditures impedes economic growth. Finally, Chowdhury (1991) concluded that we cannot generalize the relationship between defense spending and growth across countries. The results of the study implied that the conclusion about whether defense spending helps or impedes economic growth does not solely depend on the development stage in which a particular country is. Chowdhury’s study inspired this paper, because at a time in the United States when defense spending has been on the rise and is at a very high level relative to other historical amounts. Deger (1986) has suggested that defense spending helps economic growth through a “spinoff effect.” However, Deger (1966) also suggests, along with Chowdhury (1991), that defense spending can hinder economic growth in a number of ways, such as diverting available resources from domestic capital accumulation (reducing national savings available for investment leading to reduced growth), and diverting funds from other national expenditures such as healthcare and 156

education. This also takes available funds and labor (as the size of the military increases) away from potentially productive capitalist ventures. Smith and Georgian (1983) summarize the ultimate conclusion about the relationship between defense spending and economic growth: “… it depends on the nature of the expenditure, the prevailing circumstances, and the concurrent government policies” (15). 4.0 DATA AND EMPIRICAL METHODOLOGY 4.1 Empirical methodology and results GC testing, developed in Granger (1969, 1980), is a technique for determining the causal relationship between time series data, to see whether one time series can be useful in forecasting another. The GC method regresses the variable X on lagged values of X (Xt-i). When the appropriate lag interval for X is significant, regressions for the variable Y are performed. In turn, serial correlation is eliminated to leave only correlation between the pair of variables. Y Grangercauses X when the coefficients of the lagged values of Y are significant. The same goes for X Granger-causing Y when the coefficients of the lagged values of X are significant after one regresses Y on lagged values of Y. The four possible results of a GC test are: no causality, X Granger-causes Y, Y Granger-causes X, and X and Y Granger-cause each other. Sims’ (1972, 1980) work promoted GC testing and helped to operationalize the test. Many econometricians discredit the GC test, because it does not imply true causality, however, it gives important insight into whether the current value of one variable influences the future value of another variable, when that other variable’s past is considered. The following steps are the order in which to conduct a GC test: 1. Test for the presence of a unit root using the Augmented Dickey-Fuller Test (ADF). 2. Difference the data in the presence of unit root and conduct the ADF test again on the differenced data. 3. Exclude if one series is non-stationary and the other is stationary. 4. Estimate co-integration using the same order of integrated variables using the Johansen test. 5. Based on the co-integration results, use VAR or VEC to test causality. The first step in the process, testing for unit roots indication non-stationary data, uses the ADF test. The ADF test uses the following regression equations:

(1) ∆Xt = βXt-1 + Σ(pi=2)ηi ∆Xt-1+i + εt 157

(2) ∆Xt = α0 + βXt-1 + Σ(pi=2)ηi ∆Xt-1+i + εt (3) ∆Xt = α0 + βXt-1 + δt + Σ(pi=2)ηi ∆Xt-1+i + εt where ∆ is the first difference operator, X is logGDP (or logDS), p is the maximum lag error, ε is the stationary random error, and t is the time. Equation 1 tests for random walk, however one would use equation 3 which includes both the drift term (intercept, α0) and linear time trend (δt), and use the other equations when the test fails to reject the null hypothesis that β=0 (unit root is present). If β is negative and statistically significant, then the time series has no unit root or is stationary. In this regression, an optimal number of 4 lags were included on the first level based on Schwartz Bayesian Criterion (BSC). Table 1 shows the ADF test results on the level and on first differenced data. The results indicate that they are all non-stationary, and that the null hypothesis can be rejected at the 1% level for all but LogDS, implying that all of the variables are stationary after converting the series through first differencing. The second step estimates co-integration using the same order of integrated variables.11 Each of the two variables that is I(1) needs to be tested for co-integration.12 In order to test for cointegration, the Johansen method was used. The Johansen method uses the following regressions: (3) λtrace (r) = -T Σ(ni=r+1) ln(1-λi) (4) λmax (r, r+1) = -T ln(1-λr+1) where λ is the estimated values of characteristic root or the eigenvalues and T is the number of usable observations. For λtrace statistics, the null hypothesis is that against the general alternative, the number of co-integration vectors is less than or equal to r. For the λmax statistics, the null hypothesis is the number of co-integration vectors, r, against the alternative co-integration vectors, r+1, where if r=0, the alternative is r=1. The distribution of the statistics depends on both the number of nonstationary components under the null hypothesis and whether a constant or drift term is included in the co-integrating vector.

A series is integrated of order (d) or I(d) if after being differenced d times it becomes stationary. Such is the case in this test. 12 In the case where Xt and Yt are both I(d) and linear combination exists, Zt = aXt + bYt, and characteristic roots (c<0), Xt and Yt are co-integrated.

158

The results of the Johansen test are included under Table 2. If the rank of r is 0, the variables are not co-integrated. The null hypothesis of no co-integration was rejected at a 1% critical value level. Tthe GC test was then used to test causality. The VEC model. In this study the lag length of 3 was automatically chosen by Eviews as the optimal lag length for the annual data. The following criteria was used to determine the lag length of 3: SBC = T log | Σ | + N log(T) where | Σ | = determination of the variance/covariance matrix of the residuals and N = total number of parameters estimated in all equations. The GC test (Table 3) concluded that neither variable Granger-causes the other in any of the 24 of 31 countries for which data was available and which unit root did not exist (i.e. nonstationary countries). 4.2 Data The variables tested were log(GDP growth rate) and log(Military expenditure as a percentage of GDP). The data was obtained from the World Bank’s World Development Indicators Online, and includes the years from 1989-2006. Table 1: Results of ADF test Ho: Unit root vs. H1: No unit root

Afghanistan

Albania

Australia

Azerbaijan

Variable

ADF Prob.

Lag

D(GDP Growth Rate)

D(Military Expenditure (as a % of GDP)) I

D(GDP Growth Rate)

.0002***

D(Military Expenditure (as a % of GDP)) .0573***

D(GDP Growth Rate)

.0000***

D(Military Expenditure (as a % of GDP)) .0572***

D(GDP Growth Rate)

.0148***

D(Military Expenditure (as a % of GDP)) .0691*** Bulgaria

Columbia

D(GDP Growth Rate)

.0010**

D(Military Expenditure (as a % of GDP)) .0275**

D(GDP Growth Rate)

.0000***

159

Czech Republic

Denmark

El Salvator

Eritrea

Estonia

Ethiopia

Georgia

Hungary

Italy

Japan

Latvia

Lithuania

Macedonia

Netherlands

Nicaragua

D(Military Expenditure (as a % of GDP)) .0011***

D(GDP Growth Rate)

.1123****

D(Military Expenditure (as a % of GDP)) .1548****

D(GDP Growth Rate)

.0000**

D(Military Expenditure (as a % of GDP)) .0002**

D(GDP Growth Rate)

.0000**

D(Military Expenditure (as a % of GDP)) .5059***

D(GDP Growth Rate)

.0001**

D(Military Expenditure (as a % of GDP)) .0484***

D(GDP Growth Rate)

.0001**

D(Military Expenditure (as a % of GDP)) .0988***

D(GDP Growth Rate)

.0000**

D(Military Expenditure (as a % of GDP)) .0194***

D(GDP Growth Rate)

.0001***

D(Military Expenditure (as a % of GDP)) .0640***

D(GDP Growth Rate)

.0000***

D(Military Expenditure (as a % of GDP)) .0031***

D(GDP Growth Rate)

.0000***

D(Military Expenditure (as a % of GDP)) .0480***

D(GDP Growth Rate)

.0000*

D(Military Expenditure (as a % of GDP)) .1832*

D(GDP Growth Rate)

.0000*

D(Military Expenditure (as a % of GDP)) .2618*

D(GDP Growth Rate)

.0021***

D(Military Expenditure (as a % of GDP)) .0140***

D(GDP Growth Rate)

.0039**

D(Military Expenditure (as a % of GDP)) .0132**

D(GDP Growth Rate)

.0000*

D(Military Expenditure (as a % of GDP)) .4357*

D(GDP Growth Rate)

.0000**

160

Philippines

Poland

Romania

D(Military Expenditure (as a % of GDP)) .0001**

D(GDP Growth Rate)

.0000**

D(Military Expenditure (as a % of GDP)) .0098**

D(GDP Growth Rate)

.0005***

D(Military Expenditure (as a % of GDP)) .0453***

D(GDP Growth Rate)

.0032***

D(Military Expenditure (as a % of GDP)) .0731***

Slovak Republic D(GDP Growth Rate)

South Korea

Spain

Turkey

U.K.

