Thursday, December 20, 2007

Research on Predicting Performance During Selection

There are some inherent problems with using "predictors" of suitability for open positions within a company. Hunter and Hunter (1984) showed that interviewing is certainly one of the least valid predictors of future job performance (a validity coefficient of .14). Using such things as reference checks (.26), education (.22) or biographical data (.37) did not improve the odds too favorably. In fact, even the best predictor, test scores, came out at .57. So, one could make the argument that in order to be able to best predict future job performance of any person, a combination of methods should be used. Hunter, J.E. and Hunter, R.F. 1984. "Validity and utility of alternative predictors of job performance". .Psychological Bulletin. 96:72-98
Additional research, including Campion, Campion & Hudson, 1994; Huffcutt and Arthur, 1994; Huffcutt, Roth et al, 1996; and McDaniel et al, 1994, continued to show structured interviews having a maximum validity coefficient of .24. One of the most-embraced contemporary personality assessments, the Big 5, was found in Mount and Barrick's 1995 meta-analytic study to have a coefficient of .18 (uncorrected) for Conscientiousness. There is no silver bullet.
Additionally, at least in the US, one must be concerned about the legality of "testing" in employment decisions. Specifically, Griggs v. Duke Power found that any testing must show "a manifest relationship to the employment in question" and that it is up to the employer to show that it did.
To expand on the points about employing a combination of methods, one should conduct a job analysis to establish job relevance of various KSAOs (as required by Griggs, the Uniform Guidelines, CRA, etc.). A well-designed job analysis may also provide guidance with regard to selection methods (test, interview, resume screen, etc.) as well as testing constructs (cognitive ability, conscientiousness, job knowledge, etc.). Therefore, the job analysis will drive the appropriate selection methods.

No comments: