EFFECT OF ETHNIC DIVERSITY ON ECONOMIC GROWTH

Transcription

1 EFFECT OF ETHNIC DIVERSITY ON ECONOMIC GROWTH BY Liu Ka Ho Applied Economics Option An Honours Degree Project Submitted to the School of Business in Partial Fulfillment Of the Graduation Requirement for the Degree of Bachelor of Business Administration (Honours) Hong Kong Baptist University Hong Kong April 2005

2 Acknowledgement I would like to express my genuine gratitude to my supervisor Dr. Mo. His valuable advice and research finding not only guided me to do my project smoothly, but also enrich my economic sense. His support initiates me to do my best in this project. At the same time, I have to thank all the parties who gave their considerable comment to my project. 1

3 Abstract This paper employs a framework developed by Mo(2000) to study the role of ethnic diversity in economic growth. We use ordinary least squares method to estimate the direct effect of ethnic diversity on growth and indirect effects of ethnicity diversity on growth through the transmission channels. We find that one-unit increase in the index of ethnic diversity (i.e. change from complete homogeneity to complete heterogeneity), the real GDP growth rate reduces by 2.67 percentage points, or express differently, a 1% increase in the ethnicity index reduces the growth rate by about 0.306%. The direct effect contributes 38% to the total effect, while among the transmission channels, the most important one is the investment channel, followed by the human capital channel, government consumption channel and political instability channel. 2

4 Table of Contents Acknowledgement. 1 Abstract. 2 Table of Contents Introduction Literature Review Analytical Framework Data and Descriptive Statistics Empirical Analysis The Transmission Channels Limitation Conclusion Appendix References. 34 3

5 1. Introduction Economic growth is one of the most important research topics among economists. From the studies of researchers, we know that many factors such as investment, education, government policy, political stability, exchange rate are related to the economic growth. By studying the relationship of these factors and GDP growth, economists can explain or predict a country s economic growth. In the 1960s, World Bank s chief economist predicted that seven of the African countries would clearly have the potential to reach or surpass a 7% growth rate. However, the result turned out was that those countries have a negative growth and the overall real GDP per capita in Africa did not grow during the period of On the other hand, the developing countries in East Asia had a per capita GDP growth over 5% in that period. What happened to make the economists falsely estimate the growth of those African countries? If it is because of the result of poor policies, poor investment, poor political instability and low investment, is there any cause for it? Ethnic diversity (we simply call ethnicity in following parts) maybe the answer. In this paper, we will estimate 4

6 how ethnicity influences the economic growth, its direct effect on growth and indirect effects through other economic factors. The paper is organized as follows. Literature review about ethnicity is shown in the next section. In section 3, we introduce the analytical framework of this paper. The data and descriptive statistics are shown in section 4. Section 5 is the report of our regression results. Section 6 shows the effect of ethnicity on different transmission variables. Section 7 shows the limitation of this paper. The last section is the conclusion. 5

7 2. Literature Review Ethnicity is a variable which is quite new to the economists. Before 1990s, ethnicity is just a variable commonly used in social science literature. Mauro (1995) brings ethnicity to the attention of economists by using it as an instrumental variable for corruption while arguing that corruption causes slower growth and investment. After that, effects of ethnicity start to be a research topic among economic studies. Easterly and Levine (1997) conclude that Africa s growth tragedy is caused by poor public policies, while the Africa s high ethnicity can explain a significant part of why countries have chosen the poor policies. They find that high ethnicity is closely associated with low schooling, underdeveloped financial systems, distorted foreign exchange markets, and insufficient infrastructure. La Porta R. Lopez de Silanes, F. Shleifer, and R. Vishiny (1999) also show that that ethnicity has a negative effect on several indices of the quality of government (measured as property rights protection, the extent of corruption, etc) Beside public policies, Easterly with other researchers find that a variety of public goods such as roads, schools, trash pickup, libraries worsen or receive less funding 6

