The Quality of Schooling: Comment
指出Behrman和Birdsall在计算学校教育数量社会回报率时的方法错误，并用巴西东北部城市数据验证修正后结果，发现数量与质量回报率差异不大，对教育政策制定者有参考价值。
A recent article in this Review by Jere Behrman and Nancy Birdsall (1983) raises an important issue for makers of educational policy in developing countries. BehrmanBirdsall break ground both in specifying a model that treats schooling quantity and quality simultaneously and systematically, and in developing practical methods for evaluating social rates of return to schooling quantity and quality in the absence of externalities. Their model is estimated using data from Brazil, a country with an impressive recent record in expanding primary schooling. This comment has two objectives. The first is to point out methodological errors made by Behrman and Birdsall in the derivation of the social rate of return to schooling quantity. The second is to test the robustness of this (corrected) finding with a data set that differs from that used by Behrman and Birdsall. The principal finding of the BehrmanBirdsall analysis is that the estimated social rate of return to schooling quantity in Brazil is much lower than that to schooling quality (6.8 vs. 10.4 percent). this is generally the case in developing countries, the possibility of a conflict between productivity and equity goals becomes an issue to be taken into account by educational policymakers. For example, with the conclusions reached by Behrman-Birdsall, If only one-third of the students were to attend school and the resources saved by schooling only one-third of the students could be used to increase both quality and quantity by 50 percent, the estimates imply an [overall] income gain of 32 percent! (p. 941). The analysis that follows demonstrates that correct procedure yields an estimated social rate of return to schooling quantity which is only slightly lower than that to schooling quality (9.8 vs. 10.4 percent), indicating no substantial inefficiency in the allocation of educational resources between schooling quantity and quality. To test the robustness of this result, a model of the BehrmanBirdsall type is estimated using a 1979 set of micro data from one city in the Brazilian northeast. Using theoretically and empirically preferred variable definitions, very similar estimation results are obtained with respect to relative social rates of return to quantity and quality. Section I contains a correction of the errors made in deriving the social rate of return to schooling quantity. Section II is empirical and contains a summary of potential sources of bias, a description of the data set used to estimate a Behrman-Birdsall-type model, and estimation results. Section III contains a discussion of some remaining problems.