Research Report

Using Centralized Lotteries to Measure Preschool Impact

A growing number of US cities are organizing common application systems for families seeking free public preschool for their children, and many use the deferred acceptance (DA) algorithm to assign students to schools that receive more applications than they have seats.

In this report, we examine the steps taken to use a common application system and centralized lottery within the District of Columbia’s public prekindergarten program. By reconstructing and repeatedly simulating the lottery, we demonstrate that enrollment in the city’s public prekindergarten program has a large effect on families’ persistence in public schools. We conclude that the lottery evaluation methods used in the District of Columbia can be used by other cities to evaluate their own educational programs and examine the effects of public preschool.

Methodological Key Findings

  • The DA propensity score method eliminates applicant differences in the likelihood of being matched.
  • The District of Columbia’s common lottery and large-scale public prekindergarten program are well suited for evaluation using DA methods. We describe the DC context and lottery implementation details to help other cities understand whether these methods may be a good fit for them.

DC Prekindergarten Key Findings

  • Being matched to a PK3 program (public prekindergarten for 3-year-olds) causes a 30 percentage-point increase in the likelihood that a student enrolls in PK4 (public prekindergarten for 4-year-olds) the following year and leads to a 28.5 percentage-point increase in the likelihood of continuous enrollment in the system.
  • Estimated impacts for PK4 are of a smaller magnitude. Being matched to a PK4 program causes a 24 percentage-point increase in the likelihood that a student enrolls in DC Public Schools kindergarten the following year. Impacts on continuous enrollment in the system are not distinguishable from chance.
  • For PK3, on average, listing one more school is associated with a 1.1 percentage-point increase in the likelihood of being matched. This finding does not hold for PK4, for which we find no correlation between the number of schools ranked and the likelihood of a match.
  • In PK3, about 80 percent of applicants that were matched enrolled in the program, and between 50 and 60 percent of unmatched applicants eventually enrolled. PK4 applicants, both matched and unmated, enroll at higher rates than PK3 applicants.

Implications

This report follows a report on lottery applicants and application patterns in DC and will be followed by an expanded set of impact findings, building on the persistence effects identified here.