Knowledge capital, growth, and the East Asian miracle.

W ith per-capita gross domestic product (GDP) growing by an average of 4.5% annually since 1960, people in East Asia are about nine times as prosperous as two generations ago. By contrast, the average person in Latin America is only about two and a half times as prosperous. Over the past quartercentury, both theoretical and empirical analyses of possible drivers of the different growth rates seen around the world invariably assign an important role to human capital (1–4). This has led to development policies focused on increasing enrollment and retention in schools. We argue, however, that too much attention is paid to the time spent in school, and too little is paid to the quality of the schools and the types of skills developed there. Differences in school attainment cannot account for the East Asian miracle or the Latin American puzzle. When plotting the marginal effect of years of schooling attained in 1960 (5, 6) against annual growth rates between 1960 and 2000 [see the chart, (A)], East Asian countries are systematically above the schooling-growth line, which indicates that they grew faster than expected by their school attainment. Latin American countries, lying below the line, grew slower than expected. [Details of all analyses are in the supplementary materials (SM).] Empirical analyses of growth have attempted to deal with these and other anomalies by adding geographic, institutional, and cultural factors in addition to school attainment and by taking different approaches to the form and estimation of the models. Nonetheless, skepticism about the validity of such estimates has grown because the impact of various factors has been sensitive to the specification of the model (7), because of suspicion of reverse causality, i.e., growth causes more schooling rather than the other way around (8), and because other factors, such as culture or institutions, may drive both schooling and growth (9). Our prior analysis (10–12) emphasizes linkage between growth and the skills of the population measured by cognitive tests and shows that the main questions about empirical growth models are greatly lessened once skills are appropriately measured. Here, we build upon that work to demonstrate how the apparent growth anomalies mentioned above are consistent with this perspective, and we sketch implications for current development policy.