Association for Public Policy Analysis & Management

At the forefront of early childhood educational research are questions about observed heterogeneities of intervention-generated impacts on short, medium and longer-run outcomes; fadeout; and later emergence of impacts. What are the pathways through which early interventions produce impacts on longer-run outcomes? The present study proposes a framework, the missing mediator problem, and an empirical investigation to tackle this question.

Among two prototypical randomized early childhood interventions, Perry Preschool Program generated large short-term impacts on IQ which faded by age 8, while other impacts emerged in adulthood, mostly on crime-related and labor market outcomes. Some have proposed that Perry initially boosted some sets of cognitive skills, which transferred thereafter onto noncognitive mediators carrying impacts across time. By contrast, Abecedarian induced effects on cognitive factors persisting throughout development and into adulthood, as well as impacts on completed schooling by age 30 and on some health measures by age 35.

Yet, in the few literature on mediating effects of early childhood educational interventions on longer-term impacts, much of the total effects are not explained by proposed mediators. Thus, one possible explanation for heterogeneity and fadeout is that important mediators have gone unmeasured. To empirically probe if this was indeed the case, a formal mediation analysis was undertaken using data from Abecedarian (which followed around hundred children from birth into mid-adulthood). Patterns of indirect effects on completed schooling by age 30 via a series of different mediating constructs (verbal ability, fluid intelligence, behavioral and emotional scales, and grade retained) were investigated.

Results indicated that the proportion of indirect effects to the total Abecedarian-generated impact on completed schooling was: a) increasing across time (about 13% of the total effect was mediated at age 8, while more than 50% of the total effect was mediated at age 15); b) diverse in its composition; c) not due to early impact on cognitive skills. Findings therefore suggest 1) the presence of an upstream missing mediator; and 2) key role of foot-in-the-door process grade retained as downstream mediator for the persistent effects generated by the program.