Correcting Sample Selection Bias in PISA Rankings

Abstract

This paper proposes a method to account for sample selection (survivor bias) in cross-country comparisons. International assessments such as the Programme for International Student Assessment (PISA) observe outcomes only for students enrolled in school at age 15, which can distort comparisons in countries with high dropout rates. I consider a quantile-based selection correction that delivers bounds on countries' average performance rather than point estimates. Using these bounds, I construct optimistic and pessimistic rankings, with each country's true rank lying between the two. An application to PISA 2018 shows that correcting for selection bias leads to substantial changes in countries' average scores and rankings.