Bridging Disciplinary Boundaries (January 11 - 14, 2007)

As a large-scale, publicly-financed program since the mid-1960s, Head Start has been an important topic for evaluation in public policy arena. Selection bias is a critical issue since both observable and unobservable factors could affect selection into Head Start and outcomes of participants. Findings on Head Start impacts in earlier studies were inconsistent. Some recently well-designed studies found positive effects in early school years and persistent to adolescence and adulthood. However, previous findings are virtually compared to a mixed counterfactual: Although the treatment group received Head Start, the control group received childcare ranging from parental care to other center-based care. Additionally, information about Head Start participation was collected retrospectively in almost all previous studies, which might be biased due to recall errors.

This study uses data from Early Childhood Longitudinal Study Kindergarten Class of 1998-99 (ECLS-K) to examine the effects of Head Start participation in pre-k on child school achievement (test scores in math, reading, and general knowledge) and social development (social skills and behavioral problems) in fall kindergarten. With reliable Head Start participation data verified by the Head Start Verification Study, this study uses propensity score matching method to simulate random-assignment and create treatment/control groups based on children's propensities of attending Head Start programs in pre-k. OLS regressions, including school and teacher fixed effects models, are conducted among the comparable groups with matched propensity scores. Covariates include pre- and post-treatment characteristics of child, parents, family, and teacher. The reference groups are children who exclusively received parental care only, relative care, non-relative care, and other center-based care.

The results show that, compared to children in parental care only and relative care groups, Head Start participants tend to have higher math scores (effects size of 0.15 to parental care only and 0.28 to relative care) but more behavioral problems (effect size of 0.15-0.31 to parental care only and 0.37-0.65 to relative care). Compared to those who received non-relative care, Head Start participants are likely to have lower scores in math (effect size of 0.42), reading (effect size of 1.26), and social skills (effect size of 0.52), and more behavioral problems (effect size of 0.69-0.91). Compared to other center-based care, Head Start participants tend to have lower scores in math (effect size of 0.10) and reading (effect size of 0.17-0.77), but higher social skills (effect size of 0.34).

Few previous studies have examined the effects of Head Start on children' school achievement and social development shortly after participation, controlling for other childcare arrangements and both pre- and post-treatment factors. Using advanced statistical methods, this study demonstrates that, although Head Start programs tend to increase school achievement in early kindergarten year compared to parental care only and relative care and social skills compared to other center-based care, non-relative care and other center-based care are likely to do better in promoting school achievement. Head Start participation is likely to increase children's behavioral problems than other childcare arrangements. Therefore, Head Start programs should aim to improve participants' school achievement and to reduce their behavioral problems.