8 with higher ethnicity in a sample of US cities (Alesina, Baqir, and Easterly, 1997). Some researchers further differentiate the term ethnicity into ethnic fragmentation and ethnic polarization. For instance, Montalvo and Reynal-Querol (2000) show that ethnic fractionalization has a direct negative effect on economic development while religious and ethnic polarization have a negative effect on growth through the reduction of investment, the increase in government expenditure and the increase in the probability of civil wars. However, not all the researchers agree all the effects of ethnicity which we mentioned above. Arcand JL, Guillaumont P, Guillaumont Jeanneney S (2000) question the paper of Easterly and Levine (1997) about the Africa s growth tragedy. They conclude that the results in the paper of Easterly and Levine (1997) are both debatable and weak from an econometric point of view. They also raise an interesting question: if ethnicity is the cause of Africa s growth tragedy, then there is very little, if anything can be done, to change the tragedy, and the tragedy will become the fate of Africa. Other researchers, such as Sachs and Warner (1997), Rodrik (1998), Guillaumont et al. (1998) also find insignificant results for ethnicity in the growth regression models. 7

9 If ethnicity is the cause of the Africa s tragedy, is there really nothing can be done? Collier (1999) and Bluedorn (2001) may give us the solution. They suggest that democracy can diminish the negative effect of ethnic diversity on economic development, which means that the effect of ethnicity is not unavoidable. 8

10 3. Analytical Framework In this paper, we employ a framework developed in Mo(2000) to investigate the effect of ethnicity on economics growth. The framework starts with the input-output relationship, which is characterized by the general production function, (1) Y = Tf ( K, L) where Y is the total output level, T is total factor productivity, K is the capital stock, L is the labor. Total differentiation of Y: (2) dy = fdt + T( f dk f dl) Divide (2) by Y, the equation becomes: K + L (3) dy Y = dt T + Tf dk Y K + f L f L dl L We simplify the function as: (4) GR = F[ γ, IY, dll] where GR represents the real GDP growth rate, γis the growth rate of total factor productivity, IY is the investment output ratio and dll is the growth rate of labor. According to Levine and Renelt(1992), the share of investment in GDP, the rate of population growth, the initial level of real GDP, and a proxy for human capital are the four variables that robust in determining growth. The first two variables belong to the 9

11 growth component, which is related to the factor availability. The last two variables belong to the development component, which is related to the effect of social and technological changes. We include all these variables into our model, and we further define the rate of productivity growth as: (5) γ = γ ( ETHNIC, y0, HUMAN) where ETHNIC is an index for the level of ethnicity, y0 is the initial GDP per capital and HUMAN is an index for human capital stock. Therefore, our economic growth model becomes: (6) GR = F[ γ ( ETHNIC, y0, HUMAN), IY, dll] In the next section, the data and descriptive statistics are shown in detail. 10

12 4. Data and Descriptive Statistics Except the ethnicity index, all the data in our model are obtained from the panel data set assembled by Robert Barro and Jong-Wha Lee. Most of the data in Barro s and Lee s data set are from , covering 138 countries. Some of the data are also available up to The data are usually presented as an averaged five years sub-period, i.e. there are 5 five-year-averaged observations for a variable in the period Due to the fact that there are quite a lot missing observations for the period before 1970 and after 1985, in order to have a more accurate and meaningful estimation, we only use the data from the period of 1970 to 1985 in our model. The reason that we choose a fifteen-year-period is that a longer period is better for studying the determinants of the growths in total factor productivity and capital stock. We use the same definition of ethnicity as Easterly and Levine(1997) used in their paper. They primarily defined ethnicity as ethnolinguistic fragmentation, which was constructed in 1960 by a team of 70 researchers at the Miklukho-Maklai Ethnological Institute in the Soviet Union and printed in the 1964 Atlas Narodov Mira (Atlas of 11

13 Peoples of the World). It measures the probability that two randomly selected individuals in a country will belong to different ethnolinguistic groups. This variable ranges from 0 to 1. The larger number of ethnolinguistic groups and the closer the sizes of the groups are, the larger the index is. For instance, when there is only one ethnic group, the index equals to 0. The index equals to 1 when there are infinite number of ethnic groups. When there are only two ethnic groups with the same size, the index equals to 0.5. The ethnic index modified by Easterly and Levine covers 109 countries. All the data included in our regression model are closely related to the analytical framework except the growth rate of labor. We use the growth rate of population as a proxy for the growth rate of labor. Although the growth rate of population is different from the growth rate of labor and might have different effects to the GDP growth, the quality of the data on population growth is better. It is because different countries have different definitions of labor and the measurements of labor growth are also not identical. It makes the labor growth rates become incomparable. Therefore, it is common for researchers to use population growth as a proxy. The average schooling years in the total population over age 25 is used as a proxy for 12

