A substantial empirical literature on the effect of welfare benefits on fertility, marriage, cohabitation, and family structure in general has grown up over the last decade. The central tendency of this literature tends to suggest that welfare benefits have some effect in the direction many have hypothesized (e.g., higher welfare benefits leading to lesser marriage rates and greater nonmarital fertility rates) but there is considerable uncertainty surrounding the magnitude of the effect, and possibly even its existence, because of the large dispersion in estimates across different studies. This lack of consensus in research findings has weakened the impact of the research on policy-makers, who often regard the academic research as inconclusive or, alternatively, as capable of supporting any conclusion. This project will address this problem by conducting a series of replication, robustness, and reconciliation studies in an attempt to narrow the differences across past research efforts and in an attempt to build a consensus for the existence and size of effects. The project will be divided into a primary phase in which differences across studies are explained and a secondary phase in which """"""""best"""""""" models are chosen using statistical specification tests and new specifications tested. In the major and primary phase, the estimates from several major studies will first be replicated using the same data sets, models, and estimating techniques used in those papers;then the robustness of the estimates will be ascertained by estimating the models used in each paper on the data sets used in the other papers, using to the greatest extent possible the same samples, variable definitions, and econometric methods;then the differences between studies will be traced to differences in each of these factors-differences in the underlying data sets for the same variables;in variable definition;in subsampled populations;and in model specification and econometric method. The secondary phase will consist of specification tests to choose among the alternative model specifications as well as a more detailed examination of several hypotheses involving the imporance of family background, the influence of male and female wages, and alternative statistical specifications.
