Panel Paper: Evidence of Impact on Learning: Regression Discontinuity Evidence from the Preschool Expansion Grant

Saturday, November 10, 2018
Marriott Balcony A - Mezz Level (Marriott Wardman Park)

*Names in bold indicate Presenter

Shaun Dougherty, Tamika La Salle, Hannah Dostal and Carissa Scogin, University of Connecticut

Prekindergarten has long been believed to help children prepare cognitively and behaviorally for K-12 schooling. Existing and mounting causal evidence suggests that there is clear evidence of an impact on early learning (Phillips, et al., 2017; Weiland & Yoshikawa, 2013), and older evidence that tracks students later into life suggests that, in some cases, the effects may be large and long-lasting. Despite the existing evidence, the expansion of prekindergarten through federal incentives, as well as state and district incentives and investments, has resulted in a broadened prekindergarten landscape which has not yet been evaluated. In fact, the proliferation of prekindergarten options, funding structures, and variation in eligibility criteria and peer composition, suggests that we may know less about the potential effects of the current prekindergarten marketplace than we did only 10 years ago.

In this paper we use an age-based eligibility rule for prekindergarten and a regression discontinuity design, to estimate the learning impacts of an expanded, means-tested prekindergarten program in Connecticut. Connecticut was among a host of states to receiving federal Preschool Expansion Grant funds to increase the supply and quality of prekindergarten, particularly for families of less economic means. We leverage these conditions for a natural experiment, along with an innovative assessment schedule of testing students at the start and end of the academic school year, to estimate program impacts. Whereas prior studies have faced the challenge of non-random selection into prekindergarten using a kindergarten-based control group, we ensure that treatment effects are estimated using only those students, across school years, who were age-eligible for prekindergarten, and who all selected into state-subsidized prekindergarten. Using universally applied assessments of literacy and numeracy in our sample we are able to estimate effects on learning, and using self-selected surveys on behavior, we obtain suggestive, though more flawed estimates on behavioral outcomes.

Initial estimates suggest that the data satisfy the requirements for a valid RD design, and that impacts are suggestively positive on early literacy and numeracy outcomes. Collection of additional assessment data in spring 2018 and fall 2018 will improve estimate precision and statistical power, and enhance the abilities to make stronger inferences about program impacts.