and participants in the Workshop on Low Income Populations at the Institute for Research on Poverty at the University of Wisconsin-Madison for helpful comments on an earlier draft and Frances Margolin, Hoda Makar, and Simon Hotz for their editorial assistance. We especially wish to thank Robert Willis for numerous helpful discussions during the course of this study. All remaining errors are the responsibility of the authors. 8/27/99 Teenage Childbearing and Its Life Cycle Consequences: Exploiting a Natural Experiment In this paper, we exploit a “natural experiment ” associated with human reproduction to identify the effect of teen childbearing on subsequent educational attainment, family structure, labor market outcomes and financial self-sufficiency. In particular, we exploit the fact that a substantial fraction of women who become pregnant experience a miscarriage (spontaneous abortion) and thus do not have a birth. If miscarriages were purely random and if miscarriages were the only way, other than by live births, that a pregnancy ended, then women, who had a miscarriage as a teen, would constitute an ideal control group with which to contrast teenage mothers.

participants at the June preconference and December conference for their helpful suggestions and

numerous discussions. We’d also like to thank participants of the University of Michigan Population Studies Center workshop and the Applied Micro Lunch at the University of Maryland for helpful comments and suggestions. Abstracts This paper investigates to what extent the observed correlation between adolescent fertility and poor maternal educational attainment is causal. Semi-parametric kernel matching estimator is applied to estimate the effects of teenage childbearing on schooling outcomes. The matching method estimates the conditional moments without imposing any functional form restrictions and attends directly to the common support condition. Using data from the NLSY-79, kernel matching estimates suggest that half of the cross-sectional educational gaps remains after controlling for individual and family covariates. The difference between matching estimates and regressionbased estimates implies that part of the conditional difference in parametric models is due to the functional assumption. The robustness check following Altonji, Elder, and Taber (2005) reveals that a substantial amount of correlation is required within a parametric framework to make the negative effect of teen motherhood on educational attainment go away. Further evidence obtained by simulation-based nonparametric sensitivity analysis suggests that the matching estimates are quite robust with regard to a wide range of specifications of the simulated unobservables. 1 1

Abstract: Teen out-of-wedlock mothers have lower education and earnings than peers who have children later. This study uses the National Educational Longitudinal Survey of 1988 (NELS) to examine the extent to which the apparent effects of out-of-wedlock teen fertility are not causal, but are due to pre-existing disadvantages of the young women and their families. We use a novel fixed-effect matching method to study this problem. We find that mothers-to-be were substantially disadvantaged before their teen out-of-wedlock fertility. At the same time, we cannot rule out that out-of-wedlock fertility reduces education substantially, although far less than the cross-sectional comparisons of means suggest. Acknowledgment: Paul Gertler and Bryan Lincoln were quite helpful. The second author will make code available to interested replicators. Our most serious social problem [is] the epidemic of teen pregnancies and births where there is no marriage.-- President Clinton, 1995 State of the Union Address While teen mothers are very likely to live in poverty and experience other forms of adversity, our results imply that little of this would be changed just by getting teen mothers to delay their childbearing into adulthood.-- Hotz, Sanders and McElroy, 1999 As the authors of both of the quotations above agree, teen mothers have lower average education and earnings than peers who have children later. At the same time, several studies find that

The difficulties experienced by young women who give birth as teenagers have been well-documented (Maynard, 1997). For instance, teen mothers are far less likely to work compared to other women who delay child-bearing. Their lower earnings places the family at far greater risk of falling into poverty. Teen mothers are also more likely to rely on public assistance to provide the family with a minimal living standard. These and other apparent costs of teen childbearing have led to the proposal of several policy alternatives designed to alleviate them. The three demonstrations projects discussed in this book represent such alternatives. This paper will synthesize the evidence from the existing nonexperimental literature regarding the labor market outcomes and public assistance receipt of teen mothers and the results of these three experiments. The nonexperimental literature has attempted to determine whether teen motherhood itself “causes ” the outcomes that follow, or whether the characteristics of women who give birth as teens are such that they would have experienced subsequent difficulties even in the absence of the birth. Recent research suggests that perhaps little, if any, of the worse labor market outcomes of teen mothers is directly attributable to giving birth as a teen. The experimental findings, which show policies directed at improving labor market outcomes for teen mothers are, at best, only modestly effective, seem to coincide with these

for helpful comments on an earlier draft and Frances Margolin, Hoda Makar, and Simon Hotz for their editorial assistance. We especially wish to thank Robert Willis for numerous helpful discussions during the course of this study. All remaining errors are the responsibility of the authors. The views expressed herein