Association for Public Policy Analysis & Management

Who can read better, a fourth-grade student in Niger or one in Peru? Questions like this one have complicated answers because “standardized” tests are comparable only within a limited set of countries. We develop score conversion functions by administering an exam combining items from different assessments to 2,300 primary students in Bihar, India. Armed with this learning “Rosetta Stone,” we convert the scores of more than 600,000 students from 80 countries onto common scales, massively expanding coverage of comparable student microdata. We document five facts about education quality around the world: (i) learning differences between most- and least-developed countries are larger than existing estimates suggest, (ii) test scores rise with per-pupil education spending, (iii) national learning-income gradients correlate with income inequality, (iv) girls score lower in countries with more conservative gender norms, but the gender gap in math exists on average only for richer students, and (v) countries with higher test scores export more in skill-intensive industries.