This paper uses a policy implemented in Chile that obliges firms to fully fund childcare costs for their female employees, but only if they hire more than 19 women. Using plant level data from manufacturing firms, we first show that this policy has had a substantially detrimental impact on the hiring of women above that threshold, in particular since the policy has become more binding, in industrial sectors that hire fewer women and in larger firms. We then use the response of firms to study whether women workers are more or less complementary to capital than men. We find that firms that avoid the legislation by having just below 20 female workers are significantly more capital intensive than firms just above the threshold. This suggests that firms that want to avoid being subject to the regulation replace women with capital but in such a way that the capital to men ratio increases. We use our estimates to calibrate a production function and find that our results are consistent with a framework where women are weakly substitutes with capital (while men are complementary) in this emerging economy’s manufacturing sector. This does not seem to be driven by a change in skill composition of the workforce. We also find some evidence of other changes: average wages and total workforce are lower for firms who hire 20 women than those who hire just below that threshold but labor productivity is unaltered. PRELIMINARY, PLEASE DO NOT CITE ∗We thank comments from seminar participants at IADB, PUC Chile and Toronto. All remaining errors are our own. †Pontificia Universidad Catolica de Chile ‡Pontificia Universidad Catolica de Chile. §Pontificia Universidad Catolica de Chile and FinanceUC.

Over the last century, resource allocations within families changed significantly, as did marriage matching patterns. College-educated women became more likely to marry (and, to a lesser extent, have children) than less educated women. A large literature documents these patterns and proposes a variety of explanations. We review this literature. Then, we provide a unified empirical framework, which can integrate these mechanisms. We demonstrate the usefulness of that framework by employing it in decennial US censuses and showing that a combination of technological changes that increased the value of children’s education and enabled more educated women to devote more time to child-rearing are consistent with multiple behavioral changes within marriage, on the marriage market, and before marriage.