United States

Uzbekistan

.0060**

D(Military Expenditure (as a % of GDP)) .0014**

D(GDP Growth Rate)

.0000***

D(Military Expenditure (as a % of GDP)) .0960***

D(GDP Growth Rate)

.0000*

D(Military Expenditure (as a % of GDP)) .1035*

D(GDP Growth Rate)

.0000**

D(Military Expenditure (as a % of GDP)) .0972***

D(GDP Growth Rate)

.0000*

D(Military Expenditure (as a % of GDP)) .1790*

D(GDP Growth Rate)

.0000**

D(Military Expenditure (as a % of GDP)) .0229**

D(GDP Growth Rate)

D(Military Expenditure (as a % of GDP)) I

ADF regression equation: ∆Xt = α0 + βXt-1 + δt + Σ(pi=2)ηi ∆Xt-1+i + εt *** denotes significant at 5% critical value ** denotes significant at 1% critical value * has unit root LogGDP and LogDS are series in level D(LogGDP) and D(LogDS) are first differenced series

161

Table 2: Results of the co-integration test using the Johansen method λtrace λmax Afghanistan

Albania

32.37***

21.38***

Australia

19.97***

18.45***

Azerbaijan

43.21***

42.49**

Bulgaria

26.28**

17.36***

Columbia

12.00**

9.40***

Czech Republic

Denmark

8.43***

8.27**

El Salvator

29.59***

18.27**

Eritrea

23.02***

18.59**

Estonia

14.86***

11.82**

Ethiopia

22.32**

13.10***

Georgia

16.63**

10.79***

Hungary

23.35***

16.82***

Italy

17.75***

9.13***

Japan

Latvia

Lithuania

21.67***

15.39***

Macedonia

19.63***

15.48***

Netherlands

Nicaragua

34.42***

23.87**

Philippines

15.02***

14.67**

Poland

23.17**

17.71**

Romania

14.58***

10.21***

Slovak Republic 19.44**

18.25**

South Korea

18.81**

13.69***

Spain

Turkey

20.95***

16.69**

162

U.K.

20.66***

16.94***

United States

19.73***

15.84***

Uzbekistan

** denotes not significant at 5% critical value *** denotes significant at 5% critical value Table 3: Results of the Granger-causality test Country

Military expenditure→GDP Growth

GDP Growth→Military expenditure

Afghanistan

Albania

Australia

Azerbaijan

Bulgaria

Columbia

Czech Republic

Denmark

El Salvator

Eritrea

Estonia

Ethiopia

Italy

Japan

Georgia

Hungary

Latvia

Lithuania

Macedonia

Netherlands

Nicaragua

Philippines

Poland

No 163

Romania

Slovak Republic No

South Korea

Spain

Turkey

U.K.

United States

Uzbekistan

NA denotes not enough data available UR denotes unit root exists and cannot be tested 5.0 EMPIRICAL ANALYSIS The hypothesis in the introduction of this study tested the direction of causality between economic growth and military expenditure as a percentage of GDP in the member-nations of the Coalition of the Willing. Most previous studies were performed in the 1970s and 80s with older data, and used OLS estimation and cross-sectional data, under the assumption that defense spending causes economic growth, without consideration of the reverse. Most studies have used defense spending as one of the variables, and not the log of military expenditure as a percentage of GDP as this study has done. The ADF test indicates that both logGDP and logDS have unit roots in the level data. In the presence of unit roots, the variables needed to be first-differenced in order to make the series stationary because without differencing the data, the causality test would lead to misspecification. The Johansen test (in Table 2) co-integrated using the VEC model. The resulting GC test showed no Granger-causality in either direction between the two variables in any of the 24 (out of 31) countries with GC-testable data. 6.0 CONCLUSION Previous studies have used different testing techniques (OLS, GC, some cross-sectional, etc.) to analyze the relationship between defense spending and economic growth. This paper aimed to employ GC testing between the log of military expenditure as a percentage of GDP and the GDP growth rate of the 31 member nations of the Coalition of the Willing to test for a causal relationship, amidst growth in defense spending, an economy on the brink of recession, and claims that the war in Iraq will help the economy. 164

In summary, this study does not support Benoit’s (1973) early claim that defense spending causes economic growth. Furthermore, it does not help to support later studies suggestions that defense spending hinders economic growth. Based on the empirical results using GC testing, the findings of the study agree with the findings of Biswas and Ram (1986) and Chowdhury (1991) that there is no statistically significant evidence to support the idea that defense spending causes economic growth. That is to say, there is no statistically significant evidence to support the claim that defense spending causes economic growth in any of the countries involved in the Iraq war. For this time period, even though there were no large-scale significant wars, there have been many minor conflicts and large investments in military technology. Governments spending money abroad in foreign conflicts and wars would not necessarily typically benefit domestic GDP, however investments in technology and research typically should. There appears to be other exogenous factors that have a larger effect on GDP than military expenditure, or it may be that the results take longer to materialize and that countries may see benefits further down the road. Either way, countries should be weary of where they spend their money by maintaining sound fiscal policy and not relying on one area such as military expenditure to help stimulate an economy.

165

BIBLIOGRAPHY Benoit, E., 1973.Defense and economic growth in developing countries. Boston: D.C. Heath & Co. Benoit, E., 1978. Growth and defense in developing countries. Economic Development and Cultural Change 26,271-80. Biswas, B., and Ram., R. 1986. Military expenditures and economic growth in less developed countries: An augmented model and further evidence. Evidence Development and Cultural Change 34, 36 1-72. Canning, D., Amaral, L.A.N., Lee, Y., Meyer, M., and Stanley, H.E., 1998. Scaling the volatility of GDP growth rates. Economics Letters 60, 335-341. <http://polymer.bu.edu/hes/articles/calms98.pdf>. Chowdhury, A., 1991. A Causal Analysis of Defense Spending and Economic Growth. The Journal of Conflict Resolution, Vol. 35, No. 1: 80-97. Deger, S. 1986. Economic development and defense expenditure. Economic Development and Cultural Change 34, 179-96. Deger, S., and Sen, S., 1983. Military expenditure, spin-off and economic development. Journal of Development Economics 13, 67-43. Deger, S., and Smith, R., 1983. Military expenditure and growth in the less developed countries. Journal of Conflict Resolution 27 (June), 335-53. Faini, R., Annez, P., and Taylor, L., 1984. Defense spending, economic structure and growth: Evidence among countries and over time. Economic Development and Cultural Change 32, 487-98. Francis, D., 2003. War’s Mixed Impact on a Reviving Economy. The Christian Science Monitor. Sept. 16. Fredericksen, P. and Looney., R., 1982. Defense expenditures and economic growth in developing countries: Some further empirical evidence. Journal of Economic Development 7, 113-24. Fredericksen, P., 1983. Defense expenditures and economic growth in developing countries. Armed Forces and Society 9, 633-45. Grobar, L. and Porter, R., 1989. Benoit Revisited: Defense Spending and Economic Growth in LDCs. The Journal of Conflict Resolution, Vol. 33, No. 2: 318-345. 166

