The thesis of this paper is that wage rates and earnings give misleading signals to public and private decision makers regarding the social benefits of certain kinds of education and training (E&T) investments. The misleading signals are a result of the fact that (1) workers and employers prefer employment contracts which either do not recognize or only partially recognize differences in productivity among workers doing the same job and (2) important dimensions of E&T accomplishment -- the skill, knowledge and competencies actually developed -- are often not signaled to potential employers and therefore have limited influence on the allocation of workers to jobs. The result is that there are significant productivity differentials between workers who receive the same pay for the same job and some of these productivity differentials are related to dimensions of E&T accomplishment that are not efficiently signaled.
The paper develops a very simple signaling/implicit contracting model of the labor market. True productivity depends on general intellectual achievement (GIA) and educational credentials but GIA is unobservable, so pay is based on credentials and supervisory assessments of doubtful reliability. As in most signaling models, the labor market tends to overcompensate credentials and undercompensate academic achievement. The next section of the paper refutes the simple wage equals individual MRP assumption by presenting evidence of great variability of productivity across workers paid the same wage and doing the same job. The paper then tests and rejects a weaker hypothesis that can justify an inference that productivity and wage effects of GIA are equal -- namely that deviations of productivity from wages are not correlated with academic achievement. Finally the paper develops a method of estimating the true impact of academic achievement on productivity and applies it to data on the productivity of 31,399 workers.
The analysis provides strong support for signaling theory. As predicted by the theory when workers doing the same job are compared and academic achievement (the unobservable) is controlled, the years of schooling signal is negatively associated with relative productivity. When the schooling signal is controlled, academic achievement has a very strong positive effect on relative productivity. This implies that academic achievement has a larger effect on productivity than it has on wages. Academic achievement produces some private rewards for it facilitates entry into higher paying occupations and promotions into better jobs. These are the effects that are captured by standard wage regressions. In addition GIA has effects not picked up by wage regressions. In each job the individual works he/she is doing a better than average job but not receiving an appreciably higher wage as a result. The results imply that schooling raises productivity primarily by improving academic achievement as it is measured by standard tests. When it does not lead to gains on such tests, the credentials that graduates receive tend to be overcompensated. The second major implication of the results is that academic achievement is substantially under compensated if it is not signaled to the market by a credential. This tendency to underreward academic achievement may help explain why American high school students devote less time and energy to learning than their counterparts abroad.