The socioeconomic costs of teenage childbearing: evidence and interpretation.

There is little evidence to support the reasons suggested by Hoffman et al. for treating the results based on the NLSYW as outliers. There is even some evidence that might lead one to favor the NLSYW estimates. After some investigation, which of the range of within-family estimates across data sets is most accurate remains unsettled (although exploring differences in cross-sectional estimates from the sisters subsamples seems promising). In addition, we believe there is evidence to support the hypothesis that within-family estimates are upwardly biased because of within-family heterogeneity and endogeneity, but the importance and magnitude of such bias is also an open question. Although we have highlighted here what we believe to be the main points of disagreement between ourselves and Hoffman et al., we hope readers will not lose sight of the areas of agreement between the two studies, which are substantial, or of the empirical support for our key findings that Hoffman et al.'s replication study has provided. To us, the findings of both studies suggest that future research should account empirically for potential biases from (possibly unmeasured) heterogeneity in family background. Because the prevailing beliefs about the consequences of teen childbearing have been based on cross-sectional comparisons that lack detailed family background controls, these beliefs now should be open for reconsideration and should be subjected to reevaluation. Several recent empirical attempts have been made to take heterogeneity or endogeneity bias into account. These studies support this conclusion and caution against drawing causal inferences from existing estimates of the effects of teen births. We continue to recognize the limitations of currently available methods and data for accounting for unobserved heterogeneity and selectivity (e.g., Griliches 1979; Manski 1989). Therefore we encourage the enhancement of data sets and the continued empirical investigation of questions that have been raised about possible biases of sibling estimation and other methodological approaches. We hope that with new rounds of research, advances will continue to further the understanding of these important social processes. Given the difficulty of accounting adequately for selection into teen childbearing across and within populations, and even within families, and given the conflicting within-family estimates, we believe that the size of any "true effects" of teen births on socioeconomic status must be considered an open question.