Heo, UK, 1998. Modeling the Defense-Growth Relationship around the Globe. The Journal of Conflict Resolution, Vol. 42, No. 5: 637-657. Joerding, W., 1986. Economic growth and defense spending: Granger causality. Journal of Development Economics 21,35-40. Leonhardt, D., 2007. "What $1.2 Trillion Can Buy." The New York Times 17 Jan. Leontief, W., and Dutchin, F., 1983. Military spending: Facts and figures New York: Oxford University Press. Lim, D., 1983. Another look at growth and defense in less developed countries. Economic Development and Cultural Change 31, 377-84. Looney, R and Fredericksen, P. 1986. Defense Expenditures, External Public Debt and Growth in Developing Countries. Journal of Peace Research, Vol. 23, No. 4: 329-338. Mintz, A. and Stevenson, R., 1995. Defense Expenditures, Economic Growth, and the “Peace Dividend”: A Longitudinal Analysis of 103 Countries. The Journal of Conflict Resolution, Vol. 39, No. 2: 283-305. Mohan, R., 2006. Causal Relationship Between Savings and Economic Growth in Countries with Different Income Levels. Economics Bulletin, Vol. 5, No. 3: 1-12. Reuters, 2007. Ex-U.S. Treasury Head Summers: Recession Likely. CNBC 26 Nov. 22 Mar. 2008 <http://www.cnbc.com/id/21971632>. Reuters, 2008. U.S. Headed Into Recession, Goldman Sachs Says. International Herald Tribune 9 Jan. 22 Mar. 2008 <http://www.iht.com/articles/2008/01/09/business/goldman.php>. Rothschild, K., 1977. Military expenditure, exports and growth. Kyklos 26, 804-13. Schifferes, S. 2003. US names ‘coalition of the willing’. BBC News. <http://news.bbc.co.uk/2/hi/americas/2862343.stm>. Sims, C., 1972. Money, income and causality. American Economic Review 62,540-42. Shanker, T., 2008. Proposed Military Spending is Highest Since WWII. The New York Times 4 Feb. 22 Mar. 2008 <http://www.nytimes.com/2008/02/04/washington/04military.html>. Smith, D., and Smith, R., 1980. Military expenditure, resources and development. Birbeck College, London. Mimeo. Smith, R. and Georgion, G., 1983. Assessing the effect of military expenditure on OECD economies: A survey. Arms Control 4,3-15.

167

Stiglitz, J., and Bilmes, L., 2008. The Three Trillion Dollar War: the True Cost of the Iraq Conflict. W. W. Norton. "Today Exclusive." The Today Show. NBC. President and First Lady Speak Out. 18 Feb. 2008. 22 Mar. 2008 <http://www.youtube.com/watch?v=lIbdnM8Ts88>. United States. Office of Management and Budget. Executive Office of the President of the United States. Historical Tables - Budget of the United States Government. 25 Mar. 2008 <http://www.whitehouse.gov/omb/budget/fy2009/pdf/hist.pdf>. Wingfield, B., 2008. Key Forecaster Says U.S. in Recession. Forbes 14 Mar. 2008. 22 Mar. 2008 <http://www.forbes.com/2008/03/14/economy-nber-feldstein-biz-beltwaycx_bw_0314economy.html?feed=rss_news>. Starr, H., Hoole, F., Hart, J., and Freeman, J., 1984, The Relationship between Defense Spending and Inflation. The Journal of Conflict Resolution, Vol. 28, No. 1: 103-122.

168

U.S. Current Account: Why Is It Increasingly Negative?

Joshua Champagnea

Abstract: This paper investigates the reasons for the increasingly negative United States current account. The study incorporates information into a multivariate linear regression model to examine the influence of various economic indicators on the U.S. current account. The paper focuses more so on which variables create an increase in the current account and which variables cause deterioration and why the overall value of the current account is continually becoming more negative. The results show that the U.S. Current Account is negative because there is not enough government investment, savings, and private savings, along with a negative fiscal policy, combined with an increase in private investment and domestic GDP. JEL Classification: E22, E60, E62 Keywords: Current Account, Account Deficit a

Bryant University, 1150 Douglas Pike, Smithfield, RI 02917. Phone: (401) 744-1153. Email:

jchampag@bryant.edu

______________________________ The author thanks the BLS, BEA, GPO Access, and Misery Index for providing data and gratefully acknowledges the help/guidance from Professor Ramesh Mohan.

169

1.0 Introduction The U.S. current account is a major part of the economy and oftentimes dictates fiscal and monetary policy in order to create an overall surplus. The current account is made up of three parts: the balance of trade plus net factor income from abroad plus net unilateral transfers from abroad. The biggest part of the current account is the balance of trade, which is exports minus imports. Ever since the 1970s the balance of trade has become increasingly negative, yet the current account has been occasionally positive, mostly in the 1970s and 1980s, and also once in 1991. Therefore there are many other factors that can have a strong impact on the current account. The current account has been a hot topic as of late, especially since it has been in a deficit since 1992, setting new record lows every year since 1998. Recently, during the Bush administration, where the current account deficit ballooned to over $811 billion, many economists and citizens alike have come to realize how important this deficit has become. With a slumping economy and high unemployment rate, among other things, economists have come to question how dangerous this deficit is to our economy in the long run. Overall, there is very limited research that focuses on such a wide comprehension of variables and effects. Much of the research in the area focuses on the current account along with the budget deficit (known as the “twin deficits”) and whether the U.S. can sustain these deficits and continue to grow. This paper contributes to the literature on the subject in three respects. First, this paper is the first of its kind to include such a wide variety of variables over such a large time span. Second, this paper finally brings quantifiable results as to the affect of the tested variables on the current account. Lastly, this paper brings up to date the affects of the variables on the current account by using the most recent data available (2006). The rest of the paper is organized as follows: Section two gives a brief literature review. Section three outlines the empirical model. Data and estimation methodology are discussed in section four. Finally, section five presents and discusses the empirical results. This is followed by a conclusion in section six

170

2.0 Current Account Trends Current Account 100

Current Account (billions US)

0 -100

9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47

-200 -300 -400

-500 -600 -700 -800 -900 Years

Source: Bureau of Economic Analysis

The graph above shows the current account since 1960 in billions of US dollars, with 1 being 1960, and 47 being 2006. The negativity of the account becomes quite clear, beginning in 1992, and dropping sharply until the present day, reaching a low deficit of $811 billion. This has become quite problematic for the U.S. economy, as it has entered a slump in the last few years or so. Unemployment is high, the dollar is depreciating, and oil prices are skyrocketing. These are just a few problems on top of and/or related to the enormous current account deficit. The good news is that by the end of 2007, the current account deficit was only $172.9 billion (BEA), a sign that the current account is heading in the right direction. However, that is still a large deficit, and definitely something that the U.S. has to attempt to control through various governmental policies in the near future. 3.0 Literature Review One of the major areas of the current account that is researched is what is known as the “twin deficits.” Studies on the “twin deficits” which is the theoretical idea that the budget account and current account should fluctuate together have produced surprising results. In Kim and Roubini (2004), their results lend them to conclude that in the short run, a budget deficit can

171

actually lead to an improvement in the current account. This can be seen in the divergence of the two accounts from 1987 to 2001. In recent years, more research on changes in the fiscal policy and its effect on the trade balance and the current account have been conducted. In Baxter (2005) she finds that an increase of the budget deficit equal to about 1% of GDP leads to a decrease of the current account by about 0.5% GDP. Further research conducted by Erceg et al (2005) states that increased government spending and tax rate cuts do not drastically affect the trade balance and therefore the current account is relatively unchanged, meaning that the budge deficit has an even more modest affect on the current account than found in Baxter (2005). All of the above papers were reviewed in Cavallo (2005) where she inserts her own research into the idea of “twin deficits.” Her research deals with government expenditure on nontraded labor services which include, “for example, general public service, national defense, public order and safety, health, education, and others.”13 Her findings indicate that an increase in government expenditures equal to 1% of GDP lead to a mere 0.05% of GDP reduction in the current account. All of these papers combined lead to the idea that the current budget deficit is not affecting the current account as much as economists thought, and perhaps there are other causes for the increasing negativity of the current account. Lastly, in Holman (2001), she concludes that much of the current account deficit is driven by two factors: a surge in U.S. productivity coupled with the stock market boom of the 1990s which led to an increase in consumer spending, both of which caused the deficit to widen further. However, since both productivity and the stock market (thus consumer spending) have decreased since the article was written there must be another factor that is drastically affecting the widening current account deficit. 4.0 Definition of Variables CAt = β0 + β1FPt + β2NPSt + β3NGSt + β4PIt + β5GIt + β6PGt + β7TMCURt + β8GDPt + β9INFt + ε

(1)

Quote taken from Cavallo’s 2005 FRBSF Economic Letter in which she talks about her research, showing how a

large increase in expenditure on labor services leads to a very small deterioration in the current CA.

