Association for Public Policy Analysis & Management

Research has shown the positive effects of pre-K on student outcomes; however much of this has focused on means-tested programs like Head Start, which target services to low-income students and families (Deming, 2009; Ludwig & Miller, 2007; US DHHS, 2010). More recently, many states and large cities such a Boston and New York have moved toward providing universal access programs (e.g. Weiland & Yoshikawa, 2013) highlighting the importance of understanding how publicly-provided programs function at scale.

One type of publicly-provided pre-K is provided by the school district and offered in the system’s K-12 schools, usually offered in the presence of many other early childhood options. Prior research has shown that understanding this counterfactual has implications for early childhood education more broadly, particularly in the study of Head Start (Feller, Grindal, Miratrix, & Page, 2016). This study investigates the effect of attending one such example of a large-scale district-provided public pre-K program with a variety of counterfactual options with the following research questions: (a) what is the average impact of winning a lottery to attend a district-provided pre-K program? (b) how do the effects of pre-K attendance vary based on the availability of counterfactual early learning opportunities?

Data is drawn from cohorts of students who began pre-K (or were eligible for pre-K) in the 2008-2009 school year and extends as far as the data will allow, including some information from the 2017-2018 school year. Descriptively, students who attend district pre-K are more likely to be Black, Hispanic, English-language learners and eligible for free/reduced lunch. These differences highlight the need for using lotteries to identify a randomized sample. Based on parent self-reports, students who did not attend district pre-K experienced a wide variety of early childhood experiences including Head Start, private pre-K, other publicly funded pre-K, and no formal pre-K.

Causal identification draws on lotteries that exist in many of their oversubscribed pre-K programs. Across the years included in the analysis, approximately 65% of programs used lotteries to determine pre-K enrollment, and these programs were well distributed geographically across the district. This final analysis included 180 lotteries, with 6785 students having been identified as taking part in those lotteries. The pre-K attenders represent 62% of the lottery sample and the waitlist represents 38% of the lottery sample. Using the lottery sample, analysis was conducted using a two-level hierarchical model with random intercepts at the school level, and fixed effects for lottery blocks, and a vector of student covariates.

While we find no significant effect on student outcomes in Kindergarten, the third grade results are mixed, with one null result and one significant negative effect (ES = -0.11, p<0.05). Additional analyses indicate that impacts differ across administrative regions; these regions also vary in the types of pre-K programs available to students beyond the district-provided pre-K. Next steps will use more formal models to estimate the effects of the district-provided pre-K relative to specific counterfactual options.