Commission on Growth and Development Cognitive Skills and Economic Development Eric A. Hanushek Stanford University in conjunction with Ludger Wößmann University of Munich and Ifo Institute
Overview 1. Evidence on cognitive skills and economic growth a) Issues of Endogeneity b) Institutions and Education Quality c) The Implications of Improved Education Quality 2. The skill deficit of developing countries 3. What might we do? (later)
Education in Growth (Theory) 1. Augmented neoclassical growth theories Mankiw, Romer, and Weil (QJE 1992) 2. Theories of endogenous growth Lucas (JMonE 1988); Romer (JPE 1990); Aghion and Howitt (1998 3. Theories of knowledge diffusion Nelson and Phelps (AER 1966); Benhabib and Spiegel (JEconGro 2005)
Cognitive Skills: International Student Achievement Tests Measuring knowledge, not sitting in the classroom International agencies have conducted many international tests of students performance in cognitive skills since mid-1960s 12 testing occasions 36 separate test observations (age levels, subjects) Require rescaling to obtain combined measure Adjust mean and variance of separate
International Achievement Source: Hanushek and Wößmann (2007).
International Achievement
International Achievement
Education Quality and Economic Growth
Quantity of Schooling With quality control Without quality control
Additional Details Institutions Details of distribution Minimal skills Rocket scientists
Growth of GDP/pop, 1960-2000 (1) (2) (3) a (4) GDP per capita 1960-0.379-0.302-0.277-0.351 (4.24) (5.54) (4.43) (6.01) Years of schooling 1960 0.369 0.026 0.052 0.004 (3.23) (0.34) (0.64) (0.05) Test score (mean) 1.980 1.548 1.265 (9.12) (4.96) (4.06) Openness 0.508 (1.39) Protection against expropriation 0.388 (2.29) Constant 2.785-4.737-3.701-4.695 (7.41) (5.54) (3.32) (5.09) N 50 50 50 47 R 2 (adj.) 0.252 0.728 0.741 0.784 Dependent variable: average annual growth rate in GDP per capita, 1960-2000. t-statistics in parentheses. a Regression includes five regional dummies. Source: Hanushek and Wößmann (2007).
Distribution and Institutions (9) (10) GDP per capita 1960-0.287-0.297 (5.12) (5.64) Years of schooling 1960 0.022-0.031 (0.28) (0.41) Share of students above threshold of 400 2.732 (3.61) Share of students above threshold of 600 12.880 (4.35) Test score (mean) 0.942 (2.30) Openness 0.732 (2.13) Test score * openness 1.609 (2.34) Constant 1.335 3.814 (2.97) (11.24) N 50 47 R 2 (adj.) 0.719 0.785 Dependent variable: average annual growth rate in GDP per capita, 1960-2000. t-statistics in parentheses. Source: Hanushek and Wößmann (2007).
Education Quality and Effect of test score on growth 3 Openness 2.5 2 1.5 1 0.5 0 Openness 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Estimated effect of test scores on average annual rate of growth of real GDP per capita in 1960-2000, depending on degree of openness to international trade of a country. Source: Hanushek and Wößmann (2007).
Issues of Endogeneity 1. Schooling may reflect other attributes of the economy Strong relationship even if East Asian countries are excluded 2. Reverse causality International test scores not systematically related to resources
Expenditure-Performance Math performance in PISA 2003 550 500 450 Slovak Rep. Czech Rep. R 2 = 0.01 Hungary Poland R 2 = 0.15 Korea Ireland Greece Finland Japan Netherlands Canada Belgium Switzerland Australia Iceland Denmark Sweden France Germany Austria Norway Spain USA Portugal Italy 400 Mexico 350 0 10000 20000 30000 40000 50000 60000 70000 80000 Cumulative educational expenditure per student
Issues of Endogeneity Simply identifying East Asian countries Strong relationship even if East Asian countries are excluded Reverse causality International test scores not systematically related to resources Other efficiency or cultural aspects Immigrants into U.S.