172

This is the overall model used within this paper. Other models included PG and consumer spending (Holman 2001), while others used FP (Erceg et al 2005). While many previous papers have used one or more of these variables in determining their effects on the current account, none of them have comprised such a comprehensive model to include numerous variables that, according to economic theory, should have a significant impact on the overall value on the current account. CAt is the U.S. current account at year t. It is comprised of the balance of trade plus net factor income from abroad plus net unilateral transfers from abroad. The definition of the current account in this paper is consistent with the IMF which states that the CA is “The record of all transactions in the balance of payments covering the exports and imports of goods and services, payments of income, and current transfers between residents of a country and nonresidents” (IMF 2006). All CA figures were obtained from the Bureau of Economic Analysis (BEA) and are in billions of U.S. dollars. Independent variables consist of nine variables obtained from various sources. Appendix A and B provide data source, acronyms, descriptions, expected signs, and justifications for using the variables. All variables were also obtained yearly, so each is listed at year t. The first variable, FPt, is the fiscal policy of the government, which is government expenditure plus tax revenue. The second variable, NPSt, is the net private savings of the citizens in the United States. The next variable is NGSt, which is the net government savings. These two savings combined produce the overall net savings of the U.S. The fourth variable is PIt, which is the private investment, followed by the next variable GIt, which is the government investment. These two variables also combine to produce net investment in the U.S. All five of these variables were obtained from the 2008 Economic Report of the President and are in billions of U.S. dollars. The sixth variable is productivity growth, PGt, which is the overall growth in productivity of the U.S. economy. This data was obtained from the Bureau of Labor Statistics (BLS) and is listed as a percent (%). The next variable is total manufacturing capacity utilization rate, TMCURt. This is the capacity at which the manufacturing firms in the United States are operating. This data was also obtained from the 2008 Economic Report of the President and is listed as a percent (%).

173

The eighth variable is gross domestic product or GDPt for short. This is the total market value of all final goods and services produced within a given country in a given period of time (usually a calendar year). This data was obtained through the 2008 Economic Report of the President and is listed in billions of U.S. dollars. Lastly, the ninth variable is INFt, which is the annual inflation rate in the United States. Data was obtained from the Misery Index, an online database, and is listed as a percent (%). 5.0 Data The study uses annual data from 1960 to 2006. Data were obtained from the Bureau of Economic Analysis (BEA), Bureau of Labor Statistic (BLS), the 2008 Economic Report of the President (GPO Access), and the Misery Index websites. Summary statistics for the data are provided in Table 1 on the next page. Table 1: Summary Statistics Variable

Obs.

Mean

Std. Dev.

Min

Max

Current Account

-127.1762

30.39713

-811.477

18.116

Fiscal Policy

-94.82766

19.35204

-412.7

236.2

Net Private Savings

292.7085

24.25811

44.3

551.1

Net Government Savings

-80.82341

19.0658

-392.5

239.4

Private Investment

451.5851

57.45935

40.5

1357

Government Investment

99.3766

11.02342

267.7

Productivity Growth

2.255319

.2117022

-1.6

4.6

Total Manufacturing Capacity Utilization Rate Gross Domestic Product

80.8234

.6445976

91.1

4607.56

550.567

526.4

13194.7

Inflation

4.241064

.4203198

1.07

13.58

174

6.0 Empirical Results The primary objective of this study was to find the determinants of the current account deficit. The means and standard deviations, as well as the maximums and minimums, of the variables used in this study are given in Table 1. The results of the nine different variables’ effects on the U.S. current account are shown in table 2. In this regression, using a multivariate linear model, the current account in billions of U.S. dollars was regressed against various independent variables. After running a correlation test, different variables which were highly correlated were dropped in different regressions, so as to produce accurate and unbiased results. Of the nine variables, five were statistically significant at a 1% level, two were statistically significant at the 5% level, one was statistically significant at the 10% level, and one was statistically insignificant. One of the most surprising finds was the fact that productivity growth and TMCUR were both relatively insignificant, with PG only being significant at the 10% level in one regression, which contradicts the theory and results of Holman (2001). In two of the regressions the results were negative, and in the other regression PG was positive and barely significant. One of the two reasons she believed that the CA deficit was widening was PG. However, she focused her data on the economic boom of the 1990s, and while it is true that in the short run PG will serve to widen the CA deficit, in the long run, or at least the in the 47 years of data used in this study, PG becomes statistically insignificant as a factor the in the increasingly negative CA. Strangely enough, the sign of the coefficient in front of PG is positive in some cases, which would indicate that as PG increased, the CA would increase, which goes against Holman (2001) and economic theory in general. Even more interesting is the fact that the one time PG was statistically significant was the regression in which it had a positive value. Unfortunately while this variable is statistically insignificant, the results may hint that while in the short run, PG widens the CA deficit, in the long run PG may help increase CA in some indirect way. This may be a case similar to the “twin deficit” divergence, where oppositely, in the short run, an increased budget deficit actually helps to increase the CA. More research into the matter is most likely needed on the long-run vs. short-run affects of PG on the current account. Directly related to this is the TMCUR, whose negative coefficient, while expected in accordance with economic theory, contradicts the above results. As firms use more of their capacity, they produce more, which could suggest an increase in productivity. This increase in 175

productivity should increase the value of working capital, thus private investment, thus decreasing the CA. This would tend to agree with economic theory as well as Holman (2001). However, due to the extremely low t-statistic (-0.18), the results of this variable are better off being ignored. Of the two variables that were statistically significant at the 5% level, both agreed with the expected sign, but not in every regression due to various correlations. First, inflation, which produced a positive coefficient in only one of the three regressions (but when negative only slightly negative, and when positive, it was statistically significant), as expected, because one way to decrease a CA deficit is to increase inflation, which can be done by decreasing interest rates, which in turn would vary the exchange rates, making domestic products cheaper to foreign buyers, increasing exports and the balance of trade. This ultimately would serve to increase the CA, although not as much as previously thought according to the research of Erceg et al (2005). Next is PI, which as expected was also negative, and serves to widen the CA deficit. NGS was produced a positive coefficient because as NGS goes up, net savings should increase, and subsequently so should the CA. Overall this was one of the more statistically significant variables and indicates that the government should not spend money foolishly, such as on unnecessary wars in Iraq, which do not provide a measurable net benefit to the people of the U.S, and should save this money. These results correlate directly with the fiscal policy, where if the government spends less than the tax revenues that it raises (ultimately causing NGS to increase) then the CA would also increase. In line with this is GI, which was the most significant variable with the highest coefficient, indicating that government investment, in this study, appears to be the leading determinant in the CA. This shows that as government investment increases, the CA should drastically increase overall. This seems most obvious, especially in the long run, because government investments should lead to interest and income receipts, causing exponential benefits. Second to GI was NPS, which was just as significant, showing that as net private savings increases, so should the CA. As Dick Cooper claims in his article, as reviewed in the Journal of Policy Modeling (2006) “Americans save too little,”14 which is definitely true in that

“Americans save too little” is one of the three propositions Dick Cooper uses in his article “Living with global

imbalances: A contrarian view” as reviewed by the Journal of Policy Modeling

176

every year since 1960, NPS has been decreasing, yet since 1992 the CA has been decreasing. Lastly, GDP, which produced a negative coefficient as expected in two out of the three regressions (the positive value was insignificant) because a decrease in domestic GDP reduces domestic demand for foreign goods, lowering imports without affecting exports. Since GDP has also been increasing every year since 1960, just the opposite holds true. Table 2: Regression Results for the Current Account Variables

Reg I (Coefficient)

Fiscal Policy

0.339***

Net Private Savings

0.823***

Net Government Savings Private Investment Government Investment Productivity Growth

Reg II (Coefficient)

Reg III (Coefficient) 0.036

0.324*** -1.310** 12.59*** -3.309

14.276*

-14.113

-3.034

-0.831

-1.280

0.060

-0.394***

-0.049***

-0.091

8.492**

-0.364

0.9427

0.9275

0.7919

Adjusted R2

0.9325

0.9166

0.7665

F-Value

91.72***

85.28***

31.21***

No. of obs.

