Essays on Matching in Labor Economics

In this dissertation, I present three essays on matching and assignment in labor economics. The first chapter presents an integrated model of occupation choice, spouse choice, family labor supply, and fertility. Two key features of the model are that occupations differ both in wages and in an amenity termed flexibility, and that children require a nontrivial amount of parental time that has no market substitute. I show that occupations with more costly flexibility, modeled as a nonlinearity in wages, have a lower fraction of women, less positive assortative mating on earnings, and lower fertility among dual-career couples. Costly flexibility may induce high-earning couples to share home production, which rewards husbands who are simultaneously high-earning and productive in child care. Empirical evidence broadly supports the main theoretical predictions with respect to the tradeoffs between marriage market and career outcomes. In the second chapter, I use the University of Michigan Law School Alumni Survey to investigate the interaction between assortative mating and the career and family outcomes of high-ability women. Women with higher earnings potential at the time of law school graduation have higher-earning spouses and more children 15 years after graduation. As the earnings penalty from reduced labor supply decreased over the sample, women with higher-earning spouses and more children reported shorter work weeks and were less likely to be in the labor force. Decreasing the career cost of non-work may have the unintended result of reducing the labor supply of the highest-ability women, as their high-earning spouses give them the option to temporarily exit the labor force. The third chapter addresses specification choice in empirical peer effects models. Predicting the impact of altering composition on student outcomes has proven an unexpected challenge in the experimental literature. I use the experimental data of Duflo et al. (2011) to evaluate the out-of-sample predictive accuracy of popular reduced form peer effects specifications. I find that predictions of the impact of ability tracking on outcomes are highly sensitive to the choice of peer group summary statistics and functional form assumptions. Standard model selection criteria provide some guidance in selecting among peer effect specifications