Essays in Labor Economics and Political Economy

Abstract

Each of the three chapters of this dissertation analyzes an empirical question bearing on social policy. In the first chapter, I assess if there was differential selection between white and black men into World War II Army enlistment. Differential selection could represent (1) a potential artifact of well-documented historical discrimination in induction into service or (2) a potential confound in the attributing racial differences in relative postwar outcomes of veterans versus nonveterans to racial differences in the combined returns to military service plus veterans’ benefits. My analysis demonstrates that white and African American World War II enlistees appear largely representative of their birth cohorts in terms of their family background in the 1930 Census and the racial and economic exclusivity of their 1930 neighborhood, proxied for with grades on Home Owners’ Loan Corporation “redlining” maps. In the second chapter, I investigate whether political polarization is shaped by social contact among peers across racial, ethnic, income, and social and political attitudinal lines. I estimate how the characteristics of randomly-assigned freshman roommates affect students’ own attitudes. With scattered exceptions, I do not find robustly significant roommate effects. In contrast to the findings in previous studies, I find that white students assigned black roommates, as compared with those assigned white roommates, did not report more comfortable interracial interactions or higher support for affirmative action or campus diversity. But they did report lower closeness to their roommates. One possible explanation is that some white students’ friendship and deliberation with black roommates, the presumptive channels for peer influence, are impaired by implicit racial bias formed before college. In the third chapter, I examine how incarceration length affects recidivism. I estimate the joint effect of specific deterrence plus aging using quasi-random differences in average incarceration length for adults convicted of driving under the influence (DUI) and reckless driving who kill their victims as opposed to seriously injure them. Estimated impacts on the level of the recidivism rate are mainly precise zeros in Florida and on the small end of the previous literature in Georgia, while impacts expressed as a proportion of sample-mean recidivism rates are more in line with the moderately negative comparable estimates in a previous literature.