Beginner’s Guide to Education Economics

For a long time, economists measured education by counting years. Jacob Mincer’s 1974 framework — the schooling equation that eventually landed in every labor economics textbook — treated each additional year of schooling as generating a roughly constant percentage increase in earnings. The model was empirically productive for its era and still has uses. Its central flaw is that it treats a year of school in rural Bihar as equivalent to a year of school in Seoul or Chicago, which is not even approximately true. The move away from years-of-schooling toward measured learning outcomes is the most important methodological shift the education economics field has made in the past decade, and it is far from fully absorbed.

The instrument that forced the shift was PISA — the Programme for International Student Assessment, administered by the OECD. The 2003 results, published in December 2004, covered 41 countries and tested 15-year-olds in mathematics, reading, and science. The variance was startling even among OECD members: Finnish students outscored Mexican students by an amount equivalent to roughly three full years of schooling, despite Mexico spending a significant share of GDP on education by developing-country standards. Eric Hanushek at Stanford and Ludger Woessmann in Munich have spent the past several years trying to connect these test-score differences to long-run economic growth. Their finding, which has attracted considerable argument, is that a one-standard-deviation improvement in PISA-style math and science scores is associated with roughly one percentage point of additional annual GDP growth over a sustained period. If that relationship is causal — contested, but not implausible — the stakes in education quality are vastly higher than the enrollment numbers that dominate policy discourse.

James Heckman at the University of Chicago has been making a related point from a different direction. His work traces the Perry Preschool Project, a randomized intervention in Ypsilanti, Michigan, that enrolled disadvantaged children in high-quality preschool during 1962–1967 and has since followed them into their late thirties. The returns are arresting. Participants showed higher employment, lower crime rates, higher earnings, and better health outcomes relative to the control group. Heckman’s estimate of the social return to the Perry investment is somewhere between 7% and 12% per year — well above the return to spending equivalent money on remedial programs for teenagers or job training for adults. The mechanism is not primarily cognitive. Early childhood intervention builds what Heckman calls “soft skills”: self-regulation, persistence, the capacity to sit still and try. These turn out to predict labor market outcomes at least as well as IQ scores.

The implication is uncomfortable for aid agencies and finance ministries alike. The interventions with the highest documented returns are not university scholarships or secondary school expansion programs, which are politically popular because the beneficiaries are visible and vocal. They are programs aimed at 3- and 4-year-olds whose parents have neither the time nor the political leverage to lobby for them. Budget allocation in education tends to go where the organized interests are.

In developing countries the picture has further complications. Private returns to schooling in low-income settings are generally higher than in rich countries — roughly 10% per additional year of schooling versus 7% in OECD economies, on the standard Mincerian estimates. But this masks an enormous dispersion. George Psacharopoulos at the World Bank has documented that returns are highest at the primary level in the poorest countries, which argues for a very different spending priority than the one most developing country governments actually adopt. India’s Sarva Shiksha Abhiyan, the universal elementary education program launched in 2001, is attempting to reach roughly 180 million out-of-school or at-risk children through a combination of new school construction, teacher hiring, and block grants to village education committees. Enrollment numbers have improved. Attendance and learning outcomes have been harder to shift. China made nine-year compulsory education legally universal in 1986, but enforcement in poor rural counties remained patchy well into the late 1990s; the gap between legal requirement and classroom reality is a recurring theme in Chinese education statistics.

The honest question, at the end of all this, is what interventions actually move learning rather than just enrollment. Conditional cash transfer programs like Mexico’s Oportunidades have been reasonably good at getting children through the school gate. What happens once they are inside is less clear. The teacher quality literature — not a comfortable one for unions or for governments that depend on teacher unions — suggests that within-school variation in teacher effectiveness dwarfs between-school variation. A student unlucky enough to draw low-effectiveness teachers in three consecutive years faces a deficit that is very difficult to make up later. Benjamin Friedman’s argument about growth and civic culture, which I discussed in connection with the moral consequences of economic growth, applies directly here: the political conditions for sustained education investment are themselves a function of whether ordinary families feel the economy is working for them. The Easterly-Sachs debate about aid effectiveness runs through education as much as it runs through health or infrastructure — Easterly’s skepticism about top-down targets is at least partly borne out by the gap between India’s enrollment statistics and its learning outcomes. The structural reasons India’s financial system has underperformed also apply, by analogy, to its education system: federal design, state-level political economy, and a chronic mismatch between where the money is appropriated and where the implementation bottleneck actually sits.