Total Manufacturing Capacity Utilization Rate Gross Domestic Product Inflation 2

Note ***, **, and * denotes significance at the 1%, 5%, and 10% respectively 7.0 Conclusions This paper contributes to the literature of the U.S. Current Accounts. Using existing data for the 1960-2006 period for the current account, a multivariate linear model was used to empirically estimate the regression. The regression estimates indicate that in increase in the CA is associated with an increase in fiscal policy, net private savings, government investment, net government savings and inflation rates. A decrease in the CA is associated with an increase in private investment and gross domestic product.

177

Overall, given that PG has curtailed in recent years, and that FP does not drastically affect the trade balance (which is the largest part of the CA), the real economic indicators that affect the CA have to be NPS and GI, followed up with an overall improvement in the FP and NGS. First and foremost the government needs to drastically increase their investment expenditures, as it pales in comparison to private investment (GI is roughly 1/5th of what PI is). Secondly, we as Americans need to improve our savings levels, possibly through reducing consumer spending. Third, the government needs to stop spending, and increase their savings immensely. Lastly, while FP has little effect on the trade balance, it also has little effect on the deterioration of the CA if spent in the right areas, such as nontraded labor services. Obviously it is better if the government can spend less than the tax revenues it raises, and create a positive FP, which it has not had since 2001. However, if the government must spend its money, then spend it wisely and efficiently, and in places that provide a high return and net benefit to the U.S. as a whole. In the end, while the 2007 U.S. Current Account deficit decreased drastically from 2006 ($811.5 billion) to $172.9 billion, indicating that the U.S. is heading in the right direction, many more improvements to American policies are needed to bring the country back into a positive current account.

178

A: Variable Description and Data Source

Acronym

Description

Data source

Current Account in billions of dollars

Bureau of Economic Analysis

Fiscal Policy in billions of dollars

Economic Report of the President 2008

Appendix B:

NPS

Net Private Savings in billions of dollars

Economic Report of the President 2008

Net Government Savings in billions of dollars

Economic Report of the President 2008

Private Investment in billions of dollars

Economic Report of the President 2008

Government Investment in billions of dollars

Economic Report of the President 2008

Productivity Growth as a percent (%)

Bureau of Labor Statistics

TMCUR

Total Manufacturing Capacity Utilization Rate as a percent out of 100%

Economic Report of the President 2008

GPD

Gross Domestic Product in billions of dollars

Economic Report of the President 2008

INF

Annual Inflation Rate as a percent (%)

Miseryindex.com

NGS

Variables and Expected Signs

179

Acronym

Variable Description

What it captures

Expected sign

Fiscal policy

Government spending plus tax revenues

NPS

Net private savings

The overall savings of people in the U.S.

NGS

Net government savings

The overall savings of the government

Private investment

The amount of money people in the U.S. invest

Government investment

The amount of money the government invests

Productivity growth

TMCUR

Total manufacturing capacity utilization rate

GDP

Gross domestic product

INF

Annual inflation rate

Overall growth in productivity of the U.S. economy

The capacity rate at which manufacturing firms are operating

The total market value of all goods and services produced in the U.S.

The annual rate of inflation in the U.S.

Bibliography

180

Baxter, Marianne. 1995. “International Trade and Business Cycles.” In Handbook of International Economics Vol. 3, eds. Gene M. Grossman and Kenneth Rogoff, pp. 1801– 1864.Amsterdam: North-Holland. Bureau of Economic Analysis, [online data file], http://www.bea.doc.gov/bea/di/home/directinv.htm. Bureau of Labor Statistics, [online data file], http://www.bls.gov/. Cavallo, Michele. 2005. “Government Consumption Expenditures and the Current Account.” FRBSF Working Paper 2005-03. http://www.frbsf.org/ publications/economics/papers/2005/wp05-03bk.pdf. Cavallo, Michele. 2005. “Understanding the Twin Deficits: New Approaches, New Results.” FRBSF Working Paper 2005-16. http://www.frbsf.org/publications/economics/letter/2005/el2005-16.pdf. Economic Report of the President 2008, [online data file], http://www.gpoaccess.gov/eop/2008/2008_erp.pdf. Erceg, Christopher J., Luca Guerrieri, and Christopher Gust. 2005. “Expansionary Fiscal Shocks and the Trade Deficit.” International Finance Discussion Paper 825, Federal Reserve Board. http://www.federalreserve.gov/pubs/ifdp/2005/825/ifdp825.pdf. Holman, Jill A. 2001. “Is The Large U.S. Current Account Deficit Sustainable?” Economic Review First Quarter 2001, pp. 5-23. Federal Reserve Bank of Kansas City. http://www.kansascityfed.org/publicat/econrev/PDF/1q01holm.pdf. International Monetary Fund, [online data file], http://www.imf.org/external/.

181

Kim, Soyoung, and Nouriel Roubini. 2004. “Twin Deficits or Twin Divergence? Fiscal Policy, Current Account and Real Exchange Rate in the U.S.” Mimeo, Korea University and New York University. http://econ.korea.ac.kr/prof/sykim/files/fiscalus9.pdf. Misery Index, [online data file], http://www.miseryindex.us/iRbyyear.asp?StartYear=1960&EndYear=2006. Salvatore, Dominick. 2006. “Twin deficits, growth and stability of the US economy: Editor’s Introduction.” Journal of Policy Modeling Vol. 28-6, pp. 603-604. http://ideas.repec.org/a/eee/jpolmo/v28y2006i6p603-604.html.

182

The Social and Behavioral Factors that Affect Obesity in OECD countries

Sanjana Desai15 Abstract: This paper investigates the factors that affect obesity in the population of the member countries of the Organization of Economic Co-operation and Development (OECD). This study incorporates existing information of the most likely variables that would influence obesity in these member countries for a decade (1990-2000). The study looks at contributing factors that are both innate and acquired, such as Alcohol Consumption, Low-birth weight, sugar consumption, smoking (consumption of tobacco), and protein intake and public and private expenditure on education. Using macro-level data from various sources, the results generally suggest that factors like alcohol consumption, low-birth weight and sugar consumption add to obesity whereas smoking and protein intake negatively (favorably) impact obesity.

