Printer Friendly

How does racial diversity raise income inequality?

Interest in the determinants of income inequality, most commonly measured as the Gini index, has been sparked by a trend towards greater inequality in most of the advanced economies since about 1970 (Alderson and Nielsen, 2002; Atkinson, 2004). The reasons for this trend are still unknown. One possibility is that rising income inequality might be related to the rising cultural or racial heterogeneity of many national populations.

A recent cross-sectional study did indeed point to population diversity as a determinant of the Gini index. Using novel measures of ethnic, religious and racial diversity, it found that high racial diversity, but not high ethnic or religious diversity, is a robust predictor of a high Gini index (Meisenberg, 2007). The effect is approximately linear at low-to-medium levels of racial diversity, but there is no significant relationship in comparisons of countries at high to very high levels of racial diversity. For example, within the category "Latin America + Caribbean", which has the worldwide highest levels of racial diversity (Table 1), the correlation of racial diversity with the Gini index is slightly and non-significantly negative (r = -.247 when racial distances are defined by genetic distances, N = 23).

The empiric relationship between racial diversity and income inequality is no great surprise. Of the major world regions, racially diverse Latin America has the highest level of income inequality, and income inequality is higher in the United States than in the more homogeneous European and East Asian countries. If racial diversity favors income inequality, and if advanced postindustrial nations are becoming more diverse in the wake of replacement migration (Coleman, 2002), we can predict a continuing trend towards greater income inequality. But which are the causal paths that lead from racial diversity to income inequality? Two alternative but not mutually exclusive hypotheses are tested in the following study.

The first hypothesis proposes that racial diversity raises the Gini index because different racial groups tend to differ in their intellectual abilities and achievements even when they live in the same country (Lynn, 2006). Such differences are not unique to racial groups, but are commonly observed between cultural, national, linguistic and religious groups of the same racial background living in the same country (Lynn & Longley, 2006; Verster & Prinsloo, 1988). Nevertheless, ability differences between racial groups tend to be larger and more persistent than those between cultural, linguistic, religious and national groups. If races differ in intellectual ability, then racial diversity is expected to widen the range of ability levels in the country. Mental ability measured in childhood or adolescence predicts adult earnings at the level of individuals (Deary et al., 2005; Irwing & Lynn, 2006; Murray, 2002). Therefore we can expect that greater ability differences translate into greater income differences.

A second hypothesis proposes that racial diversity raises the Gini index because, ultimately as a result of biological evolution, people are most altruistically inclined toward those who are similar to themselves (Meisenberg, 2007b; Rushton, 1989; Rushton and Bons, 2005). This leads to a more individualistic and less altruistic social ethos in countries with a highly diverse population. Social solidarity is restricted to groups that are defined on racial, ethnic, religious or local lines, rather than extending society-wide; economic exploitation is more extreme; there is less concern for marginalized groups and individuals; and there is less redistribution of wealth from the rich to the poor. In the following, these two hypotheses are examined in turn.

General Methods

The Gini index

The primary data source is the World Income Inequality Database (WIID2a) of the United Nations University, available at www.wider.unu.edu/wiid/wiid.htm. There are more than 100 results listed for some countries, together with the type of data and an estimate for the quality of each study. Only data from 1990 and later were used except in two cases where data from the 1980s were used for lack of more recent results. Only data that were based on disposable income, net income or consumption were included. Data based on gross income were not used. For each country an average was formed from the results of the highest-quality data in these categories. Gini indices for 123 countries could be computed with this procedure, and Ginis for 10 additional countries were extrapolated from other sources as described in Meisenberg (2007). A listing of the resulting 133 Ginis can be found in Meisenberg (2007). Table 1 shows the Gini index together with the average (country-level) racial-diversity scores for different world regions.

Racial diversity measures

The construction of the racial-diversity indices is described in Meisenberg (2007), along with a listing of country-level racial diversity scores for 198 countries and territories. Racial distances were weighted either by genetic distance according to Cavalli-Sforza and Feldman (2003), or by IQ difference according to Lynn (2006). These two measures are highly correlated (r = .969). To maximize the main effects, logarithmic transformations were used for analyses spanning the entire range of countries with available Gini index; and untransformed racial diversity scores for analyses of country samples that include few countries with very high racial diversity. Racial diversity weighted by genes or by IQ were used alternatively in regression models.

Other variables

The following country-level data were used for correlational studies:

1. lgGDP is the logarithm of gross domestic product adjusted for purchasing power, averaged for the years 1990-2005. Data are from the World Development Indicators of the World Bank, which can be purchased at http://econ.worldbank.org. The logarithmic transformation was used because of the highly skewed nature of GDP worldwide, which approximates to a normal distribution in the logarithmic form.

2. Education was calculated by averaging the standardized scores of 4 measures: (1) Average years of schooling of adults over the age of 25 from the Barro-Lee dataset, averaged for the years 19902000. The data are available at: http://www.cid.harvard.edu/ciddata/Appendix%20Data%20Table s.xls. (2) The school life expectancy (1999-2003 average) from UNESCO at: http://stats.uis.unesco.org/TableViewer/tableView.aspx. (3) The combined enrolment ratio for primary, secondary and tertiary schools in 2002, from the 2005 Human Development Report of the United Nations (http://hdr.undp.org/reports/global/2005). (4) The arcsine-transformed averages of the adult literacy rates in 1990 and 2002, from the 2005 Human Development Report.

3. IQ is the average IQ in the country according to Lynn & Vanhanen (2006) and Lynn (2006). Measured scores are available for 114 countries. Lynn & Vanhanen (2006) also provide estimates for the IQs of 79 additional countries based on the IQs of neighboring countries with similar population and culture. The estimates were included in the present study, yielding a total of 193 countries with IQ scores.

