Earnings, Schooling, Ability, and Cognitive Skills
利用肯尼亚和坦桑尼亚的城市工资劳动力调查数据，区分认知成就、先天能力和教育年限对收入的影响，以检验人力资本、筛选和证书主义假说。
Conventional estimates now available for a large number of countries generally indicate that the social returns to education are positive, large, and competitive with returns to investment in physical capital.' That such estimates are good guides for public resource allocation has, however, been questioned. The heart of the problem lies in the interpretation of the positive relationship between the education and the earnings of workers: whether, as the conventional estimates assume, the coefficient of the education variable in the earnings function measures the effect on the productivity of workers of human capital acquired in school. It has been hypothesized *that education in part, or instead, represents screening for native ability and motivation, or credentialism, and that as a consequence conventional measures of the social benefit of education are substantially upward biased.2 In this paper we attempt to distinguish the influence on earnings of cognitive achievement, native ability, and years of education as a means of adjudicating the human capital, screening, and credentialist hypotheses. Our econometric analysis is based on two rigorously comparable micro data sets from Kenya and Tanzania, generated by surveys of the urban wage-labor force specifically for this study. These data sets contain the usual variables found in earnings function estimates of the benefits of schooling-individual earnings, years of education, and years of employment experience. In addition, they contain two variables-measures of the worker's cognitive skills and of his or her reasoning ability-not previously found in studies of developing countries and only rarely found in studies of the education-earnings relationship in developed countries.3 With these variables we can estimate the direct effects on earnings of cognitive skills, ability, and years of schooling. By using them to estimate educational production functions and educational attainment functions, and linking these functions with the earnings function in a recursive framework, we can also assess the various indirect effects on earnings of ability and years of schooling. Having data sets from two countries very similar with respect to size, resource endowments, structure of production and employment, and level of development means that not only can we subject our results to the usual statistical tests, but we can also assess their replicability. Both Kenya and Tanzania have nearly achieved the objective of universal primary education while university enrollments remain at less than 1 percent of the relevant age group. The important policy issues re* Boissiere: Development Research Department, The World Bank, 1818 H Street, Washington, D.C. 20433; Knight: Institute of Economics and Statistics, Oxford University; Sabot: Williams College, Williamstown, MA 01267. We are grateful to the Educational Testing Service of Princeton for the design of tests used in this study and to J. Armitage, J. Behrman, J. Hausman, D. Hendry, D. Jamison, and an anonymous referee for their insights and advice. Helpful comments were also received from participants in seminars at Oxford and Yale universities. The views presented here are our own; they should not be interpreted as reflecting those of the World Bank. 'George Psacharopoulos (1973; 1981) contains a listing of 44 countries in which rate of return studies had been conducted and of the estimates obtained. 2For instance, Kenneth Arrow (1973), Mark Blaug (1976), Samuel Bowles and Herbert Gintis (1976), John Riley (1979), Michael Spence (1976), and Lester Thurow (1975). 3For attempts to control for ability and/or for cognitive achievement in studies for the United States, see Jere Behrman et al. (1980), Gary Chamberlain and Zvi Griliches (1977), Griliches and William Mason (1972), Michael Olneck (1977), Paul Taubman and Terence Wales (1974), Taubman (1975), and David Wise (1975); see also the survey articles by Griliches (1977; 1979). In most instances the data refer to special subgroups in the population and clear distinction cannot be made between natural ability and cognitive skills acquired in school-