Key Words: Obesity, social and behavioral factors, JEL classification: I1, J1,

Department of Economics, Bryant University, 1150 Douglas Pike, Smithfield, RI02917. Phone: (401) 232 8437 Email: sdesai@bryant.edu The author gratefully acknowledges the help/guidance from Dr. Ramesh Mohan

183

1.0 INTRODUCTION: Obesity is one of the greatest public health challenges of the 21st century. The number of those affected continues to rise at an alarming rate, particularly among children. Obesity is already responsible for 2-8% of all health costs and 10-13% of deaths in different parts of the region. (World Health Organization).Obesity is the extreme condition of being overweight. Obesity can be defined as an adult who has a Body Mass Index (BMI) of 30 or higher. BMI can be calculated by a person’s weight (in kilograms) divided by the square of his or her height (in meters). Obesity and its related diseases are more prevalent among groups with low socioeconomic status. Those on lower incomes tend to consume more meat, fat and sugar, and those on higher incomes, more fruit and vegetables. In addition, poorer groups usually have less access to sport and fitness facilities, which limits the exercise they take. (WHO) Obesity is a health problem in its own right. Physical exercise has become a leisure activity; people have air-conditioned cars and buy their food from supermarkets. Along the same lines, dietary habits have undergone a major change as well. Fat consumption is on the rise, fast food outlets are found everywhere and an increasing number of people are consuming processed foods at every meal.Obesity is more prevalent amongst the urban poor than the economically affluent. The urban poor live in more crowded and unhealthy conditions and are also subject to severe stress brought about by their marginal situation. (WHO) The study of obesity is a very vital one in these current times. Obesity has increased sharply in populations that include both children and adults, since the mid-seventies. Data from two NHANES surveys show that among adults aged 20–74 years the prevalence of obesity increased from 15.0% (in the 1976–1980 survey) to 32.9% (in the 2003–2004 survey) the implications to health from being obese are immense. Being overweight or obese increases the risk of many diseases and health conditions, including the following: • • • • • • • •

Hypertension (high blood pressure) Osteoarthritis (a degeneration of cartilage and its underlying bone within a joint) Dyslipidemia (for example, high total cholesterol or high levels of triglycerides) Type 2 diabetes Coronary heart disease Stroke Gallbladder disease Sleep apnea and respiratory problems 184

•

Some cancers (endometrial, breast, and colon) (Department of Health and Human Services, 2008) Overall there are a variety of factors that play a role in obesity. This makes it a complex

health issue to address. Obesity results from an energy imbalance. This involves eating too many calories and not getting enough physical activity, when calories consumed is more than energy burned. Body weight is the result of genes, metabolism, behavior, environment, culture, and socioeconomic status. Behavior and environment play a large role causing people to be overweight and obese. These are the greatest areas for prevention and treatment actions. (Adapted from U.S. Surgeon General’s Call to Action to Prevent and Decrease Overweight and Obesity, 2001)

In this study, the latest data of the past decade on obesity and related statistics that were released by the OECD were employed (Health Data). These records are augmented with countrylevel data pertaining to a wide variety of social, economic, and environmental variables found to be important determinants of obesity-related measures in past studies. This article provides an empirical assessment of the different factors that contribute to the growth of the obese population over the decade (1990-2000) in OECD countries. 2.0 OBESITY TRENDS: The chart below shows the comparative percentage of adults that can be labeled as obese in the OECD countries; it shows men and women separately. More men than women are overweight but obesity is pretty evenly distributed in all OECD countries. The upward trend of the obesity rates in OECD countries (both developing and industrial) in the past two decades reinforces the idea of ‘globesity’- a global obesity epidemic. This epidemic adversely affects economic costs, according to UCLA researchers, obesity raises a person’s health care costs by 36 percent and their medication costs by 77 percent. (OECD health data 2006) In figure 2, it can be observed that obesity in the OECD countries has increased from approximately 25 units in 1990, to approximately 40 units in 2000 and it can be seen to continue to rise rapidly in 2002. Thus we can say that obesity truly is one of the greatest health challenges of the 21st century and the number of those affected continues to rise at an alarming rate.

185

Figure 1 Percentage of adults considered obese in OECD countries. Percentage of adults considered obese ( Body Mass Index exceeds 30kg/m²) 2004 or latest year available Women

Men

Source: OECD (2005), Health at a Glance – OECD Indicators 2005, OECD, Paris; OECD (2006), OECD Health Data 2006, OECD, Paris.

Figure 2 Percentage of individuals with weight problems in OECD countries.

Source: Loureiro, Maria.L, Nayga Rodolfo .M., (2005).

3.0 LITERATURE REVIEW: With the increasing concerns of the rapidly growing obesity figures across the western world and also in some parts of Asia, the consequent research and studies on this subject matter have also increased. The factors that lead to obesity are financial, economic, biological, social 186

and behavioral. Which factor is most significant has not been agreed upon unanimously. Some feel that unhealthy eating habits and no exercise causes obesity whereas others blame it on genetics or forced economic conditions. According to Peralta-Alva Porqueras, (2006) there has been an increase in the obesity rate of both males and females in the United States since the early 1960’s. Porqueras goes on to say that the obesity rates have even doubled in that time period, from the early 1960’s to 2006. Taking this test a step further, a study done by Richards and Patterson (2007) tested that eating unhealthy foods is the reason that there has been a rise in obesity levels in the United States. Their test reveals that people have a strong addiction to carbohydrates, which cause excessive nutrition intake. A study done by Rashad (2006) looked at the relationship of things like exercise, smoking among other factors and their relationships to obesity. This study found that people who smoke are less likely to be obese, since they generally have a lower body mass index (BMI). The data also said that higher the number of years that you were educated, lower the risk of being obese. In general terms, if you graduated high school, but did not go on to college, you are more likely to be obese than someone who is a college graduate, but less likely to be obese than someone who did not graduate from high school. The study also found that you are more likely to be obese if you are either married or widowed than if you are divorced. Smith, Stoddard and Barns (2007), completed a study to find if people’s income was a key determinate to why people are obese. The study did find that people who believe that they have a chance of becoming unemployed have a greater chance to gain weight. In other words, a temporarily unemployed person is more likely to fill his insecurity with food rather than productive workouts. The study also found that people with a good job and comfortable health insurance were more likely to have a small weight gain. Loureiro and Nayga (2005), combined the social and economic factors that contribute to obesity and specifically observed the environment in which their sample population lived – rural or urban, and their country’s per capita income. By including things like GDP and rural area, they were able to better determine what causes obesity. They concluded that the population living in rural areas carries negative and statistically significant coefficients, whereas per capita income level affects the incidence of obesity rates.

187

A study completed by Goel (2006), looked at the obesity rate in developing countries. The study concluded that there are both endogenous and exogenous factors that affect obesity, as well as demand driven (economic and health) factors and supply driven (technology and government) factors. It is important to look at these along with the individual’s consumption to get a complete picture of obesity. The trends of obesity have been increasing rapidly in most developed countries, especially in the OECD member countries (see figure 2). To my knowledge there has been only one other study done on the factors that affect obesity in OECD countries by Loureiro and Nayga in 2005. That study however only looked at socioeconomic factors that affect obesity. To fill the void and to add to the minimal studies done on the concerning topic of obesity this paper looks at the social and behavioral factors that affect obesity in OECD countries. 4.0 DATA AND EMPIRICAL METHODOLOGY 4.1 Data: The data employed in this research came from a variety of international organizations and databases. Annual data was collected from twenty-seven member countries of the Organization of Economic Development and Co-operation (OECD), with the exception of Luxemburg, Turkey and Greece which were omitted from the country-list due to insufficient data available for the independent variables needed. A cross section of the averages over the decade of 1990-2000 was taken for 27 member countries of the OECD. The log of the averages was taken into consideration for further study. Data for the dependent variable of OBESITY (total, males and females) and the independent variable of SMOKING and ALCOHOL was collected from the 1st edition of the OECD Health Data 2004. The data for the variables of PROTEIN and SUGAR was collected from the FAO Nutrition Database. PUB_PRI_EXP and L_BIRTH_WT data was country-specific and was thus collected from individual country databases, the source for the public and private expenditure variable was Education at a Glance, OECD, Paris, 2003 and individual country mappings and national data sources. For low birth weight in Australia, AIHW National Perinatal Statistics Unit 2003; in Canada the Health statistics divisions; Czech Republic from the Czech Statistical Office; in Denmark the National Board of Health; in Finland, STAKES, National Research and Development Centre for Welfare and Health; Germany Federal Statistical Office - Official 188

