Job Testing

A somewhat surprising conclusion to this article:

Will Job Testing Harm Minority Workers?

DAVID H. AUTOR
Massachusetts Institute of Technology (MIT) - Department of Economics; National Bureau of Economic Research (NBER)
DAVID SCARBOROUGH
Unicru, Inc.

August 1, 2004

MIT Department of Economics Working Paper No. 04-29

Abstract:
Because of the near-universal finding that minorities fare poorly on standardized tests, the use of such tests for employment screening is thought to pose an equity-efficiency trade-off: improved selection comes at a cost of screening out more minority applicants. This paper investigates the consequences of standardized testing for minority employment and productivity. We analyze the experience of a large, geographically dispersed retail firm whose 1,363 stores switched from paper to electronic job applications during 1999 and 2000. Both hiring methods use face to face interviews, but the test-based screen also places substantial weight on a computer administered personality assessment. We find strong evidence that testing yielded more productive hires - increasing median employee tenure by 10 percent, and slightly lowering the frequency at which workers were fired for cause. Consistent with prior research, minorities applicants performed significantly worse on the employment test. Had stores initially screened workers in a manner uncorrelated with the test, simple calculations suggest that testing would have lowered minority hiring by approximately 10 to 25 percent. This did not occur: employment testing had no measurable impact on the racial composition of hiring at the firms 1,363 sites; and, moreover, productivity gains were uniformly large among both minority and non-minority hires. As we show formally, these results imply that employers were, in effect, statistically discriminating prior to the introduction of employment testing - that is, their hiring practices already accounted for expected productivity differences between minority and non-minority applicants. Consequently, testing improved the accuracy of selection within each applicant group (minorities, non-minorities) without generating measurable cross-group shifts in hiring.