When Do Informational Interventions Work? Experimental Evidence From New York City High School Choice

Sarah R. Cohodes, Sean P. Corcoran, Jennifer L. Jennings, Carolyn Sattin-Bajaj

Research output: Contribution to journalArticlepeer-review

1 Scopus citations

Abstract

Despite evidence that informational interventions can influence K–12 school choices, we know little about the mechanisms through which they work and the factors that produce heterogeneity in student responses. Through a school-level randomized controlled trial conducted in 473 New York City middle schools serving 115,000 eighth graders, we evaluated three counselor-delivered informational interventions that were designed to help students avoid low-graduation high schools, but differed in their level of individual customization and mode of delivery (paper or online). Every intervention reduced likelihood of application to and enrollment in low-graduation-rate schools (those below the city median of 75%). Simplified paper interventions had the largest impacts and produced lower heterogeneity in effects across subgroups than customizable digital formats. A key mechanism by which interventions worked was through new information replacing students’ default first-choice schools that had low graduation rates and guaranteed admission. We conclude that informational interventions to support school choice can be effectively implemented at scale via school counselors, but that intervention design can lead to differences in who engages, with consequences for inequality.

Original languageEnglish (US)
JournalEducational Evaluation and Policy Analysis
DOIs
StateAccepted/In press - 2023

All Science Journal Classification (ASJC) codes

  • Education

Keywords

  • admissions
  • counseling
  • economics of education
  • educational policy
  • experimental research
  • high schools

Fingerprint

Dive into the research topics of 'When Do Informational Interventions Work? Experimental Evidence From New York City High School Choice'. Together they form a unique fingerprint.

Cite this