population statistics; Hungary Central Statistical Office ;Italy National Institute of Statistics; Japan Ministry of Health, Labor and Welfare-Vital statistics of Japan; Korea National Statistical Office-Annual Report on the Vital Statistics; Mexico’s information found from the Ministry of Health; Netherlands’ data from the Health Interview Survey; data for New Zealand from the Health Information Service; Norway’s data from the Medical Birth Registry of Norway (MBRN); data for Poland from the Central Statistical Office of Poland; Slovak Republic’s data from the Statistical Office of the Slovak Republic; Spain-Instituto Nacional de Estadistica, vital statistics and Sweden’s information from The National Board of Health and Welfare. The complete list of variable definitions and data sources are presented in Appendix A. 4.2 Methodology: In this empirical study, we estimate the regressions with ordinary least squares (OLS) to account for the non observed heterogeneity across countries. Three regression models have been included in this study; the first includes total OBESE population and ALCOHOL, L_BIRTH_WT, SUGAR, SMOKING, PROTEIN and PUB_PRI_EXP as explanatory variables, the second model has OBESE_FM as the dependent variable with the same independent variables and the third model has OBESE_ML as the dependent variable again with the same independent variables. This study distinguishes between total obese population (OBESE), female population who are considered obese (OBESE_FM) and population of males (OBESE_ML) who are considered obese because of clear, observed differences. The regression models have the following functional forms:

OBESEi t = β0 ALCOHOLi t +β1 L_BIRTH_WTi t + β2 SUGARi t – β3 SMOKINGi t – β4 PROTEIN i t +β5 PUB_PRI_EXP + ε i t

II)

OBESE_FMi t = β0 ALCOHOLi t +β1 L_BIRTH_WTi t + β2 SUGARi t – β3 SMOKINGi t – β4 PROTEIN i t +β5 PUB_PRI_EXP + ε i t

III)

OBESE_MLi t = β0 ALCOHOLi t +β1 L_BIRTH_WTi t + β2 SUGARi t – β3 SMOKINGi t – β4 PROTEIN i t +β5 PUB_PRI_EXP + ε i t

189

Subscripts i and t, respectively, denote a country and a specific year. The Variable OBESE equals population of total individuals with BMI greater than 30 kg/m 2 in country i and time t. OBESE_FM is Obese population as a percentage of females with a BMI>30 kg/m2. OBESE_ML is the obese population as a percentage of males with a BMI>30kg/m2. ALCOHOL is Consumption of alcohol in each country in liters per capita (15+); L_BIRTH_WT is the number of live births weighing less than 2500 grams as a percentage of total number of live births; SUGAR is the variable for all quantities of sugar in its centrifugal, refined state, expressed in kilograms per capita per year; SMOKING is the consumption of tobacco, as a percentage of population of daily smokers in each country; PROTEIN is the total protein intake in Grams per capita per day. PUB_PRI_EXP is Total public and private expenditure for educational institutions as a percentage of GDP. Finally, εit is the stochastic error term. The complete variable definitions and corresponding expected signs are presented in Appendix B. Summary statistics for data are provided in table 1. Table 1: Summary Statistics of Dependent and independent variables obesity

alcohol L_birth_wt Protein

Pub_pri Smoking Sugar

Mean

0.98

0.97

0.74

2.00

0.75

1.46

1.54

Median

1.00

0.77

2.00

0.75

1.44

1.57

Maximum

1.38

1.16

0.87

2.07

0.84

1.68

1.66

Minimum

0.34

0.70

0.51

1.90

0.63

1.30

1.21

Std. Dev.

0.24

0.13

0.09

0.04

0.05

0.09

0.11

5.0 EMPIRICAL RESULTS: As discussed earlier, the OLS regressions were estimated to minimize the sum of the squared residuals’ of the data. Table 2 represents the results obtained from this estimation. The results reinforce some of the conclusions from previous studies and add significant information regarding the effect of various social and behavioral factors on obesity in OECD countries. As shown in table 2 under model I, the independent variables that affect the incidence of obesity (OBESE - as a percentage of total individuals with a BMI>30), in a positive and statistically significant way are sugar consumption, Low birth weight and alcohol consumption. Further, other factors such as consumption of protein intake and tobacco consumption in 190

smoking all carry negative and statistically significant coefficients. Other variables, however, such as public and private expenditure are not statistically significant. In model II, the independent variables that affect the incidence of obesity (as a percentage of female population with a BMI>30), in a positive and statistically significant way are sugar consumption and low birth weight. Further, other factors such as protein intake and smoking all carry negative and statistically significant coefficients. The explanatory variable of public and private expenditure for educational institutions and alcohol consumption is not statistically significant. In model III, the independent variables that affect the incidence of obesity (as a percentage of male population with a BMI>30), in a positive and statistically significant way are sugar consumption, low birth weight and alcohol consumption. Further other factors such as protein intake and smoking carry negative and statistically significant coefficients. Like the previous two models, public and private expenditure on education as a percentage of GDP is not significant. This set of results reinforces and expands those found by Rashad (2006) and Loureiro and Nayga (2005). In the study done by Rashad (2006) he concluded that people who smoke are less likely to be obese. Direct interpretation of the coefficients of the explanatory variables helps assess the relative impact of each of the variables, ceteris paribus. Based on the results obtained through the OLS models, with an increase in obesity by one unit of BMI in kg/m2, is caused by an increase of 0.690 grams of birth weight in the total obesity population, an increase of 0.695 grams of birth weight in obesity of female population and an increase of 0.667 grams in birth weight of the obesity in male population. Studies have shown that babies who are born large are more likely to end up fat as adults. Scientists say a person’s weight at birth, as a preschooler and as a teen seem to have a strong connection to weight problems in adulthood. Further, an increment of one unit of kg/m2 of the total obesity population is caused by an increase of 1.557 kilograms of sugar consumption, 1.450kgs in female obesity population and an increase of 1.748kgs of sugar consumption in males. The coefficient of sugar consumption in females is lower than the coefficient of sugar consumption in males because, females are known to be more diet-conscious than their male counterparts are and thus consume less sugar, which then reduces its impacts on obesity.

191

While, the same increase in total obesity and obesity in males is caused by an increase of alcohol consumption by 0.332 and 0.504 liters respectively. In the female population of obesity, alcohol is not a significant factor that contributes to obesity which can be possible because of difference in choice of alcohol consumed and quantity of consumption of alcohol which is greater in males than in females. Females also opt for more light beers and diet mixers like diet coke and zero sprite which are consumed with alcohol in comparison to full sugar content mixers consumed by males. Whereas an increase of a single unit of kg/m2 of obesity is caused by a decrease of 1.924 grams of protein intake for the total population, a decrease of 2.053 grams of protein intake for the obese female population and a decrease of 1.711 grams of protein intake for the male obese population. Further the same increase in obesity is caused by a decrease of the consumption of tobacco by 1.527 percent of daily smokers of the total obese population, a decrease of 1.532 percent of daily smokers from the female obese population and a decrease of 1.546 percent of daily smokers from the male obese population. The decrease of daily smokers which causes a unit increase in obesity is almost identical across all three models, which means that smoking has the same effect on both males and females who are less likely to be obese. The reason for the negative and statistically significant coefficient for smoking as a factor that influences obesity is that consumption of tobacco which contains high levels of nicotine, suppresses appetite. Public and private expenditure on education is not a significant factor contributing to obesity because the trend of highly educated population comprising of developed countries like United States, Australia and the United Kingdom have the highest rates of obesity and some developing countries in Asia who also have high rates of obesity do not spend as much on education as do their western counterparts. Table 2: Regression results I)

