Greater use of employment tests for selecting workers will have important effects on the economy. First, the rewards for developing the competencies measured by the tests will rise and this will increase the supply of workers with these competencies. Employment tests predict job performance because they measure or are correlated with a large set of developed abilities which are causally related to productivity and not because they are correlated with an inherited ability to learn. Our economy currently under-rewards the achievements that are measured by these tests and the resulting weak incentives for hard study have contributed to the low levels of achievement in math and science.
Greater use of tests to select workers will also change the sorting of workers across jobs. Its impacts on total output depends on the extent to which the developed abilities measured by employment tests--academic achievement, perceptual speed and psychomotor skills--have larger impacts on worker productivity in dollars in some occupations than in others. This question is examined by analyzing GATB revalidation data on 31,399 workers in 159 occupations and by reviewing the literature on how the standard deviation of worker productivity varies across occupations. The analysis finds that indeed such differentials exist and therefore that reassigning workers who do well on a test to occupations where the payoff to the talent is particularly high will increase aggregate output. The magnitude of the output effect was estimated by reweighting the GATB revalidation data to be representative of the 71 million workers in the non-professional and nonmanagerial occupations and then simulating various resorting scenarios. Selecting new hires randomly lowered aggregate output by at least $129 billion or 8 percent of the compensation received by these workers. An upper bound estimate of the productivity benefits of reassigning workers on the basis of three GATB composites is that it would raise output by $111 billion or 6.9 percent of compensation. Reassignment based on tests had an adverse impact on Blacks and Hispanics but greatly reduced gender segregation in the work place and substantially improved the average wage of the jobs held by women. These results are based on a maintained assumption--the models of job performance which were estimated in samples of job incumbents are after corrections for measurement error and selection on the dependent variable yield unbiased estimates of true population relationships--that is almost certainly wrong. The biases introduced into the calculation by this assumption lower the estimated costs of introducing random assignment of workers to jobs, exaggerate the benefits of greater test use and exaggerate the changes in demographic composition of occupational work forces.
The paper concludes with a discussion of ways in which employment tests can simultaneously strengthen incentives to learn, improve sorting and minimize adverse impacts on minority groups.