Studies examining welfare‐to‐work program effectiveness present mixed and sometimes discrepant findings, partly due to research design, data, and methodological limitations. Using administrative data on Missouri and North Carolina welfare recipients, we substantially improve on past estimation approaches to identify the distinct effects of each state’s welfare‐to‐work subprograms—assessment, job search assistance and job readiness training, and more intensive programs designed to augment human capital. More intensive training is associated with greater initial earnings losses but also greater long‐run earnings gains. The negative program impacts we observe in quarters immediately following participation turn positive by the second year after participation.

This article examines match bias arising from earnings imputation. Wage equation parameters are estimated from mixed samples of workers reporting and not reporting earnings, the latter assigned earnings of donors. Regressions including attributes not used as imputation match criteria (e.g., union) are severely biased. Match bias also arises with attributes used as match criteria but matched imperfectly. Imperfect matching on schooling (age) flattens earnings profiles within education (age) groups and creates jumps across groups. Assuming conditional missing at random, a general analytic expression correcting match bias is derived and compared to alternatives. Reweighting a respondent‐only sample proves an attractive approach.