14 human capital stock. We also add the variable, political instability, into our model to capture the political effect. Political instability is defined as average of the number of assassinations per million population per year and the number of revolutions per year over the period. The annual growth rate of a variable is approximated by finding the compound interest rate over the period The correlation coefficients and descriptive statistics of variables are presented in table 1. TABLE 1 Correlation Coefficients and Descriptive Statistics GR ETHNIC y0 POPG IY HUM GOV INSTAB GR 1.00 ETHNIC y POPG IY HUM GOV INSTAB Mean Std. Dev Note. GR = growth rate of real GDP, ETHNIC = index for ethnicity, y0 = real GDP per capita at 1970, POPG = population growth rate, IY = ratio of private investment to real GDP, HUM = average schooling years in the total population over age 25, GOV = share of government consumption in real GDP, INSTAB = index for political instability. 1 For instance, the annual real GDP growth rate over 1970 to 1985 is estimated by finding r in the formula, GDP1970*(1+r) 15 =GDP

15 5. Empirical Analysis We use the ordinary least squares method (OLS) for regressions to estimate the effect of ethnicity on the economic growth and the White Heteroskedasticity-adjusted t statistics are reported. The results are shown in the following table: TABLE 2 The Effect of Ethnicity on the Growth Rate Estimation: B1 B2 B3 B4 Dependent variables Independent variables GR GR GR GR ETHNIC (-3.05) (-1.40) (-2.45) (-2.81) y0-1.41e e e e-06 (-1.00) (-2.48) (-3.93) (-1.60) POPG (1.16) (1.09) (1.75) (1.19) IY (5.43) HUMAN (3.99) GOV (-3.08) INSTAB Constant (3.74) (0.83) (2.02) (4.81) R No of obs Note: White heteroskedasticity-adjusted t statistics are reported. The t statistics are in parentheses. 14

16 TABLE 2 (Con t) The Effect of Ethnicity on the Growth Rate Estimation: B5 B6 B7 B8 Dependent variables Independent variables GR GR GR GR ETHNIC (-2.83) (-1.19) (-1.67) (-1.07) y0-2.96e e e e-06 (-2.09) (-3.73) (-5.08) (-4.00) POPG (0.98) (2.31) (1.72) (2.06) IY (4.01) HUMAN (3.14) GOV (-2.93) INSTAB (-4.07) (4.48) (-3.32) (-4.31) (3.80) (3.08) (-2.99) (-3.84) Constant (4.82) (0.64) (4.18) (1.44) R No of obs Note: White heteroskedasticity-adjusted t statistics are reported. The t statistics are in parentheses. In table 2, estimation B1 indicates that ethnicity has a significant negative effect on the real GDP growth rate when all the transmission channels are excluded in the regression. Estimations B2 to B5 show the effects of ethnicity to the real GDP growth rate when the possible transmission channels, which are the share of investment, human capital stock, share of government consumption and political instability, are added to the regression independently. As expected, the magnitude and significant 15

17 levels of the ETHNIC coefficient in B2 to B5 are smaller that in B1. This shows that there may be muticollinearity problems between ethnic and the mentioned transmission channels. From estimation B1 to B2, B6, B8, we can observe that when more and more variables are added to the estimations, the coefficients and significant levels of ETHNIC decrease continuously. The magnitude and the significant level of the ETHNIC coefficient are the lowest in estimation B8 when all the possible transmission channels are included. In the next section, we will estimate the direct effect of ethnicity and the indirect effect of the transmission channels caused by ethnicity to the GDP growth. 16

18 6. The Transmission Channels There are four transmission channels in our model, they are: the Investment Channel, the Human Capital Channel, the Government Consumption Channel and the Political Instability Channel. We will estimate each channel individually first and then analyze their effects when they are all included in the regression. 6.1 The Investment Channel When ethnicity is high (e.g. many ethnic groups of same size), in rent-seeking models, the resources spent by each group in order to obtain political influence (e.g. time, labor, etc.) can be considered as a social cost that has a negative effect on economic growth because it implies a non-productive usage of these inputs. This would clearly reduce the investment in the productive sector (Montalvo and Reynal-Querol, 2003). If the growth rate of real GDP depends on the share of investment and the share of investment is affected by ethnicity, the effect of ethnicity on the real GDP growth rate can be decomposed as: dgr GR GR IY (7) = + dethnic ETHNIC IY ETHNIC To find out the effect of ethnicity to share of investment, we estimate the effect of ethnicity on share of investment together with the non-transmission variables by OLS 17

