Chapter 5 rigorously tests the observable implications of the argument with longitudinal analysis. I conduct two panel analyses that isolate the effects of educational expansion from the “contaminating” effects of affirmative action policies. First, drawing on annual household surveys conducted by the census bureau, I construct a synthetic panel of birth cohorts to test the hypothesis that better-educated Brazilians situated in the lower classes are mostly likely to self-darken over time. The analysis supports this hypothesis and finds that this relationship holds across diverse cultural regions of Brazil. Next, I introduce an original panel dataset of Brazilian municipalities in 2000 and 2010 to explore whether spatial variation in educational expansion causes higher rates of reclassification within Brazil. Fixed-effects analysis again supports the hypothesis, showing that greater rates of high school and university attendance correlate with greater black identification. Additional analysis indicates that the hypothesized patterns are clearest in urban centers, and are not conditional on the presence of state-level affirmative action policies.

Chapter 6 turns to affirmative action. I begin with a discussion of two affirmative action-based hypotheses, one instrumental and the other symbolic. Both hypotheses point to these race-targeted policies as explanations for the reclassification reversal. I then test these hypotheses in several ways. First, I analyze priming and list experiments to probe for evidence of strategic manipulation in response to affirmative action. Second, I return to the municipal panel dataset and conduct a difference-in-difference analysis of state-level affirmative action on identification. And finally, I analyze an original panel dataset of university students, constructed from embargoed surveys held by the Ministry of Education in Brazil, to compute difference-in-difference estimates of the effects of affirmative action usage on the identifications of university applicants. Overall, evidence is mixed and inconsistent. Evidence suggests that, as part of the broader array of policies that expanded education, affirmative action does boost the effects of education. But the reclassification reversal cannot be reduced to, nor solely explained by, affirmative action policies.

This opening chapter raises the research questions that motivate this book. It briefly introduces the state of mineral exploration in China and the ensuing impacts on the Chinese economy and society and pinpoints the puzzling existence of a contained resource curse in China. After critically reviewing the existing debate on the resource curse, it proposes an original theory about how mineral resources affect state–capital–labor relations, which can explain the empirical observations in China. This chapter lays out the roadmap of the whole book and explains the research methods and data sources for the empirical analysis in the following chapters.

Political transitions in a multiethnic setting are often marred with violence between groups. Why does ethnocommunal violence during a country’s democratic transition erupt in some places, and not others? Why does it subsequently decline? Existing explanations typically focus on the weakness of state capacity, the power struggle between political candidates around elections, or the structural factors such as economic inequality between groups. However, in many countries in transition, administrative units with similar characteristics have varying levels of violence. In this book, I build on Albert O. Hirschman’s framework of exit, voice, and loyalty, and I argue that excluded local elites mobilize violence to leverage their demands for greater inclusion in local politics during democratic transition. When these demands are accommodated, violence will decline. This chapter provides an overview of my theory, contributions to related literature, methodological approaches, key findings, and plan of the book.