Individual Submission Summary

Direct link:

Ethnic, Linguistic and Religious Heterogeneity and Preferences for Public Goods

Thu, August 30, 2:00 to 3:30pm, Sheraton, Back Bay D


Sarah Berens (University of Cologne) & David Brady (University of California, Riverside)

One of the most prominent claims in political economy is that social heterogeneities weaken public support for public goods and redistribution. Despite the appeal of this theoretical argument, empirical tests outside certain global regions remain surprisingly scarce. This study sheds new light on this prominent argument by making use of a large set of survey data and more rigorous statistical techniques. Latin America and the Caribbean provide a useful setting to investigate these claims given these regions feature poor and developing countries, with many heterogeneities, and active political debates about public goods and redistribution. This study tests three hypotheses about the relationship between various forms of heterogeneity and individual-level support for public goods redistribution (henceforth “preferences”): (1) heterogeneity is negatively associated with preferences, (2) heterogeneity is positively associated with preferences; and (3) heterogeneity is negatively associated with preferences among the dominant group.

We use a maximum of five waves of survey data from the Americas Barometer to examine individual-level preferences in a maximum of 24 Latin American and Caribbean countries. Based on individual-level responses aggregated and weighted into country-level measures, we calculate Herfindahl indices for ethnic (2008, 2010, 2012, 2014, 2016), linguistic (2008, 2010, 2012, 2014), and religious (2010, 2012, 2014, 2016) heterogeneity. For preferences, we asses four different questions regarding whether the government “should implement strong policies to reduce income inequality between the rich and poor” (2008, 2010, 2012, 2014, and 2016); and be responsible for “creating jobs” (2008, 2010, 2012) “providing health care services,” (2010, 2012) and “ensuring the well-being of the people” (2008, 2010, 2012).

For every outcome and measure of heterogeneity, we analyze both between-country variation with multilevel models and within-country variation with difference-in-difference models (DD, i.e. models with fixed effects for country and time). The models adjust for a variety of individual-level characteristics (i.e. sex, age, wealth, education, and employment) and country-level factors (i.e. GDP per capita, inflation, government effectiveness, and income inequality). Models also adjust for relevant individual-level ethnic, linguistic, and religious identities.

The results provide no support for the first hypothesis, but partially support the second hypothesis. Across a large number of models, zero coefficients for ethnic, linguistic and religious heterogeneity are significantly and negatively associated with any of the four individual-level preferences. Indeed, in the multilevel models focused on between-country variation, most models show ethnic, linguistic, and religious heterogeneity are significantly positively associated with individual-level preferences. Thus, the Latin American and Caribbean countries with greater heterogeneities exhibit higher public support for public goods and redistribution. In the DD models focused on within-country variation however, the coefficients for heterogeneity are positively signed but mostly insignificant.

For the third hypothesis, we interacted the country-level measures of heterogeneity with an individual-level binary measure of whether the respondent is a member of the majority (or largest plurality) group. In most models and for most outcomes and most measures of heterogeneity, the coefficients for the key interaction terms are insignificant. However, a few models reveal positive and negative interactions. Specifically, the interaction for religion is significantly positive for the public well-being (in multi-level and DD models) and healthcare (in multi-level models) preferences. We offer an interpretation of this variation in the majority-group-member effect.

In sum, our study features rigorous empirical test of a prominent political economic argument with two different modeling strategies, four different measures of heterogeneity, four different preferences, and a maximum of 24 countries and five waves of survey data. Overwhelmingly, the analyses provide little evidence that heterogeneities are negatively associated with preferences for public goods and redistribution. There is some evidence of a positive association, especially when analyzing between-country variation. There is also some evidence of a negative association between heterogeneities and preferences among the dominant group. However, the evidence of positive associations and interactions are not robust. Therefore, we ultimately conclude that heterogeneities do not have a major influence on preferences for redistribution and public goods.