4. Corruption was averaged from the reverse of Transparency International's Corruption Perception Index, available at http://www.transparency.org (average for the years 1999-2005), and the corruption domain of the Heritage Foundation's 2006 Index of Economic Freedom (http://www.heritage.org/research/)

5. Economic Freedom was averaged from two sources: (1) Areas 2-5 of the Fraser Institute's economic freedom index (Gwartney, Lawson et al., 2007) were averaged over the years 1980-2005. Area 1 was excluded because it is distinct from areas 2-5 both conceptually and factorially (r = -.015). (2) 9 of the 10 areas of the Heritage Foundation's 2006 Index of Economic Freedom (http://www.heritage.org/research/) were subjected to principal components analysis with 2-factor varimax rotation. Corruption was excluded because it was considered conceptually different from economic freedom. The first principal component of this factor rotation correlated with areas 2-5 of the Gwartney & Lawson index at r = .891. These two variables were averaged into one combined economic freedom index, which correlated with corruption at r = -.829. To avoid excessive collinearity, corruption and economic freedom were used alternatively in regression models. Only the variable producing the better fit was retained.

6. Freedom/Democracy is the average of two measures: Political Freedom, defined as the scores of political freedom (political rights + civil liberties) from Freedom House at http://www.freedomhouse.org/research/freeworld, average 19882005; and Democracy, defined as Vanhanen's democracy index, average 1990-2004, from the Finnish Social Science Data Archive at http://www.fsd.uta.fi/english/data/catalogue/FSD1289/. These two highly correlated measures (r = .847, N = 179 countries), and the average of the two, were used alternatively in regression models.

7. The square root of the country's land surface was determined from data in the World Fact Book of the CIA: www.cia.gov/cia/publications/factbook. The logarithm of the population density in 1997 was formed from data in the World Development Indicators of the World Bank, available from http://econ.worldbank.org.

Results

Relationship between Gini index and racial diversity

Table 2 shows that the Gini index is negatively related to many development indicators. The strongest relationship is with IQ, but also high lgGDP, education, democracy and economic freedom are associated with a low Gini index while high corruption is related to a high Gini index. As expected, many of the correlations between these predictors are quite high. The racial diversity measures have substantial positive correlations with the Gini index although their correlations with most of the other variables are insignificant.

The independent effects of the predictors were investigated in regression models. Initial models contained IQ, education, the logarithm of GDP (lgGDP), corruption, the average of political freedom and democracy, the logarithm of racial diversity, a categorical measure of communist history, the square root of the country's surface area, and the logarithm of its population density. In order to capture non-linear effects, both the centered linear terms and quadratic terms were included in the models.

Economic freedom was allowed to substitute for corruption when this improved the fit of the model, political freedom or democracy were allowed to substitute for their average, and a categorical measure for "ex-communist" (for the countries of eastern Europe and the former Soviet Union) was allowed to substitute for "communist history." Models were started with the average of "racial diversity by genes" and "racial diversity by IQ," and one of these two variables was allowed to substitute for the average if this improved the fit of the model. Models were developed to maximize the adjusted R2, and non-predictors were dropped. Table 3 shows the results.

Model 1 shows the prediction of the Gini index in the absence of racial diversity. The Gini index (measured mainly during the 1990s) is reduced by high IQ, high GDP, high population density and a history of communist rule, and raised by high education. IQ and education show saturation effects at high levels. Countries with extremely high and extremely low levels of political freedom tend to have a low Gini index. Model 2 adds interaction effects between the continuously measured variables in model 1. There is a strong interaction effect between education and lgGDP. The negative sign of the interaction implies that the Gini index is lowest in countries in which the average educational level of the population is congruent with the level of economic development. Actually, the Gini index is lowest in countries in which the average educational level is a bit lower than expected for the country's wealth. High education in a poor country and very low education in a very rich country both raise the Gini index. This finding is replicated in models 4 and 6.

Model 3 adds the logarithm of racial diversity to model 1, with a highly significant effect of racial diversity. Weighting of racial diversity by IQ produces a marginally better fit than weighting by genetic distance, but the difference is not nearly significant. Model 4 adds interaction effects to model 3.

Correlational analyses at the country level are subject to spatial autocorrelation, also known as Galton's problem (Eff, 2004). In essence, data from different but geographically, culturally and historically similar countries are not strictly independent of each other and tend to show similar correlations. This can lead to type 1 errors. To control for spatial autocorrelation, dummy variables for the major world regions were introduced one at a time to models 3 and 4. Those showing evidence for effectiveness were then introduced jointly, and non-predictors were removed. After these manipulations the racial diversity effect was reduced, especially by the inclusion of Latin America, but it remained statistically significant. Models 5 and 6 are equivalent to models 3 and 4, respectively, but limited to the 109 countries outside of Latin America and the Caribbean. The results show that the worldwide relationship between Gini index and racial diversity is not merely caused by a coincidence of high Gini index and high racial diversity in the countries of Latin America and the Caribbean.

Racial diversity and variance in cognitive ability

The spread of mental ability in the national population is measured as the dispersion of ability levels on standardized mental tests administered to representative samples of the national population.

The most useful data come from two sets of international assessments of school achievement: the Third International Mathematics and Science Study (TIMSS), organized by the International Study Center of the Lynch School of Education at Boston College; and the Programme for International Student Assessment (PISA) of the OECD. A recent study (Lynn et al., 2008) showed that the pooled results of these assessments correlate at r = .92 with the IQs published by Lynn & Vanhanen (2006). This shows that at the country level, mental ability tests ("IQ tests") and tests of school achievement measure the same or nearly the same construct. These measures of intellectual proficiency have been claimed to be better indicators of "human capital" than traditional measures of education such as years in school and educational degrees (Weede, 2004).

