We use a novel method to measure racism at both the individual and the country level. We show that our measure of racism has a strong negative and significant impact on economic development, quality of institutions and education. We then test different hypotheses concerning the origin of racism and its channels of impact in order to establish causality. We find that racism is not correlated with any possible measure of coexistence of different racial or ethnic groups, like ethno-linguistic fragmentation, share of migrants, or ethnically-motivated conflicts among others. Racism has a negative effect on social capital measured as generalized trust and voice and accountability. More importantly, we show that for former colonies, racism is strongly correlated with the presence of extractive institutions during the colonial time, even when we control for current institutions, current GDP per capita or current education. We argue that extractive colonial institutions not only had a negative impact on the political and economic institutions of the colonized countries, but also shaped the cultural values of the population. We claim that colonial powers instilled racism among the population of their colonies in order to weaken their ability for collective action, justify their own role as extractive elite in the eyes of the ruled and facilitate the internal cohesion of the elite. We also show that, at the individual level and using country fixed effects, racism is negatively correlated with those cultural values that one would expect if an extractive elite would be able to decide the cultural values of the society they control: lower trust, higher obedience, lower respect for others, lower feeling of control of one's live, lower preference for democracy, higher support for military intervention of the government, lower preference for political participation, lower valuation of civil rights, higher preference for state intervention in the economy, lower support for economic competition, and higher acceptance of dishonest behavior. We finally show that racism still has a significant impact on our outcome variables even when we control for these potential cultural correlates.