19 method: (8) IY = ETHNIC E-06 y POPG (-2.61) (3.32) (0.24) (6.20) R 2 = 0.26 No. of observations = 101 Equation (8) indicates that ethnicity has a negative effect on the share of investment. The direct effect of ethnicity on the real GDP growth rate is shown in the coefficient of ETHNIC in estimation B2, while that in B1 incorporates the direct effect and the indirect effect from the share of investment. As the indirect effect from the investment channel is captured in estimation B2, the magnitude of the coefficient in B2 is expected to be smaller than that in B1. Base on equation (8), regression B1 and B2, we can calculate the role of the share of investment by using equation (7). The results are reported in Table 3. We can see that about 38% of the growth rate reduction is due to the investment channel. TABLE 3 The Investment Channel Direct Effect Investment channel(a) Total effect(b) (a)/(b) *( ) = [ (a)= ] Note: The summation inside the brackets [ ] is calculated total effect based on Equation (8) and B2, while the estimated total effect is drawn from the result in B1. 18

20 6.2 The Human Capital Channel As the ethnic group which controls the government usually chooses the policy which favors its own ethnic group (Easterly and Levine, 1997), public good such as schooling brings less satisfaction to everyone in a high ethnicity country because of the preferences for language of instruction, curriculum, location, etc. Therefore, less of the public good is chosen by society (Alesina and Spolaore, 1997). Alesina, Baqir, and Easterly (1997) also found that a variety of public goods including schools and libraries worsened or received less funding with higher ethnicity in a sample of U.S. cities. Obviously, ethnicity has a negative effect to human capital stock. If the real GDP growth rate is affected by human capital stock, the effect of ethnicity can be decomposed as: dgr GR GR HUMAN (9) = + dethnic ETHNIC HUMAN ETHNIC We estimate the effect of ethnicity on human capital by following regression: (10) HUMAN = ETHNIC y POPG (-1.22) (6.91) (-2.63) (5.31) R 2 = 0.76 No. of observations = 88 As we expected, equation (10) indicates that ethnicity has a negative effect on the level of human capital. The direct effect of ethnicity on the real GDP growth rate is shown in the coefficient of ETHNIC in estimation B3, while that in B1 incorporates 19

21 the direct effect and the indirect effect from the human capital channel. As the indirect effect from the human capital channel is captured in estimation B3, the magnitude of the coefficient in B3 is expected to be smaller than that in B1. Base on equation (10), regression B1 and B3, we can calculate the role of the human capital by using equation (9). The results are reported in Table 4, showing that the human capital channel account for about 15% of the effect of ethnicity. 2 TABLE 4 The Human Capital Channel Direct Effect Human capital channel(a) Total effect(b) (a)/(b) *( ) = [ (a)= ] Note: The summation inside the brackets [ ] is calculated total effect based on Equation (10) and B3, while the estimated total effect is drawn from the result in B The Government Consumption Channel When ethnicity is high, government increases public expenditure in order to mitigate potential conflict. It contributes a negative effect on economic growth since it diverts resources from the private sector. (Montalvo and Reynal-Querol, 2003). Therefore, if the real GDP growth rate is affected by share of government consumption to real GDP 2 We use the estimated effect of ethnicity on human capital in equation (10) for calculation in Table 4, although the coefficient is statistically insignificant. Statistical insignificance may be due to the absence of a functional relationship between the variables or to multicollinearity among the independent variables or because the relationship is relatively small. Since the estimated result is consistent with the calculated results shown in Table 4, we adopt the last interpretation. 20

22 and share of government consumption to real GDP is affected by ethnicity, we can decompose the effect by: dgr GR GR GOV (11) = + dethnic ETHNIC GOV ETHNIC To find out the indirect effect of ethnicity to share of government consumption, we estimate the effect by following regression: (12) GOV = ETHNIC -8.65E-06 y POPG (1.17) (-2.65) (-0.13) (9.00) R 2 = 0.13 No. of observations = 109 Equation (12) indicates that ethnicity has a positive effect on the share of government consumption. The direct effect of ethnicity on the real GDP growth rate is shown in the coefficient of ETHNIC in estimation B4, while that in B1 incorporates the direct effect and the indirect effect from the share of government consumption. As the indirect effect from the government consumption channel is captured in estimation B4, the magnitude of the coefficient in B4 is expected to be smaller than that in B1. Base on equation (12), regression B1 and B4, we can calculate the role of the share of government consumption by using equation (11). According to the results reported in Table 5, the government consumption channel contributes about 11% of the growth 21

