Does Diversity Drive Down Trust?

1. Does Diversity Drive Down Trust? Eric M. Uslaner Professor of Government and Politics University of Maryland College Park, MD 20742 http://www.bsos.umd.edu/gvpt/uslaner

2. There is a controversy about the effect of diversity on trust. Some(esp. Marschall and Stolle, 2004) argue that diversity promotes trust byputting people into contact with others not like themselves. • Others suggest that diversity leads to lower trust. Alesina and Putnam argue that racial diversity and fractionalization leads to lower levels of trust -- because minorities are less trusting. The more diverse a society is, the more minorities it obviously has. • We know that it is easier to trust people like yourself: From social identity theory, we are predisposed to trust in-groups more than out-groups (see various works by Brewer & Messick and Tajfel).

3. This literature on trust follows from an older literature on racialcontact and threat.  Some contact literature says that interaction withpeople of different backgrounds leads to greater tolerance and out-group trust. • Other literature argues that racial contact, especially when theshare of minorities is high, may lead to increased levels of racialdiscord--this is the "racial threat" argument made in the 1940s by V.O. Key, Jr. about racialized politics in the American South, which were most vehement and nasty in areas with high shares of African-Americans--the "racial threat" argument, which has been confirmed by more recent work on voting for racist candidate David Duke in Louisiana in the 1990s).

4. Overall, there is less than overwhelming support for either argument. Nor is it clear that these two perspectives are contradictory. • First, the racial threat hypothesis does not contradict the racial contact (more positive) argument because we rarely control for context. • Racial threat seems rarely accompanied by actual sustained contact between members of the majority and minority groups. The effects of context depend upon context: Are there real opportunities for minorities and majorities to interact?

5. There are not powerful effects of diversity on trust at either the individual or aggregate levels. • At the individual level, there are at best modest relationships with the diversity of friendship networks and the diversity of memberships in clubs and trust in Putnam's Social Capital Benchmark Study. I examine both generalized trust and trust in racial and ethnic groups. • Does having a friend from a different background to your own lead to more trust overall or to trust in other ethnic or racial groups? There is little support for this thesis. Only a handful of correlations exceed .10 and none exceed .14, as shown in the following table. Of course, these measures tell us little about the frequency of contact with people of different backgrounds. • These findings are consistent with Marschall and Stolle (2004)—who find that interracial contact alone is insignificant in predicting social trust in two Detroit Area Studies.

6. Dependent VariableIndependent Variabletau-b / tau-c • Generalized trust Have black friend .015 • Have Hispanic friend .037 • Have Asian friend . .072 • Have white friend .122 • Number friends different background .046 • Trust own ethnic group Have black f riend -.022 • Have Hispanic friend -.017 • Have Asian friend .045 • Have white friend .044 • Number friends different background .008 • Trust blacks relative • to own group Have black f riend .094 • Have Hispanic friend .042 • Have Asian friend .054 • Trust whites relative to own group Have Hispanic friend .085 • Have Asian friend .066 • Have white friend .073 • Number friends different background .087 • Trust Asians relative to own group Have black f riend .111 • Have Hispanic friend .077 • Have Asian friend .133 • Have white friend .102 • Number friends different background .138 • Trust Hispanics relative to own group Have black f riend .112 • Have Hispanic friend .120 • Have Asian friend  .105 • Have white friend .044 • Number friends different background .113 Each of the measures of friendship refer to friends outside one’s own ethnic/racial group.

7. In cross-national aggregate analyses, virtually every measure of fractionalization (diversity) hasat best modest relations (negative) with trust. This holds for the traditional Eastery-Levine measure (avelf in the literature) and newermeasures by Fearon and Alesina. • Even these miniscule relationships vanish in multivariate analyses.

8. The measure of trust is an aggregate estimate from the World Values Surveys of 1990 and 1995. To increase the sample size, I imputed values for countries not included in these surveys. The variables used to impute trust are: gross national product per capital; the value of imports of goods and services; legislative effectiveness; head of state type; tenure of executive (all from the State Failure Data Set); distance from the equator (from Jong-sung You of Harvard University); and openness of the economy (from Sachs and Warner, 1997; data available at http://www.cid.harvard.edu/ciddata/ciddata.html ). • R2 = .657, standard error of the estimate = .087, N = 63.

17. Trust and Ethnic Homogeneity in the American States • An aggregated trust score (from national surveys in the 1990s) for the American states is uncorrelated with a measure of ethnic homogeneity (fractionalization) constructed by Richard Winters.

18. Generalized Trust and State Minority Population in the United States • The share of minorities in a state is more strongly related to the level of generalized trust (though even here the relationship is not strong).

19. Ethnic Homogeneity and Minority Population Share • Minority population and ethnic heterogeneity in a state are strongly related. Even a measure of fractionalization (a Herfindahl index) cannot distinguish between segregated and integrated neighborhoods at high levels of integration such as the state or the nation. If some states/nations have large minority populations, they will have higher scores on fractionalization indices regardless of whether the minorities are integrated or segregated.