OBESITY AND TOTAL

PROTEIN SMOKING SUGAR L_BIRTH_WT PUB_PRI_EXP ALCOHOL C R2

Coefficient -1.924*** -1.527*** 1.557*** 0.690*** -0.094 0.332* 3.891

Std. Error 0.534 0.231 0.197 0.257 0.433 0.176 1.211

t-Statistic -3.600 -6.591 7.899 2.683 -0.217 1.881 3.212

Prob. 0.0019 0.0000 0.0000 0.0147 0.8301 0.0753 0.0046

0.877088

192

F-statistic II)

OBESITY AND FEMALE

PROTEIN SMOKING SUGAR L_BIRTH_WT PUB_PRI_EXP ALCOHOL C R2 F-statistic III)

22.59704

Coefficient -2.053*** -1.532*** 1.450*** 0.695*** -0.293 0.217 4.595

Std. Error 0.530 0.230 0.195 0.255 0.430 0.175 1.202

t-Statistic -3.870 -6.660 7.406 2.724 -0.683 1.237 3.821

Prob. 0.0010 0.0000 0.0000 0.0135 0.5028 0.2310 0.0012

Std. Error 0.627 0.271 0.231 0.301 0.508 0.207 1.421

t-Statistic -2.728 -5.686 7.556 2.211 0.202 2.432 2.027

Prob. 0.0133 0.0000 0.0000 0.0395 0.8418 0.0251 0.0569

0.873831 21.93194

OBESITY AND MALE

PROTEIN SMOKING SUGAR L_BIRTH_WT PUB_PRI_EXP ALCOHOL C R2 F-statistic

Coefficient -1.711*** -1.546*** 1.748*** 0.667** 0.102 0.504** 2.882 0.843271 13.83538

Note: ***, **, and * denotes significance at the 1%, 5%, and 10%

6.0 CONCLUDING REMARKS: There has been a relentless rise in obesity in the last two decades in OECD countries in particular. This rising trend is costly both in terms of personal suffering and the associated social and economic costs. In order to minimize these costs, it is urgent to devote more research to finding cost – effective responses to curb the growth of obesity. Using macro-level data from various sources, the results of this study suggest that factors such as protein intake, sugar consumption, tobacco consumption, low-birth weight and alcohol consumption significantly influence obesity in OECD countries. While this study provides some interesting findings, due to data limitations and collection challenges, a variable like public and private expenditure on education was just a proxy for the education level or more specifically population of educated individuals in OECD countries. This analysis is also just for OECD countries, and it is possible that the same results may not apply to 193

non-OECD countries. This study could be replicated for future studies for other countries like developing countries or a continent to assess robustness of the findings. 6.1 Limitations: In most countries, data on weight and height are self-reported. Evidence suggests that both men and women underestimate their weight and/or overestimate their height, so leading to an underestimate of the true prevalence of overweight and obesity problems. For example, evidence from Canada is that 13.3% of women and 15.4% of men were obese in 2003 based on self reported data, whereas when actual measures were used in 2004, 22.5% of women were obese as were 22.3% of men. In Canada, New Zealand, Australia, the United Kingdom and the United States, actual measurements were made. The percentages for these five countries are among the highest in the OECD region. (OECD health at a glance, 2006) 6.2 Policy Implications: OECD has not yet carried out any rigorous study on the cost-effectiveness of different measures to prevent obesity or treat its health consequences, and so the OECD does not have a ready-made “global strategy” to tackle the epidemic of obesity. However, in the related area of alcohol consumption, they have carried out a recent review of the experience in several OECD countries which shows that a combination of instruments can help achieve the goal of reducing 3

alcohol consumption . These instruments include public education campaigns, curbs on advertising, restrictions on sales to young people, and increased taxation. Similar instruments are often suggested as prime candidates to curb obesity, but there is little evidence as to their likely effectiveness in this task (Obesity and Health, OECD forum 2004).

194

Appendix A: Variable Description and Data Source

Acronym

Description

Data source

OBESE

Obese population calculated as a percentage of total population with a BMI>30kg/m2

OECD HEALTH DATA 2004, 1st edition

OBESE_FM

Obese population calculated as a percentage of females with a BMI>30kg/m2

OECD HEALTH DATA 2004, 1st edition

OBESE_ML

Obese population calculated as a percentage of males with a BMI>30kg/m2

OECD HEALTH DATA 2004, 1st edition

ALCOHOL

Consumption of alcohol in each country in liters per capita (15+)

OECD HEALTH DATA 2004, 1st edition

SMOKING

Tobacco consumption, as a percentage of population of daily smokers

OECD HEALTH DATA 2004, 1st edition

PROTEIN

Total protein intake in Grams per capita per day

FAO Nutrition database

SUGAR

All quantities of sugar in its centrifugal, refined state, expressed in kilograms per capita per year.

FAO Nutrition database

PUB_PRI_EXP

Total public and private expenditure for educational institutions as a percentage of GDP.

OECD HEALTH DATA 2004, 1st edition

L_BIRTH_WT

Number of live births weighing less than OECD HEALTH DATA 2500 grams as a percentage of total number 2004, 1st edition of live births.

Appendix B: Variables and Expected Signs

195

Acronym

Variable

Description

Expected Sign

ALCOHOL

Alcohol Consumption

Consumption of alcohol in each country in liters per capita (15+)

SMOKING

Tobacco Consumption

Tobacco consumption, as a percentage of population of daily smokers

PROTEIN

Total Protein Intake

Total protein intake in Grams per capita per day

SUGAR

Sugar Consumption

All quantities of sugar in its centrifugal, refined state, expressed in kilograms per capita per year.

PUB_PRI_EXP

Total public/private education expenditure

Total public and private expenditure for educational institutions as a percentage of GDP.

+/-

L_BIRTH_WT

Low Birth Weight

Number of live births weighing less than 2500 grams as a percentage of total number of live births.

REFERENCES

196

Bleich, Cutler, Murray and Adams, (2007) “Why Is The Developed World Obese”, No 12954, NBER Working Papers from National Bureau of Economic Research, Inc . Department Of Health and Services (2008, March 4). Retrieved November 17, 2007, from Centers For Disease Control and Prevention: http://www.cdc.gov/nccdphp/dnpa/obesity/ Goel, Rajeev.K., (2006).Obesity: Economic and Financial Perspective. Journal of Economics and Finance, Fall 2006, v. 30, issue. 3, pp. 317-24. Loureiro, Maria.L, Nayga Rodolfo .M., (2005) International Dimensions of Obesity and Overweight Related Problems: An Economic Perspective. American Agricultural Economic Association. 87 (Number 5, 2005): 1147–1153. OECD (2005), Health at a Glance – OECD Indicators 2005, OECD, Paris; OECD (2006), OECD Health Data 2006, OECD, Paris Porqueras, Peralta-Alva., (2006), “Macroeconomics of Obesity”, Society for economic Dynamics, 687. Richards, M. Patterson ,Tegene.(2007) Contemporary Economic Policy, vol. 25, issue 3, pages 309-324 Rashad, (2006-01), “Structural Estimation of Caloric Intake, Exercise, Smoking, and Obesity”, No 11957, NBER Working Papers from National Bureau of Economic Research, Inc Ruhm, (2000), “Are Recessions Good For Your Health?”,The Quarterly Journal of Economics, 2000, vol. 115, issue 2, pages 617-650. Smith, Stoddard and Barnes, (2007) “Why the Poor Get Fat: Weight Gain and Economic Insecurity”, No 2007-16, Working Papers from School of Economic Sciences, Washington State University.

197

Empirical Economic Bulletin, Spring 2008, Vol. 1

Articles inside

Determinants of Gasoline Price: Can Consumer Spending Make a Difference?, Jonathan Stachelek

Jeffrey Haydock

Factors Affecting Health Care Spending, Jennifer Resende

Differing Immigrant Wages: Causes of Differences in Immigrants Standards of Living, Sahana Zutshi

The Social and Behavioral Factors That Affect Obesity in OECD Countries, Sanjana Desai