23 rate reduction. 3 TABLE 5 The Government Consumption Channel Direct Effect Government consumption channel(a) Total effect(b) (a)/(b) *( ) = [ (a)= ] Note: The summation inside the brackets [ ] is calculated total effect based on Equation (12) and B4, while the estimated total effect is drawn from the result in B The Political Instability Channel Some researchers show that religious and ethnic polarization have a negative effect on growth through the increase in the probability of civil wars (Montalvo and Reynal-Querol, 2000). Although we do not include the effect of civil wars in our definition of political instability, it is conceivable that ethnicity make the overall political environment more instable. If the real GDP growth rate is affected by political instability, we can decompose the effect of ethnicity to political instability by equation (13): dgr GR GR INSTAB (13) = + dethnic ETHNIC INSTAB ETHNIC We estimate the effect of ethnicity on political instability by following regression: (14) INSTAB = ETHNIC E-05 y POPG (1.35) (-4.44) (-1.74) (4.02) R 2 = 0.13 No. of observations = The reason to use the insignificant result in equation (12) is similar to footnote 2 22

24 Equation (14) indicates that ethnicity has a positive effect on the political instability. The direct effect of ethnicity on the real GDP growth rate is shown in the coefficient of ETHNIC in estimation B5, while that in B1 incorporates the direct effect and the indirect effect from the political instability channel. As the indirect effect from the political instability channel is captured in estimation B5, the magnitude of the coefficient in B4 is expected to be smaller than that in B1. Base on equation (14), regression B1 and B5, we can calculate the role of the share of government consumption by using equation (13). The results are reported in Table 6. It indicates that about 12% of the growth rate reduction is due to the political instability channel. 4 TABLE 6 The Political Instability Channel Direct Effect Political instability channel(a) Total effect(b) (a)/(b) *( ) = [ (a)= ] Note: The summation inside the brackets [ ] is calculated total effect based on Equation (14) and B5, while the estimated total effect is drawn from the result in B1. 4 The reason to use the insignificant result in equation (14) is similar to footnote 2 23

25 6.5 Decomposition of the Transmission Channels Since the share of investment, the human capital level, the share of government consumption and the political instability level are not independent of each other, analyzing the channels individually will lead to biased results. Therefore, we should analyze all the plausible channels simultaneously. The effects of ethnicity and different channels can be calculated by following equation: dgr GR GR TV (15) = + dethnic ETHNIC TV TV ETHNIC where TV = IY, HUMAN, GOV and INSTAB. Refer to the results of estimation B1, B8 and equations (8), (10), (12), (14), the direct impact of ethnicity, effects of the four different channels are shown in table 7: TABLE 7 Decomposition of the Transmission Channels Human Capital Channel Model Direct Impact(a) Investment channel(b) (c) B *( ) *( ) {0.377} = {0.321} = {0.137} Government Political Instability Overall Effect consumption Channel(d) Channel(e) * * = {0.088} = {0.094} [ ] Note: The number inside the brackets [ ] is the calculated total effect that is equal to the summation of (a), (b), (c), (d), (e), while that inside the braces { } is the ratio of the respective channel to the overall effect. 24

26 According to regression B1, the overall effect of ethnicity on the real GDP growth rate is or in elasticity terms. This overall effect can be decomposed into: (1) the direct impact, which accounts for 37.7% of overall effect; (2) investment channel, which account for 32.1% of overall effect; (3) human capital channel, which account for 13.7% of overall effect; (4) government investment channel, which account for 8.8% of overall effect, and; (5) political instability channel, account for 9.4% of overall effect. 25

