Catholic School Effects on Achievement: New Insights From Machine Learning Techniques

• In Event: 34.095 - MTCC Roundtable Session 11
  In Roundtable Session: 34.095-7 - International Perspectives on School Choice and Market Reform

Abstract

Do Catholic schools provide superior learning environments than do their public-school counterparts? Two generations of researchers have investigated this question in the United States, with no consensus. In this paper, we use a machine learning technique called random forests to estimate Catholic school effects in Ontario, Canada. We argue that random forests can accommodate complex interactions that underlie student learning and that Catholic and public schools are more readily comparable in Ontario than in the United States. We use achievement data from Grade 3 provincial assessments in reading, writing, and math, along with an extensive set of controls, for 32,317 students. Our results suggest that there are no Catholic school effects in Ontario.

Authors

• Zahide Alaca, University of Toronto

• Scott Davies, University of Toronto - OISE