TIMSS results for science and mathematics were used from the 1995, 1999 and 2003 assessments. The results of the 1999 assessment are from Martin et al. (1999) and Mullis et al. (1999), available at http://isc.bc.edu/timss1999i/math_achievement_report.html and http://isc.bc.edu/timss1999i/science_achievement_report.html. The results of the 1995 and 2003 assessments are from exhibits 1.1 and 1.3 in Martin et al. (2004) and Mullis et al. (2004), available at http://timss.bc.edu/timss2003i/intl_reports.html. These sources contain the dispersion of scores (5th to 95th percentile) in graphic form.

The 5th and 95th percentile scores of the PISA assessments are published in numeric form at http://pisacountry.acer.edu.au/ for 2003, and at http://www.oecd.org/dataoecd/30/18/39703566.pdf for 2006. The differences between these scores were calculated for all domains (mathematics, reading, science and problem solving in 2003; mathematics, reading and science in 2006).

For all tests of ability or achievement, the dispersion of raw scores depends on the relationship between the difficulty of the test items and the performance level of the tested group. For example, the 5%-to-95% performance range of an easy test will be systematically wider in a low-scoring than a high-scoring group, and the reverse is true for a difficult test. Therefore the dispersion for each domain in each assessment was residualized against the mean score, and the residuals were standardized to a mean of zero and a standard deviation of one. The correlations between the residuals obtained from different assessments were small but positive. The correlation between the average dispersion on all domains of all TIMSS assessments and the average dispersion on all domains of all PISA assessments was only r = .256, although the correlations between different assessments within the same program were substantially higher. This suggests that sampling bias in one or both of the assessment programs is substantial.

Table 4 shows the correlations of test score variability on each individual assessment program (vTIMSS and vPISA), and the combined data from both programs (vTIMSS/PISA), with country-level variables. Overall the only noteworthy correlations are with racial diversity (especially when weighted by IQ) in either the untransformed or log-transformed form, and with the measures of political freedom and democracy.

In order to be a credible mediator of racial diversity effects on the Gini index, ability dispersion must be predicted by racial diversity in the presence of other predictors; and ability dispersion must be an independent predictor of the Gini index in the presence of other predictors. Table 5 shows the best-fitting regression models for the prediction of ability dispersion in the two assessments. The procedures were the same as those used for the models in Table 3. The racial diversity measures were used either with or without the logarithmic transformation with the aim of maximizing the main effect.

Table 5 shows that the overall explained variance is modest, and not all results replicate across the two assessment programs. For example, effects of the country's size (sqrtArea) and population density (lgPopDens) are small and inconsistent, and the same is true for the political indicators of democracy, political freedom and communist history. This is expected if either the accuracy of the outcome measures is low or if most of the chosen predictors are genuinely unrelated to the outcomes. Besides education and corruption, measures of racial diversity are the only consistent predictors. Racial diversity weighted by IQ is more predictive than racial diversity weighted by genes, which is expected if the effect is genuine. Statistical significance levels for the racial diversity measures are nonetheless moderate: p = .028 for TIMSS, r = .002 for PISA, and r = .019 for the combined results.

Variance in cognitive ability and income inequality

A greater variance of cognitive ability can mediate the effect of racial diversity on the Gini index only if greater variance in cognitive ability leads to a higher Gini index. Table 4 does indeed show a non-significant trend for a positive association of the Gini index with ability dispersion. However, Table 2 shows far higher correlations of the Gini index with several other predictors. Therefore regression models were examined in which the Gini index was predicted by ability dispersion along with the other predictors.

Consistent predictors of the Gini index included racial diversity (positive), lgGDP, and communist history (both negative). However, there were no significant effects of ability dispersion on the Gini index. In the PISA assessment (N = 51 countries), the effect of ability dispersion had a positive sign with p = .282; in TIMSS (N = 46 countries), the effect had a negative sign with p = .198; and in the combined set (N = 63 countries), it had a positive sign with p = .413. Thus the observed trends are inconsistent and far from even approaching statistical significance.

Racial diversity and big government

If a high level of racial diversity reduces social solidarity, and especially the redistribution of wealth from the rich to the poor, then we can predict that racially diverse societies have "small" governments that stay lean and mean by minimizing spending for social welfare and other redistributive programs. The following analyses examine whether racial diversity reduces the size of government, and whether smaller government leads to greater income inequality.

A Big Government variable was formed from two sources: (1) Area 1 of the Gwartney, Lawson et al. (2007) economic freedom index ("Size of Government") was used directly. (2) The second principal component of a two-factor varimax rotation of nine of the 10 dimensions (excluding corruption) of the Heritage Foundation's Index of Economic Freedom was used. The components loading most strongly on this principal component are "Fiscal Policy" (low taxes, r = .764), "Government" (low government expenditures, no state-owned enterprises, r = .709), and "Labor Freedom" (flexible hours, no minimum wage, freedom of employers to lay off redundant employees, r = .561). "Small government" as defined by this factor is also associated with more corruption (r = .309). This principal component correlates at r = .605 with area 1 of the Gwartney, Lawson et al. (2007) index. Before averaging, the two components were standardized and reversed such that high numbers correspond to "bigger" government. For those 131 countries for which all relevant variables are available, correlations of Big Government were r = -.419 with the Gini index, -.312 with the logarithm of racial diversity by genes, and r = -.224 with the logarithm of racial diversity by IQ. Big Government is also positively associated with democracy (r = .405), lgGDP (r = .349), political freedom (r = .320) and IQ (r = .197), and negatively with corruption (r = -.429).

The question of whether racial diversity is an independent and negative predictor of Big Government was explored in regression models with other plausible predictors. Model 1 in Table 6 shows that racial diversity reduces the size of the government independent of the other predictors in the worldwide sample. The size of the government is also reduced independently by a high level of economic freedom and by high population density, and increased by excessive economic wealth. Also a very high level of democratization increases the size of the government, although a transition from no democracy to some democracy does not have this effect.

When dummy-coded world regions were added individually to model 1, Latin America (+ Caribbean) strongly reduced big government while reducing the racial diversity effect to non-significance (p = .122). However, when the analysis was limited to the 108 countries outside Latin America and the Caribbean, the main effect of racial diversity was again statistically significant (p = .030, model 2 in Fig. 6). The reason for this result is that within Latin America and the Caribbean there is no negative but actually a positive relationship between the logarithm of racial diversity and big government (r = .427, p = .042, N = 23). Thus the negative relationship between racial diversity and big government does not apply to comparisons of countries with very high levels of racial diversity. This is the same pattern that is also observed for the positive relationship between racial diversity and the Gini index.

Big government and the Gini index

Table 7 shows three models in which the Gini index is predicted by Big Government. Model 1 predicts the Gini index with Big Government and other predictors but without a measure of racial diversity. It resembles model 1 in Table 3, but shows that Big Government clearly reduces the Gini index independent of the other predictors. Model 2 includes racial diversity as an additional predictor. It corresponds to model 3 in Table 3, but with Big Government as additional predictor. Although racial diversity itself emerges as an independent positive predictor of the Gini index in this model, it attenuates the effect of Big Government only to a slight extent. Model 3 includes interaction effects. As in the models of Table 3, there is a highly significant interaction between education and lgGDP (but not education and big government). This interaction does not detract from the independent effects of big government and racial diversity. The racial-diversity effects in models 2 and 3 are not quite as impressive as those in models 3 and 4 of Table 3, indicating that the Big Government variable does indeed capture some of the mediating mechanisms between racial diversity and income inequality.

When dummy-coded world regions were added singly to model 2, Latin America (+ Caribbean) attenuated the main effect of racial diversity to borderline non-significance (p = .062) but left the effect of Big Government intact with p = .001. All other world regions left both the racial diversity effect at p < .01 and the Big Government effect at p [less than or equal to] .001. Also in model 3, the addition of Latin America reduced the racial diversity effect to non-significance (p = .209) but left the effect of Big Government at p = .002. All other world regions left the racial diversity effect at p < .05 and the Big Government effect at p < .005.

lgGDP, which tended to be negatively related to the Gini index in the models of Table 3, has no effect on the Gini index when Big Government is one of the predictors. Thus any inequality-reducing effect of national wealth appears to be mediated by the proportionately greater government budgets that are typical for wealthy nations. However, the still intact interaction between education and lgGDP shows that this is indeed an interaction of education with national wealth, not an interaction of education with government expenses.

Discussion

Determinants of the Gini index

The regression models of Table 3 give some general insights into the determinants of income inequality:

1. Rising IQ lowers the Gini index up to an IQ of 90 to 95. Above this limit, IQ no longer influences the Gini index. A reasonable explanation is that institutions and procedures of collective bargaining play a major role in the containment of income inequality, presumably by reducing the power imbalance between the rich and the poor (Golden and Londregan, 2006). A certain intelligence is required to create such institutions and procedures, and to use them for their intended purpose.

2. High education tends to raise the Gini index, presumably because education enables educated people to acquire wealth at the expense of those with less education.

3. Education interacts with lgGDP. A well developed educational system in a poor country is associated with a high Gini index, presumably because education creates elites that redistribute the limited national wealth for their own benefit; and countries with low levels of education but great material wealth usually are poorly developed but resource-rich countries in which the exploitation of the country's natural resources enriches a (more or less corrupt) elite.

4. Both very high and very low levels of political freedom reduce the Gini index, perhaps because both extremely dictatorial and extremely democratic societies tend to impose restrictions on the acquisitive activities of the business elite.

5. Even during the 1990s, when most of the Ginis used in this study were recorded, a history of communist rule still reduced income inequality. Although income inequality has been rising fast in many successor states of the former Soviet Union, most of the ex-communist countries of Eastern Europe still have fairly low income inequality.

6. The reason for the inequality-reducing effect of high population density is uncertain. In regression models, a high level of urbanization fails to reduce the Gini index independent of population density, nor does it diminish the effect of population density on the Gini index. A high proportion of agricultural employment raises the Gini index somewhat, but without substantially reducing the effect of population density.

7. High racial diversity raises the Gini index. Although racial diversity weighted by IQ emerged as marginally more predictive than racial diversity weighted by genetic distance in the models of Tables 3 and 7, the difference is trivial and not nearly significant.

The main limitation of the racial-diversity indices is that they are based only on the proportions of racial groups from which the national population was formed after the year 1500. They take no account of the extent to which these racial groups have merged into a single gene pool. For example, Paraguay has high racial diversity scores because the origin of the population is presumed to be 50% European and 50% Amerindian. However, it can be argued that racial diversity in Paraguay is near-zero because almost everyone is a mestizo. The fact that many of the countries with the highest diversity of racial origins (most of them in Latin America and the Caribbean) also have extensive hybridization between races could be responsible for the observation that racial diversity no longer predicts small government or a high Gini index in this group of countries. Besides hybridization, the extent of race-based privilege and discrimination in the society is a likely determinant of economic inequalities.