20. There is one national measure, the relative segregation of minority groups within a country (estimated by the Minorities at Risk project), that matters mightily. • It is not ethnic diversity that matters by itself--but where people live within a society. When a minority group is segregated within a society, the opportunities for contact with members of the majority are limited—and hence building of generalized trust becomes more difficult. Concentrated minorities are more likely to develop a strong identity that supercedes a national sense of identification—and to build local institutions and political bodies that enhance this sense of separateness.

21. Unfortunately, obtaining this sort of data for a large number of jurisdictions (more countries, the American states) seems impossible at this time. • Marschall and Stolle (2004) do have data on neighborhood ethnic heterogeneity in Detroit—and they show that contact across racial lines does lead to greater trust, but only in racially heterogeneous neighborhoods. My findings are broadly consistent with theirs—the distribution of racial and ethnic groups, not their sheer numbers, is the key factor in mediating how ethnic and racial fractionalization shapes trust. • Ethnic or racial segregation seems to be a strong barrier to the development of generalized trust—and the effects may be particularly pronounced for the ethnic or racial minority that is likely to feel excluded from power and greater resources.

23. Generalized trust is lower in former and present Communist nations regardless of any other factor leading to trust. • The relationships with the various measures of fractionalization and diversity are all slightly higher when we eliminate present and former Communist countries from the analysis, but the differences are generally small. For Alesina’s ethnic fractionalization index, the correlation with trust rises only from r = -.320 (N = 84) to r = -.370 (N = 63). • For the residential segregation measure, the effects are much more pronounced: The correlation rises from r = -.427 (N = 68) to r = -.585 (N = 49) as in the graph in the next slide. In multivariate regressions excluding the former and present Communist countries, the impact of residential segregation on trust is significant and large.

24. Part of the reason why the ethnic segregation measure may shape trust more than the other indicators of fractionalization or diversity might lie in the fact that there is at least a moderate relationship between ethnic segregation and inequality—and inequality is the strongest determinant of generalized trust. • Using the Galbraith estimates for economic inequality for 1994 (the year with the largest number of cases), we find a moderate relationship with generalized trust—which is strengthened when we exclude the countries with the lowest levels of minority segregation (Austria, Singapore, and Jordan). The relationship is not powerful. • The logic here is that ethnic residential segregation is likely to be compounded by a lower economic status for the minority group, breeding resentment toward the majority and a low level of generalized trust.

27. However, there is little support for this proposed explanation. Of the various measures of population diversity, the group concentration measure has one of the lower correlations with economic inequality. The simple correlation is .388, compared to .465 for Alesina’s ethnic fractionalization, .456 for Fearon’s ethnic fractionalization measure, .507 for Fearon’s measure of the population share of the largest group, and .434 for the Easterly-Levine index of fractionalization. • Most measures of population diversity are related to economic inequality. Countries with a large share of minorities have less equal distributions of wealth, regardless of whether the minorities are isolated or live alongside majority groups.

28. Group Isolation and Corruption • There is one key area in which the group concentration measure of ethnic fractionalization is distinctive: Nations with high levels of group concentration are more likely to have a weak rule of law and greater corruption. The following figures show the relationships between group concentration and the Transparency International Corruption Perceptions Index for 2004 and the index of legal and property rights from freetheworld.com

30. These relationships are moderate, but they are much stronger than we find for any of the other measures of ethnic fractionalization. • For the TI 2004 Corruption Perceptions Index, representative correlations are: • Group concentration: -.485 (N = 74) • Alesina ethnic fractionalization: -.386 (N = 91) • Fearon ethnic fractionalization: -.372 (N = 88) • For legal and property rights from freetheworld.com, representative correlations are: • Group concentration: -.568 (N = 65) • Alesina ethnic fractionalization: -.420 (N = 81) • Fearon ethnic fractionalization: -.376 (N = 78)

31. For a wide range of measures of corruption and legal fairness, including several measures of contract enforcement from Alesina’s contract enforcement data set, the World Bank Governance measure of corruption in 2004, and perceptions of corruption aggregated from the Transparency International Global Corruption Barometer surveys of 2004 for grand corruption, petty corruption, and corruption in education, the parliament, the police, political parties, the registry office, utilities, and tax collection, the group concentration index had far stronger correlations than any other fractionalization or diversity measure.

33. Does Diversity Shape Trust? • The evidence is hardly conclusive, but it seems that when contact occurs between people of different backgrounds, trust may grow. However, it is unclear that contact among adults matters as much as contact among children. Uslaner (2002) finds that having a friend of an opposite race as a young person leads to lower levels of in-group trust among adults. However, there is no strong effect for adults. • When ethnic and racial groups live in separate areas, there is a lower probability of interaction between people (either young or older) of different backgrounds. This is where racial (ethnic) threat may overwhelm any positive aspects of contact—and, indeed, contact across ethnic or racial boundaries may itself be viewed as unacceptable.

34. When group concentration is high, this may lead to strong in-group trust and low-out group trust. Gambetta (1993) and Uslaner (2005) have both linked high in-group trust and low out-group trust to corruption. Ethnic group leaders may play on fears of outsiders to justify their own corruption—and this will lead to clientelistic politics and will in turn lower generalized trust.