Self selection in the labor market

Often models of labor markets have assumed that firms know the productivity of all appplicants and pay wages proportionate to those productivities. In the presence of heterogeneity among the labor force and imperfect information by employers, this assumption is overly strong. A more common practice is for firms to offer a wage for a given job classification, and to test applicants to try to ensure a minimal level of performance.' The tests used by employers typically include a trial hiring period during which the applicant's performance is carefully monitored as well as perusual of the applicant's education record, previous job experience, and behavior during an interview. Since tests are costly to administer, inaccurate, and imprecise, firms try to discourage applications from workers who do not meet their hiring standards. One way of discouraging the less qualified is to require applicants to pay a fee for being tested.2 Imposing a cost for being tested discourages applications from individuals who believe their probability of passing the examination is low (as well as from poorer individuals if the marginal utility of income is decreasing). The use of an application fee has the effect of converting a one-part test into a two-part test; only workers who both perceive their probability of passing the test to be high and who actually pass the test are hired. For example, if an apprenticeship program is viewed as a test, then below-market wages during the apprenticeship discourage applications from workers who would be less likely to successfully complete their apprenticeships (those workers who successfully finish the apprenticeship program will receive an increase in their wages). The difference between an applicant's wage in the training program and the wage he could obtain elsewhere is the fee for being tested.3 This interpretation of apprenticeship implies that wages increase with job tenure, not because of the acquisition of human capital, but rather as a consequence of the combination of tests and wages to sort workers. The debate here is similar to that over sorting versus human capital theories of education; however, by focusing on sorting versus human capital explanations of low-wage apprenticeships, we can more readily generate testable hypotheses which distinguish between the two models. For example, the human capital approach predicts that productivity per manhour increases with job tenure, while a pure sorting model predicts that, for individual workers, their productivity is independent of their job tenure.4 *University of California-San Diego and Bell Laboratories, respectively. We benefited from illuminating conversations with Joseph Stiglitz. We also wish to thank Peter Fishburn, Bruce Greenwald, Henry Landau, Ed Zajac, and George Borts for helpful comments. The beginning of this research program was supported by a Sloan Foundation grant for Weiss. Work was continued while Guasch was a Research Fellow at CORE, Universite Catholique de Louvain, Belgium. Issues of productivity differences among workers paid a uniform wage has been discussed by Melvin Reder (1955, 1969), William Brown, and Weiss (1976), among others. 2 One potential difficulty with this test-cum-fee strategy is that firms may have an incentive to take the fees and announce that workers have failed, without testing them, i.e., "take the money and run." In our model this problem can be ignored since firms, in general, are making positive profits. If they renege on their explicit contract to test applicants and hire those who pass, they would not attract future applicants and would find the expected value of their stream of profits to have fallen. In cases where reneging on contracts is profitable, we would expect legal instruments to arise to enforce contracts. 3Although, in principle, the screening mechanisms available to firms are numerous, we will focus specifically on the use of testing strategies, with or without application fees. The nontesting strategies, where the firm uses the wage offered as a means of maximizing labor input per dollar, have been analyzed in our 1980b article and by Weiss (1980). 4Of course, measurements of the average productivity of workers may show increases with job tenure if the