Racial diversity, ability dispersion and income inequality

The "meritocratic" hypothesis proposes that racial diversity increases the dispersion of mental ability levels in the population, which in turn raises income inequality through market forces. The first step of this two-step process finds cautious support in data from standardized international assessments of student achievement. In interpreting the results, we must be aware of the limitations in the TIMSS and PISA assessments. Both assessment programs include a smallish number of participating countries, most of these countries have high levels of economic development and low levels of racial diversity, and the enlistment of a country sample that is representative not only for the average level but also the variance of achievement or ability in the country is difficult.

Although the raw correlations reported in Table 4 are weak, the regression models of Table 5 do show statistically significant effects of racial diversity. Taken together, the relative congruence of results from two independently conducted assessment programs argues that the postulated causal effect of racial diversity on ability dispersion is genuine.

The translation of ability differences into income differences is often assumed to be the outcome of market forces, based on the common observation that higher mental ability leads to higher income in comparisons between individuals (Deary et al., 2005; Irwing & Lynn, 2006; Murray, 2002). Several authors have proposed that rising income inequality is caused by a rising demand for complex skills, especially technical skills (Acemoglu, 2002; Galor and Moav, 2000). However, others have pointed out that rising income inequality is not driven by technicians and engineers, whose relative incomes have, if anything, declined (at least in the United States), but by high-level office workers (Morris & Western, 1999, p. 635). Thus rising inequality is not driven by a rising demand for technical skills, but by the rising skill of corporate management to channel a greater share of their companies' resources into their own pockets. Whether this particular skill is closely related to the kind of intellectual ability measured by IQ tests and school achievement tests is uncertain. The failure to find independent effects of ability dispersion on income dispersion suggests it is not. Thus corporate power structures rather than skill-based market forces appear to be responsible for recent developments of income inequality. The effect of education in Tables 3 and 7 points to a possible role of educational credentialing systems in this process, but cognitive ability does not seem to be the decisive factor.

The failure to find independent effects of ability dispersion on income inequality might be due to the limitations of data quality. However, Table 5 shows that data quality is not so poor as to conceal an effect of racial diversity on ability dispersion. This implies that high income inequality is not a necessary attribute of "meritocratic" societies, at least not in comparisons between the kinds of societies that participate in international school achievement tests. Functionalist explanations that are based on market forces and individual differences in achievement are unconvincing. Conflict theories inspired by Marx or Darwin, which propose that power imbalances determine the extent of economic inequalities, appear more fruitful.

Racial diversity, size of government, and income inequality

An alternative explanation for the effect of racial diversity on the Gini index invokes redistributive policies that are, in turn, dependent on political values in the society. The inequality-reducing effect of redistributive policies is well established, at least in economically advanced societies (Atkinson, 2004). Redistributive policies express values of distributive justice that are held by the politically relevant sections of the population. These values vary on several dimensions, the most important of which appear to be equality, efficiency, need and merit (Michelbach et al., 2003, p. 524). Egalitarian motives, in particular, have been demonstrated repeatedly in laboratory settings (Dawes et al., 2007; Scott et al., 2001). Distributive justice norms are known to differ between countries. For example, compared to Americans, Indians give greater weight to need than to merit (Berman et al., 1985). This might be one reason why the Gini index is 38.9 in the United States but only 32.3 in India (Meisenberg, 2007).

We do not know what causes distributive justice norms to vary across countries and time, but racial diversity is one possible determinant. For example, the movement against federally funded social welfare programs in the United States during the 1990s was accompanied by the stereotype of the "black welfare mother." In this case it is plausible that redistributive policies were cut back or terminated because the beneficiaries of such policies were perceived as belonging to a different race. Attitudes to welfare spending in the United States depend on people's exposure to welfare recipients of their own or a different race (Luttmer, 2001).

It has been proposed that, as an extension of kin-directed altruism, people are most altruistically inclined toward those who are similar to themselves. Empirical evidence has been presented showing that sympathy and altruism are favored mainly by similarity in traits with high heritability (Rushton, 1989; Rushton and Bons, 2005). If genetic similarity is the determining factor, then we must expect that altruism is reduced specifically toward members of different races but not necessarily members of different cultural groups. In consequence, social solidarity and redistributive policies are reduced by racial diversity but not cultural diversity. Indeed measures of ethnic and religious diversity from Meisenberg (2007) do not reduce the size of the government the way racial diversity does, nor do they raise the Gini index (data not shown). Thus Rushton's theory of genetic similarity detection is not only a theory of personal preferences, but also a theory of race prejudice and its societal consequences.

The observation that a measure of "big government" is negatively related to both racial diversity and the Gini index suggests that redistributive government policies play a substantial role in the relationship between racial diversity and income inequality. The second step of the proposed causal sequence, the reduction of the Gini index by "big government," is unsurprising, although it contradicts those who describe the role of the government in the economy as purely predatory. The first step, which is the reduction of the size of government in response to high racial diversity, is more interesting. An effect of racial diversity in reducing social services has been demonstrated at the community level in the United States (Alesina et al., 1999), and the present results show that a similar effect also exists at the country level.

Promising areas for future research include specific institutional correlates of racial diversity. For example, public expenditures for health and education benefit, potentially at least, everyone irrespective of race. Therefore they are unlikely to be reduced by high racial diversity. However, poverty is more likely to be associated with race in the public mind, and support for the poor is therefore more likely to be reduced in racially diverse societies. Also societal traits that depend on social solidarity, such as collective bargaining and the unionization of laborers, are predicted to be low in racially diverse societies. Because political values are the most likely mediators of these institutional outcomes, survey-derived measures of values and attitudes should be investigated for racial-diversity effects both at the country level and the community level. The reported negative relationship between racial diversity and social participation in American communities (Alesina and La Ferrara, 2000) points to the possibility of rather general effects of racial diversity on social behavior and attitudes.

