ESSAYS IN DEMOGRAPHIC ECONOMICS

Abstract

In the first of the essays, I reassess the relationship between premarital cohabitation and marital instability both theoretically and empirically. It has become a stylized fact that premarital cohabitation is positively correlated with the likelihood of marital dissolution. This is counter-intuitive because economists expect that couples learn about each other during cohabitation and would only get married if they anticipated a successful marriage. One prominent explanation for the antithetic empirical evidence is self-selection of individuals with lower prospects of successful marriages into premarital cohabitation. Using U.S. data from 1988 to 2002 and duration models, I demonstrate that the positive relationship between premarital cohabitation and marital instability has weakened over time, and that the two are no longer associated with each other. A strong decline in this association within the group of more educated women drives the result. I hypothesize that a decline in the benefits of marriage has led to greater cohabitation and hence less self-selection within this group. Causal modeling using matching and panel models uncovers a negative effect of cohabitation on marital instability.

In the second essay, I investigate the effect of teenage childbearing on high school completion, and why alternative sets of instruments result in differing coefficient estimates. The three main reasons for these discrepancies are defective instruments, treatment effect heterogeneity, and multiple mechanisms by which instruments affect the treatment. I use two instruments, age at menarche and the occurrence of a miscarriage, to investigate which of these is likely to hold. While I do not find significant treatment effect heterogeneity, I found some indications that there may be problems with the instruments' validity. Furthermore, miscarriage disproportionately affects very young teenagers which are in turn more likely to drop out of high school. This may explain differences in instrumental variable estimates if age at birth is an additional explanatory variable.