(forthcoming, 1996), "Is There a Link between School Inputs and Earnings? Fresh Scrutiny of an Old Literature", in Gary Burtless (Ed.), Does Money Matter? The Effect of School Resources on Student Achievement and Adult Success, Washington, D.C.: Brookings Institution.

The paper reviews the literature on the impact of school resources on earnings and educational attainment. Three strong patterns characterize the existing research. First, papers which find no significant effect of school inputs on outcomes tend to examine resources at the school actually attended, while papers which do find an effect typically use average school inputs by state. The other two patterns have to do with time/age effects: papers which find no link also typically examine workers who were educated between 1960 and the early 1980's, and who were 32 or younger at the time of the wage observation. Papers which do find a link most frequently study older workers who were educated in the first half of the century.

Similar patterns emerge in the literature on educational attainment and school resources, although the literature is surprisingly small.

Five sets of hypotheses to explain these patterns are examined: 1) structural change, 2) age-dependence of the relation between school inputs and workers' earnings, 3) the possibility that school resources matter most in earlier grades, 4) measurement error and lack of power in the school-level studies, and 5) aggregation bias and other biases in the state-level studies. At present, one of the most viable hypotheses appears to be the structural change hypothesis, whereby the link between school spending and student outcomes has weakened over time. Such changes could have arisen from diminishing returns, increasing bureaucratization and centralization of public schools, and rising teacher unionization. Econometric problems might also explain some of the differences between the school-level and state-level studies. The school-level studies may not have high power in tests of the impact of teacher salary and length of the school year, but they clearly have high power in tests of the impact of other inputs such as class size and teacher education. There is some evidence that the state-level studies suffer from a wide range of robustness problems, related primarily to aggregation bias, several types of omitted variable bias and non-random migration. The results are also not robust to allowing for non-linear returns to education.

In addition to evaluating whether school inputs are statistically significant, the review also asks whether the observed elasticities of earnings with respect to school spending are meaningful from a policy perspective. Estimates of the rate of return to spending per pupil and class size reveal a similar pattern to that noted above: studies which examine resources of the actual school or school district obtain much smaller rates of return than do the state-level studies. But using even the most optimistic elasticities of earnings with respect to spending per pupil, the net social rate of return to such spending is negative at plausible rates of discount. In stark contrast, the social rate of return to an extra year of education is much higher.

The review concludes that perhaps the best types of educational reform are to undertake policies which increase the length of time students remain in school, rather than to increase spending per pupil further. The paper also calls for an extension of the traditional 'education production function', to allow for the role of intangible inputs in the education process. Examples of such non-financial inputs include academic standards and the incentive systems under which teachers and pupils operate.