References

Acemoglu, Daron

(2002). Technical change, inequality, and the labor market. Journal of Economic Literature 40: 7-72.

Alderson, Arthur S. and F. Nielsen

(2002). Globalization and the great U-turn: Income inequality trends in 16 OECD countries. American Journal of Sociology 107: 1244-1299.

Alesina, Alberto, R. Baqir and W. Easterly

(1999). Public goods and ethnic divisions. Quarterly Journal of Economics 114: 1243-1284.

Alesina, Alberto and E. La Ferrara

(2000). Participation in heterogeneous communities. Quarterly Journal of Economics 115: 847-904.

Atkinson, Anthony B.

(2004). Increased income inequality in OECD countries and the redistributive impact of the government budget. In: G.A. Cornia (ed): Inequality, Growth and Poverty in an Era of Liberalization and Globalization, pp. 221-248. Oxford: Oxford University Press.

Berman, John J., V. Murphy-Berman and P. Singh

(1985). Cross-cultural similarities and differences in perceptions of fairness. Journal of Cross-Cultural Psychology 16: 55-67.

Cavalli-Sforza, Luigi Luca and M.W. Feldman

(2003). The application of molecular genetic approaches to the study of human evolution. Nature Genetics Supplement 33: 266-275.

Coleman, David A.

(2002). Replacement migration, or why everyone is going to have to live in Korea: A fable for our times from the United Nations. Philosophical Transactions of the Royal Society of London B 357: 583-598.

Dawes, Christopher T., J.H. Fowler, T. Johnson, R. McElreath and O. Smirnov

(2007). Egalitarian motives in humans. Nature 446: 794-796.

Deary, Ian J., M.D. Taylor, C.L. Hart, V. Wilson, G.D. Smith, D. Blane and J.M. Starr

(2005). Intergenerational social mobility and mid-life status attainment: Influences of childhood intelligence, childhood social factors, and education. Intelligence 33: 455-472.

Eff, E. Anthon

(2004). Spatial and cultural autocorrelation in international datasets. Department of Economics and Finance Working Paper Series, Februar 2004. Retrieved February 2006 from: http://www.mtsu.edu/~berc/working/spatial%20autocorrelation%20z. pdf.

Galor, Oded and O. Moav

(2000). Ability-biased technological transition, wage inequality, and economic growth. Quarterly Journal of Economics 115: 469-497.

Golden, Miriam A. and J.B. Londregan

(2006). Centralization of bargaining and wage inequality: A correction of Wallerstein. American Journal of Political Science 50: 208-213. Gwartney, James, R. Lawson et al.

(2007). Economic Freedom of the World: 2007 Annual Report, available at http://www.freetheworld.com/release.html.

Irwing, Paul and R. Lynn

(2006). The relation between childhood IQ and income in middle age. Journal of Social, Political and Economic Studies 31: 191-196.

Luttmer, Erzo F.P.

(2001). Group loyalty and the taste for redistribution. Journal of Political Economy 109: 500-528.

Lynn, Richard

(2006). Race Differences in Intelligence. An Evolutionary Analysis. Augusta (GA): Washington Summit.

Lynn, Richard and T. Vanhanen

(2006). IQ and Global Inequality. Athens (GA): Washington Summit.

Lynn, Richard and D. Longley

(2006). On the high intelligence and cognitive achievements of Jews in Britain. Intelligence 34: 541-547.

Lynn, Richard, G. Meisenberg, J. Mikk and A. Williams

(2008). National IQs predict differences in scholastic achievement in 67 countries. Journal of Biosocial Science, in press.

Martin, Michael O., I.V.S. Mullis, E.J. Gonzalez, K.D. Gregory, T.A. Smith, S.J. Chrostowski, R.A. Garden and K.M. O'Connor

(1999). TIMSS 1999 International Science Report. Chestnut Hill (MA): ISC.

Martin, Michael O., I.V.S. Mullis, E.J. Gonzalez and S.J. Chrostowski

(2004). TIMSS 2003 International Science Report. Chestnut Hill (MA): ISC.

Meisenberg, Gerhard

(2007). Does multiculturalism promote income inequality? Mankind Quarterly 47: 3-39.

Meisenberg, Gerhard

(2007b). In God's Image. The Natural History of Intelligence and Ethics. Brighton: Book Guild.

Michelbach, Philip A., J.T. Scott, R.E. Matland and B.H. Bornstein

(2003). Doing Rawls justice: An experimental study of income distribution norms. American Journal of Political Science 47: 523-539.

Morris, Martina and B. Western

(1999). Inequality in earnings at the close of the twentieth century. Annual Review of Sociology 25: 623-657

Mullis, Ina V.S., M.O. Martin, E.J. Gonzalez, K.D. Gregory, R.A. Garden, K.M. O'Connor, S.J. Chrostowski and T.A. Smith

(1999). TIMSS 1999 International Mathematics Report. Chestnut Hill (MA): ISC.

Mullis, Ina V.S., M.O. Martin, E.J. Gonzalez and S.J. Chrostowski

(2004). TIMSS 2003 International Mathematics Report. Chestnut Hill (MA): ISC.

Murray, Charles

(2002). IQ and income inequality in a sample of sibling pairs from advantaged family backgrounds. American Economic Review 92: 339343.