27 7. Limitation In this paper, we use the same definition of ethnicity as Easterly and Levine(1997). However, there are some critiques among this index. Firstly, some researchers argued that Easterly and Levine did not consider the difference between ethnic fragmentation and ethnic polarization (Arcand JL, Guillaumont P, Guillaumont Jeanneney S, 2000). Ethic polarization refers to the case when there are two ethnic groups with the same size (the ethnicity index = 0.5). Montalvo and Reynal-Querol (2000) showed that ethnic fractionalization has a direct negative effect on economic development while religious and ethnic polarization have a negative effect on growth through the reduction of investment, the increase in government expenditure and the increase in the probability of civil wars. Secondly, this index is a bit outdated. The index is basically formed in 1964, which is 40 years ago. Ethnicity may have changed due to factors including the education policy, the language of instruction, and the change of official language. So the data may not be able to reflect the real situation of ethnicity. 26

28 The reason why we insist in using the definition of ethnicity as Easterly and Levine is that we want to keep our model simple, so we did not consider differentiating the concept between ethnic fragmentation and ethnic polarization. Also, the data in our model started at 1970, the time difference between other variables and the ethnicity index is acceptable. Besides the index of ethnicity, the definition of political instability is also arguable. Different researchers have different definitions on political instability. Since all the variables except ethnicity are in the same data set assembled by Robert Barro and Jong-Wha Lee, we also choose the political instability in this data set for the reason of simplicity. However, this data may not be the best data reflecting political instability. 27

29 8. Conclusion Ethnicity is indeed an important and complex factor to growth. As stated in the part of limitation, this report certainly does not give a complete and perfect framework about the effect of ethnicity on growth. However, we have tried to show that ethnicity has a direct effect and an indirect effect through the four transmission channels: the investment channel, the human capital channel, the government consumption channel and the political instability channel. We find that with one-unit increase in the ethnicity (i.e. change from complete homogeneity to complete heterogeneity), the real GDP growth rate reduces by 2.67 percentage points, or express differently, a 1% increase in the ethnicity index reduces the growth rate by about 0.306%. The direct effect accounts for about 38% of overall effect. Among the transmission channels, the most important one is the investment channel, which accounts for 32%, and then followed by the human capital channel, political instability channel and government consumption channel respectively. These four channels are what we are interested in our model. However, we do not mean that ethnicity is solely related to these four channels. For other possible channels, they are out of the scope of this paper. 28

30 Appendix A. Data Source: Variables Involved Code Source Real GDP per capita GDPSH4 Summers and Heston (1988) Total population POP Summers and Heston (1988) Average schooling years in the total HUMAN Barro and Lee population over age 25 Measure of political instability PINSTAB Barro and Lee Ratio of nominal domestic INVWB World Bank investment to nominal GDP Ratio of nominal public domestic INVPUB World Bank investment to nominal GDP Ratio of nominal government GOVWB World Bank consumption expenditure to nominal GDP Index of ethnolinguistic fractionalization ELF60 Atlas Narodov Mira(1964) 29

31 B. Data Set Obs. Country Name GR ETHNIC y0 POPG IY HUM GOV INSTAB 1 Algeria NA Angola NA NA NA 3 Benin NA NA NA 4 Botswana NA Burkina Faso NA Burundi NA NA Cameroon NA Cape verde NA 0.13 NA NA NA NA NA 9 Central African Rep Chad NA NA Comoros NA 0.62 NA NA NA NA NA 12 Congo NA NA NA Egypt NA NA Ethiopia NA NA Gabon NA NA Gambia NA NA Ghana Guinea NA NA Guinea-Bissau NA 0.22 NA NA NA NA NA 20 Cote d'ivoire NA NA Kenya NA Lesotho Liberia Madagascar NA NA 25 Malawi Mali Mauritania NA Mauritius NA Morocco NA Mozambique NA NA 31 Niger Nigeria NA Rwanda NA NA

32 34 Senegal NA Seychelles NA 0.07 NA NA NA NA NA NA 36 Sierra Leone NA Somalia NA South africa NA Sudan Swaziland NA NA Tanzania NA Togo Tunisia Uganda Zaire Zambia Zimbabwe Bahamas, The NA 0.69 NA NA NA NA NA NA 49 Barbados Canada NA Costa Rica Dominica NA 0.69 NA NA NA NA NA NA 53 Dominican Rep El Salvador Grenada NA 0.75 NA NA NA NA NA NA 56 Guatemala NA Haiti Honduras NA Jamaica NA Mexico Nicaragua NA NA Panama St.Lucia NA 0.58 NA NA NA NA NA NA 64 St.Vincent & Grens. NA 0.02 NA NA NA NA NA NA 65 Trinidad & Tobago United States Argentina NA Bolivia Brazil Chile Colombia

33 72 Ecuador Guyana Paraguay Peru Suriname NA NA NA 77 Uruguay Venezuela Afghanistan NA NA NA Bahrain NA NA NA NA Bangladesh NA Myanmar (Burma) NA China NA NA NA NA NA Hong Kong NA India Indonesia NA Iran, I.R. of NA NA Iraq NA Israel Japan Jordan Korea Kuwait NA Malaysia Nepal Oman NA NA NA 97 Pakistan NA Philippines Saudi Arabia NA Singapore NA Sri Lanka NA Syria Taiwan Thailand United Arab 105 Emirates NA NA NA Yemen, N.Arab NA Austria Belgium Cyprus

34 110 Denmark Finland France Germany, West Greece Hungary NA 0.28 NA NA NA Iceland Ireland Italy Luxembourg NA NA Malta NA Netherlands Norway Poland NA NA NA NA NA NA NA 124 Portugal Spain Sweden Switzerland NA Turkey NA United Kingdom Yugoslavia NA 0.17 NA NA Australia Fiji NA New Zealand NA Papua New Guinea NA 135 Solomon Islands NA NA NA NA NA NA NA NA 136 Tonga NA NA NA NA NA NA NA NA 137 Vanuatu NA 0.22 NA NA NA NA NA NA 138 Western Samoa NA 0.83 NA NA NA NA NA NA Note. GR = growth rate of real GDP, ETHNIC = index for ethnicity, y0 = real GDP per capita at 1970, POPG = population growth rate, IY = ratio of private investment to GDP, HUM = average schooling years in the total population over age 25, GOV = share of government consumption in GDP, INSTAB = index for political instability. 33

35 REFERENCES Alesina, Alberto, and Enrico Spolaore, (1997), On the Number and Size of Nations, Quarterly Journal of Economics, CII (1997), Alesina, Alberto, Reza Baqir, and William Easterly, (1997), Public Goods and Ethnic Divisions, Harvard University and World Bank mimeo, March Annett Anthony, (2001), Social Fractionalization, Political Instability, and the Size of Government, IMF Staff Papers, Vol. 48, No. 3, pp Arcand JL, Guillaumont P, Guillaumont Jeanneney S, (2000), How to make a Tragedy: on the Alleged Effect of Ethnicity on Growth, Journal of International Development, J. Int. Dev. 12, pp Bluedorn, J. C., (2001), Can Democracy Help? Growth and ethnic divisions, Economics Letters 70, Collier P., (1999), The Political Economy of Ethnicity, in: B. Plleskovic, J. E. Stiglitz (Eds.), Annual World Bank Conference on Development Economics, Easterly, William and Ross Levine, (1997), Africa s Growth Tragedy: Politics and Ethnic, Divisions, Quarterly Journal of Economics, Vol. 112, pp Guillaumont P, Guillaumont Jeanneney S, Brun J-F, 1999, How instability lowers African growth, Journal of African Economies, March: La Porta R. Lopez de Silanes, F. Shleifer, and R. Vishiny, (1999), The Quality of Government, Journal of Law Economics and Organization 15, 1, Levine, Ross, and David Renelt, (1992), A Sensitivity Analysis of Cross-Country Growth Regressions, American Economic Review 82: Mauro, Paolo, (1995), Corruption and Growth, Quarterly Journal of Economics, CX (1995), Mo, Pak Hung, (2001), Corruption and Economic Growth, Journal of Comparative Economics 29, pp

36 Mo, Pak Hung, (2002), Human Capital and Economic Growth: Alternative Estimation Methods, BRC Working Papers, Hong Kong Baptist University Montalvo, J. G., and Reynal-Querol, M., (2000), The Effect of Ethnic and Religious Conflict on Growth, IVIE Working Paper. Montalvo, J. G., and Reynal-Querol, M., (2003), Ethnic diversity, political systems and conflict, WeltTrends, 38:44 60, 2003a. Rodrik D., (1998), Trade policy and economic performance in sub-sharan Africa, NBER Working Paper No Sachs J, Warner A, (1997), Sources of slow growth in African economies, Journal of African Economies 6: