ABSTRACT
Since most of the empirical studies on ethnic diversity and trust have cross-sectional designs, they only occasionally provide evidence on the causality of the observed relationship. Cross-sectional studies are susceptible to problems of reversed causality and unobserved heterogeneity. Most correlative studies assume that ethnic diversity impacts trust and not the other way around. However, the opposite may be true as well. People who exhibit high levels of trust may differ in comparison to those with low levels in their preferences for or in the importance they attach to the ethnic composition of their environment, resulting in the self-selection of individuals who are high and low in trust into different settings. At the same time, contextual characteristics that are correlated with ethnic diversity or minority group size—such as education levels, unemployment rates, poverty, and crime rates—may be what actually brings about the observed relationships. Although regression analyses control for the impact of related contextual characteristics, standard analyses of cross-sectional correlative data cannot completely rule out the possibility that the observed associations between diversity and social cohesion are spurious. Multicollinearity problems, the restricted availability of official statistics, and the uncertainty about important impact factors make cross-sectional studies susceptible to biases.