Rushton, J. Philippe

(1989). Genetic similarity, human altruism, and group selection. Behavioral and Brain Sciences 12: 503-559.

Rushton, J. Philippe and T.A. Bons

(2005). Mate choice and friendship in twins. Evidence for genetic similarity. Psychological Science 16: 555-559

Scott, John T., R.E. Matland, P.A. Michelbach and B.H. Bornstein

(2001). Just deserts: An experimental study of distributive justice norms. American Journal of Political Science 45: 749-767.

Verster, J.M. and R.J. Prinsloo

(1988). The diminishing test performance gap between English speakers and Afrikaans speakers in South Africa. In: S.H. Irvine and J.W. Berry (eds): Human Abilities in Cultural Context. Cambridge: Cambridge University Press.

Weede, Erich

(2004). Does human capital strongly affect economic growth rates? Yes, but only if assessed properly. Comparative Sociology 3: 115-134.

Gerhard Meisenberg *

Ross University, Dominica

* Address for correspondence: Department of Biochemistry, Ross University, Medical School, Picard Estate, Dominica Email Gmeisenberg@rossmed.edu.dm
Table 1: Mean and standard deviation for Gini index and racial
diversity scores in different world regions. Racial distances are
weighted either by genetic differences or IQ. "Middle East"
includes only the Muslim-majority countries. Greece, Cyprus and
Israel are lumped with "Catholic Europe." N = number of countries.

Region                 N     Gini index

Protestant Europe      7     28.0 [+ or -] 2.9
Catholic Europe        9     32.0 [+ or -] 3.8
English-speaking       6     34.0 [+ or -] 3.0
Ex-Communist           28    34.7 [+ or -] 7.4
Latin America/Car.     23    50.5 [+ or -] 6.7
Middle East            10    37.9 [+ or -] 5.3
South/Southeast Asia   10    39.4 [+ or -] 6.6
East Asia              5     36.3 [+ or -] 6.4
Asian Communist        3     37.1 [+ or -] 1.5
Pacific Islands        1     50.4
Sub-Saharan Africa     31    48.7 [+ or -] 9.7
Total                  133   40.9 [+ or -] 10.3

Region                 Racial diversity

                       N     by genes             by IQ

Protestant Europe      8      3.2 [+ or -] 4.1     5.8 [+ or -] 4.5
Catholic Europe        11     2.2 [+ or -] 3.0     5.7 [+ or -] 10.3
English-speaking       6     22.7 [+ or -] 19.7   18.4 [+ or -] 14.7
Ex-Communist           28     2.4 [+ or -] 4.7     2.6 [+ or -] 3.9
Latin America/Car.     37    66.4 [+ or -] 27.7   52.0 [+ or -] 21.0
Middle East            24    13.9 [+ or -] 21.4    7.8 [+ or -] 10.3
South/Southeast Asia   15    19.5 [+ or -] 29.7   15.4 [+ or -] 23.0
East Asia              6      9.3 [+ or -] 14.4    8.0 [+ or -] 13.5
Asian Communist        4      0.1 [+ or -] 0.3     0.1 [+ or -] 0.2
Pacific Islands        14    22.7 [+ or -] 22.6   16.1 [+ or -] 18.7
Sub-Saharan Africa     45    12.6 [+ or -] 19.1    8.5 [+ or -] 15.2
Total                  198   21.6 [+ or -] 29.9   16.6 [+ or -] 23.0

Table 2: Correlations of the Gini index with predictors.
N = 131 countries. Educ, education; lgGDP, logarithm of
GDP; Corr., corruption; EcoFr, economic freedom; PolFr,
political freedom; Demo, democracy; RDg, racial diversity,
weighted by genes; RDi, racial diversity weighted by IQ;
lgRDg, lgRDi, logarithm of racial diversity by genes and
IQ, respectively. Correlations above .175 are significant
at p<.05, and correlations above .280 are significant
at p< .001.

        Gini    IQ      Educ.   lgGDP   Corr.   EcoFr

IQ      -.587
Educ.   -.440    .819
lgGDP   -.451    .761    .866
Corr.    .440   -.576   -.675   -.813
EcoFr   -.403    .617    .691    .779   -.860
PolFr   -.301    .514    .655    .748   -.714   .745
Demo    -.464    .689    .755    .804   -.705   .721
RDg      .372   -.102   -.018    .050    .030   .014
lgRDg    .453   -.142   -.032    .026   -.012   .043
RDi      .370   -.039    .074    .148   -.045   .080
lgRDi    .364   -.015    .113    .185   -.142   .170

        PolFr   Demo    RDg    lgRDg   Rdi

IQ
Educ.
lgGDP
Corr.
EcoFr
PolFr
Demo    .853
RDg     .135    -.029
lgRDg   .116    -.046   .861
RDi     .219     .065   .965   .854
lgRDi   .250     .122   .833   .962    .871

Table 3: The best-fitting regression models explaining
the Gini index. The standardized p coefficient is reported.
Models 5 and 6 exclude the countries of Latin America and
the Caribbean. Excom, ex-communist; lgPopDens, logarithm
of population density; Edu, education. For other
abbreviations, see legend of Table 2.
* p<.05; ** p<.01; *** p<.001.

