Skip to main content
Open Access Publications from the University of California

Estimating Preschool Impacts when Counterfactual Enrollment Varies: Bounds, Conditional LATE and Machine Learning

  • Author(s): Berkes, Jan
  • Bouguen, Adrien
  • et al.

We study the impact of preschools and the issue of close substitutes in a Cambodian context where newly built formalized preschools are competing with existing alternative early childcare arrangements. In addition to estimating the reduced-form impact of a vast preschool construction program using a random assignment, we implement several empirical techniques to isolate the impact on children who would have stayed at home if they had not been enrolled in the newly built preschools. We argue that this parameter is both critical for the preschool literature and, because it does not depend on the quality of alternative preschool, is often the only parameter that can be comparable across studies and contexts. Our results show that after one year of experiment, the average intention-to-treat impact on cognitive and socioemotional development measures is significant but small in magnitude (0.05 SD). Our analysis, however, suggests that the impact on the children who would have stayed at home will likely be high and significant, and can be bounded, under a set of reasonable assumptions, between 0.13 SD and 0.45 SD. Under heavier assumptions, we have evidence that the impact on the children who would have stayed at home is around 0.2 SD, closer to our low bound. In a context where infrastructures are improving in low-income countries, our analysis suggests that accounting for close substitutes is crucial to produce more external valid statements on programs’ performance and make appropriate policy recommendations.

Main Content
Current View