What explains U.S. growth? Favorable economic institutions Early substitution of quantity for quality Higher education
Education Quality and Growth, 1960-2000: Sub-samples (5) (6) (7) (8) Developing countries a OECD sample Low-income countries b High-income countries b GDP per capita 1960-0.262-0.301-0.063-0.294 (1.77) (5.81) (0.28) (6.38) Years of schooling 1960 0.025 0.025 0.006 0.152 (0.20) (0.26) (0.05) (1.70) Test score (mean) 2.056 1.736 2.286 1.287 (6.10) (4.17) (6.98) (5.37) Constant -5.139-3.539-6.412-2.489 (3.63) (1.96) (4.52) (2.86) N 27 23 25 25 R 2 (adj.) 0.676 0.830 0.707 0.783 Dependent variable: average annual growth rate in GDP per capita, 1960-2000. t-statistics in parentheses. a Non-OECD countries. b Countries below/above sample median of GDP per capital 1960. Source: Hanushek and Wößmann (2007).
Further Robustness Checks Fertility rate Geography (latitude; fraction of land in tropics) Alternative specification checks: Recent tests, early tests, different grade levels, math/science separately, different growth periods
Implications of Reform Speed of reform 10, 20, 30 years Magnitude of reform ½ standard deviation Half distance of Mexico, Indonesia, Chile to OECD U.S., Germany to East Asian Full impact felt 35 years after completion of reform
Improved GDP with Moderately Strong Knowledge Improvement 40% (0.5 s.d.) 30% 20% 10% 0% 2005 2010 2015 2020 2025 2030 2035 2040 2045 2050 2055 2060 2065 2070 2075 2080 year 10-year reform 20-year reform 30-year reform typical education spending percent additions to GDP
Trends in test scores 560 Japan Korea 540 Japan Norway Netherlands Finland N. Zealand Canada Australia UK 520 500 480 N. Zealand Korea UK Netherlands Australia Canada Finland Belgium Germany Sweden USA France Sweden France Belgium Norway USA Germany Italy Italy 460 1975 2000
Trends in growth rates and in test scores
Other Benefits of Improved Cognitive Skills Individual earnings Developed countries Developing countries
Estimated Returns to Cognitive Skills Country Study Estimated effect a Notes Ghana Glewwe (1996) 0.21**-0.3** (government) 0.14-0.17 (priv.) Alternative estimation approaches yield some differences; math effects shown generally more important than reading effects, and all hold even with Raven s test for ability. Ghana Jolliffe (1998) 0.05-0.07* Household income related to average math score with relatively small variation by estimation approach; effect is only observed with offfarm income, and on-farm income is not significantly related to cognitive skills. Ghana Vijverberg (1999)? Income estimates for math and reading with nonfarm selfemployment; highly variable estimates (including both positive and negative effects) but effects not generally statistically significant. Kenya Morocco Pakistan Pakistan South Africa Tanzania Boissiere, Knight, and Sabot (1985); Knight and Sabot (1990) Angrist and Lavy (1997) Alderman, Behrman, Ross, and Sabot (1996) Behrman, Ross, and Sabot (forthcoming) 0.19**-0.22** Total sample estimates: small variation by primary and secondary school leavers.? Cannot convert to standardized scores because use indexes of performance; French writing skills appear most important for earnings, but results depend on estimation approach. 0.12-0.28* Variation by alternative approaches and by controls for ability and health; larger and more significant without ability and health controls. 0.25 Estimates of structural model with combined scores for cognitive skill; significant effects of combined math and reading scores which are instrumented by school inputs Moll (1998) 0.34**-0.48** Depending on estimation method, varying impact of computation; comprehension (not shown) generally insignificant. Boissiere, Knight, and 0.07-0.13* Total sample estimates: smaller for primary than secondary school Sabot (1985); Knight leavers. and Sabot (1990) *significant at 0.05 level; **significant at 0.01 level. a. Proportional increase in wages from a one standard deviation increase in measured test scores.
Other Benefits of Improved Cognitive Skills Individual earnings Developed countries Developing countries Income distribution
Inequality of Educational Quality and of Earnings Earnings inequality 4.5 4.0 CAN USA 3.5 IRE UK 3.0 AUS 2.5 2.0 NET FIN GER DEN SWE NOR BEL SWI 1.5 1.3 1.5 1.7 1.9 Test score inequality
Conclusions Powerful effects of cognitive skills on individual earnings, on the distribution of income, and on economic growth Importance of both minimal and high level skills Complementarity of skills and the quality of economic institutions Robustness of relationship between skills and growth The current situation in developing countries is much worse than generally pictured on the basis just of school enrollment and attainment