              Model 1     Model 2     Model 3

IQ            -.390 ***   -.444 ***   -.346 **
IQ2           .194 **     .327 **     .275 ***
Education     .378 *      .166        .324 *
Education2    -.215 *     .501 *      -.194 *
lgGDP         -.322 *                 -.261
lgGDP2        -.206 *     .248 *      -.118
Corr                                  .168
PolFr2        -.257 ***   -.302 ***   -.194 **
Excom         -.391 ***   -.416 ***   -.281 **
lgPopDens     -.272 ***   -.265 ***   -.251 ***
lgRDi                                 .274 ***
lgRDi2                                .092
RDi
IQ x Edu                  -.313
Edu x lgGDP               -.912 ***
N             132         132         132
R2            .618        .637        .675

              Model 4     Model 5     Model 6

IQ            -.391 ***   -.471 ***   -.623 ***
IQ2           .403 ***    .300 ***    .663 ***
Education     .214        .446 *      .383 **
Education2    .450 *      -.188 *     .887 ***
lgGDP                     -.316
lgGDP2        .275 *      -.113       .322 **
Corr          .220        .193        .239
PolFr2        -.241 ***   -.171 *     -.211 **
Excom         -.344 ***   -.281 **    -.272 **
lgPopDens     -.254 ***   -.241 **    -.185 **
lgRDi         .228 ***
lgRDi2        .117 *
RDi                       .217 **     .136 *
IQ x Edu      -.304                   -.899 ***
Edu x lgGDP   -.768 **                -.776 **
N             132         109         109
R2            .707        .648        .733

Table 4: Correlations of test score variability
(residuals with mean score) on the TIMSS and
PISA assessments (vTIMSS and vPISA), and both
assessments combined (vTIMSS/PISA). Racial
diversity (RD) is weighted either by IQ or by
genetic distance, and is used either directly
in logarithmic transformation. * p<.05; ** p<.01.

               vTIMSS   vPISA     vTIMSS/PISA

RDgenes        .136     .204      .212
lgRDgenes      .078     .252      .210
RDiq           .214     .300 *    .292 *
lgRDiq         .144     .372 **   .317 **
Gini index     .060     .172      .181
IQ             .015     .147      .067
Education      .052     .281 *    .181
lgGDP          -.135    .328 *    .109
Corruption     .171     -.211     -.068
Pol. Freedom   .107     .290 *    .256 *
Democracy      .075     .405 **   .286 *
N              46-51    52-55     64-72

Table 5: Racial diversity as a predictor of ability
dispersion in TIMSS and PISA (vTIMSS and vPISA),
and the combined results from these two assessments
(vTIMSS/PISA). sqrtArea, square root of the country's
surface area; lgPopDens, log-transformed population
density; RDiq and lgRDiq, untransformed and
log-transformed racial diversity, weighted by IQ.
Standardized betas are reported.
* p<.05; ** p<.01; *** p<.001.

                 vTIMSS      vPISA       vTIMSS/PISA
                 predicted   predicted   predicted

IQ2              .273                    .142
Education        .427 *      .415        .498 **
lgGDP2           .142        .188        .191
Corruption       .538 *      .669 **     .398
Corruption2                  -.508 ***   -.244 *
Pol.Freedom2     -.257                   -.325 *
Democracy                    .372 *
Yearscommunist               -.362 *     -.317 *
RDiq             .307 *
lgRDiq                       .408 **     .275 *
sqrtArea                     -.241
lgPopDens        .146        -.294 *
R2               .410        .550        .396
N                49          53          69

Table 6: Big Government predicted by
racial diversity and other variables.
Model 1 is for all countries with a
Gini index for which complete data are
available. Model 2 is the corresponding
model limited to countries outside
Latin America and the Caribbean. The
standardized p is reported.
* p<.05; ** p<.01; *** p<.001.
Communist: History of communist rule.
RDgenes, racial diversity weighted by
genes. EcoFr, economic freedom. lgPopDens,
logarithm of population density.

             Model 1     Model 2

IQ           -.317 *     -.270 *
lgGDP        .863 ***    1.042 ***
lgGDP2       .406 ***    .370 ***
EcoFr        -.493 ***   -.711 ***
EcoFr2       -.233 **    -.311 ***
Democracy    .231        .301 *
Democracy2   .207 **     .240 **
Communist    .102
lgPopDens    -.264 ***   -.258 ***
lgRDgenes    -.288 ***
RDgenes                  -.158 *
RDgenes2                 .149 *
N            131         108
R2           .513        .579

Table 7: Big Government as a predictor of the Gini
index. lgRDiq, logarithm of racial diversity,
weighted by IQ; lgPopDens, logarithm of population
density; Edu, education; lgGDP, logarithm of GDP.
N = 131. * p<.05; ** p<.01; *** p<.001.

                 Model 1     Model 2     Model 3

IQ               -.467 ***   -.431 ***   -.446 ***
IQ2              .206 **     .258 ***    .235 **
Education        .354 **     .267 *      .276 *
Education2       -.236 **    -.224 **
lgGDP2           .143
Corruption       .179        .209        .182
Pol. Freedom2    -.218 ***   -.181 **    -.183 **
lgRDiq           .189 **     .169 **
lgRDiq2          .075        .086
Excommunist      -.332 ***   -.245 **    -.292 ***
lgPopDens        -.344 ***   -.318 ***   -.329 ***
Big Government   -.299 ***   -.232 ***   -.214 **
IQ x lgPopDens                           .086
Edu x lgGDP                              -.382 ***
R2               .677        .701        .722
COPYRIGHT 2014 Council for Social and Economic Studies
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2014 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Meisenberg, Gerhard
Publication:The Journal of Social, Political and Economic Studies
Article Type:Report
Geographic Code:1USA
Date:Mar 22, 2014
Words:7739
Previous Article:One man's fight for stewardship and professionalism: a lesson in business ethics.
Next Article:The World Until Yesterday: What Can We Learn from Traditional Societies?
Topics:

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters