Sibling difference (or family Fixed Effects, "FE") methods are a well-known strategy for addressing selectivity bias due to omitted family-level variables. However, they face concerns over efficiency, generalizability and within-family selectivity. Recent advances in Propensity Score Matching (PSM) by Arkhangelsky and Imbens (2018) provide an alternative approach to estimating treatment effects in clustered data that may address some of these concerns by utilizing family-average treatment information. Using “Add Health” and NLSY79 data, we illustrate this approach in family/sibling samples and compare cluster PSM treatment effects of teenage childbearing on years-of-schooling to family FE and conventional PSM estimates. Preliminary results indicate that the PSM cluster estimates are smaller than conventional PSM estimates, and more similar to the (nearer-zero) family FE estimates. We discuss the findings in the context of recent work on method choice and heterogeneous effects in the literature on the educational consequences of teenage childbearing.
Presented in Session 62. Fertility Timing: Causes and Consequences