Early childbearing is often cited as an important cause of adverse socioeconomic outcomes. Policies aimed at reducing such adversity, however, have been accused of exacerbating the problems they are intended to remedy. Empirical assessment of the true effects of early fertility and welfare incentives are problematic due to unobserved differences across individuals, and the ability of individuals to influence the timing of their births. These analytical difficulties stem from an inability to perform controlled social experiments. Twin births occur randomly, and hence provide a natural experiment for evaluating the effects of early fertility and welfare policies. The primary impediment to the use of this approach, the rarity of twin births, has been removed by our algorithm that links children to their mothers and identifies twin births in the Census Public Use Microdata Samples. Preliminary analyses using this approach have provided important insights into the effects of early childbearing. Many open questions remain, however. This project seeks to resolve those questions by adding data from internal Census files to the analysis samples constructed from public use data. Whether estimates from the twins approach are representative of the effects of early first births has been called into question. We provide conditions under which the results generalize, and propose a plan by which to test those conditions. Finally, relatively little is known about the initial unmarried spells of women who give birth out of wedlock. The duration of such spells, however, is of clear importance for the well-being of the mother and her children. We propose to provide a baseline statistical analysis of these spells, then utilize our twin birth data to estimate the effect of welfare benefits on their mean duration.
