Much of the research on early childhood education (ECE) and Head Start (HS) has focused on whether program participation affects children’s school readiness. Insufficient attention has been given to identifying the processes that may lead to better outcomes. Recent studies, however, suggest that common global and structural measures of ECE quality are not strongly predictive of children’s school readiness (Mashburn et al., 2008). Without empirical evidence about which aspects of program quality are strongly predictive children’s outcomes, policy makers struggle to identify the most efficient ways to improve underperforming Head Start programs.  This study investigates whether HS improves children’s school readiness compared with other ECE programs, and seeks to understand which dimensions of ECE quality predict larger improvements in child outcomes.   

We use data from the HSIS, combining the random-assignment design with propensity score matching to compare HS attendees with similar children from the control group who attended other (non-HS) ECE programs.  The sample is limited to children in the four-year-old cohort who are observed in spring of the Head Start year.  A nearest neighbor (n=4) with replacement matching algorithm was employed using demographic, as well as child skill and behavior measures from the baseline survey. Statistical matching was important for the analysis, because these groups of children differed on a number of important observed characteristics. The final analysis sample includes 212 ECE children matched with 774 HS attendees.

In preliminary analyses, we estimated regression-adjusted mean differences in child outcomes and program quality across the matched groups. Because the matched groups were similar at baseline, we assume that any differences in observed program quality were due to random assignment, and that differences in spring child outcomes are due to children’s ECE experiences. Results indicate that HS programs were of higher quality than other ECE program across several potentially important quality measures (ECERS-R es. 55; Arnett CIS es .29; frequency of language and literacy and math es.40-.49) as well as the provision of other social services (es .65). Despite these sizable differences in program quality, we find little to no difference in children’s outcomes. Of the three language and pre-academic skill outcomes considered, only the measure of language favored the Head Start group (es .16, other es .01-.02).  There were no significant differences found for children’s behavior outcomes (es from .02-.01). 

Next, we will estimate instrumental variable mediation models (Reardon, Unlu, Shu , & Bloom, 2012) which will examine the extent to which program quality predicts child outcomes. We expect that grantee by treatment will provide a strong set of instruments, and that we will be able to differentiate between dimensions of ECE quality in these models. Given these initial patterns of findings, we expect our results will confirm findings from earlier studies that show that these process measures of child care quality do not predict academic or behavioral outcomes for year-olds in center-based settings. We will interpret the policy and practice implications of these findings, with respect to how state and federal programs might think about quality improvement efforts.