Association for Public Policy Analysis & Management

Substantial research suggests that high-quality preschool programs improve children’s school readiness and reduce their behavioral problems, especially for low-income/disadvantaged children (Gormley & Gayer, 2005; Duncan & Magnuson, 2013; Reynolds et al., 2011; Schweinhart, 2005). Because resources are scarce, policy makers must consider whether to fund one year of preschool for a larger number of children or two years of preschool for a smaller number of children. Research has examined the effects of program duration (Arteaga et al., 2014; Reynolds et al., 2011; Tarullo, Xue & Burchinal, 2013; Wen et al., 2012), but program duration effects can be confounded with starting time effects because children start preschool at different ages; children who attend for two years usually start at age three and children who attend for only one year start at age four.

Our research overcomes this challenge by examining the effect of participating in a second year of Head Start among children who all started at age four. To do this, we use an identification strategy that exploits state cutoff dates for kindergarten-age eligibility. Not all children start kindergarten in the year they turn five, because they might have missed the age cutoff date, which varies by state. Children that are very close to the cutoff date, but on different sides of it, should be very similar to each other, as if they had been randomly assigned to either side of the cutoff date. However, children who are just shy of the cutoff date cannot start kindergarten and will probably continue in Head Start for an additional year. In this way, the exogenous state rule creates a “treatment group” of children who attend two years of Head Start and a “control group” who attend only one year. Because we do not have perfect compliance, we use a fuzzy regression discontinuity design.

The data for the study comes from the Early Childhood Longitudinal Study-Birth Cohort (ECLS-B), which provides information on primary child care arrangement, Head Start preschool participation, number of days and hours/week of Head Start participation, and school readiness outcome measures (i.e., mathematics and reading scores), as well as comprehensive individual, parental, teacher and center variables that can be used as controls. About 25% of low-income children attend Head Start, which provides a suitable sample size for our analysis.

Preliminary results show that a second year of Head Start participation increases students’ kindergarten reading and mathematics scores, a finding which is robust to different specifications and bandwidth sizes. In further analysis, we will: 1) expand the results by examining the data by subgroups (e.g., part-time and full-time Head Start participants) and 2) conduct robustness checks by testing the effect of a second year of preschool for children who attend other types of preschools. Because our analytic approach will disentangle the confounding effects caused by age at start of preschool, our results will provide new, clearer evidence about the effects of program dosage. Policymakers can use these findings to make more informed, evidence-based allocations of scarce resources to different preschool options.