Selected for Attainment: Estimating Early Child Health Effects on Adult Outcomes

Robert G. White, University of Wisconsin at Madison

Sample selection bias is a chronic problem in longitudinal studies but one which is only rarely addressed in studies concerning ties between health and socioeconomic status. This paper adopts two alternate methods for handling sample selection bias. Both methods are applied to examine the magnitude of bias in the effects of early health on educational attainment and adult socioeconomic status. A sample selection model and a multiple imputation method for handling missing data are each adapted to account for different types of nonresponse and the possibility for changing severity in sample selection bias over time. Estimates from both methods are significantly larger than effects in the absence of any treatment for sample selection. The results demonstrate the importance of explicitly taking account of sample selection bias and suggest the possible magnitude of downward bias in standard estimates of socioeconomic gradients in health.