Wilfried De Corte

Weighing job performance predictors to both maximize the quality of the selected workforce and control the level of adverse impact

Considerable thought has been given to the effects of various strategies of weighing predictor information on adverse impact, minority hiring, and quality of the selected workforce. However, these efforts do not solve the dilemma faced by employers who want to achieve an optimally qualified workforce but at the same time want to eliminate adverse impact. To remove this limitation, the present article shows how constrained nonlinear programming can be used to combine job performance predictors into a predictor composite such that the average quality of the composite selected employees is maximized, the intended overall selection rate is achieved, and the adverse impact ratio remains within acceptable bounds. Although the new procedure allows for situations in which the performance criterion is multidimensional, a further extension is needed to handle multistage selection decisions.

A note on the success ratio and the utility of fixed hiring rate personnel selection decisions

Using simple simulation methods, the author verifies that the formulas used to estimate both the expected success ratio of a selection and the expected average criterion score of the selectees overestimate the real gain one may expect to obtain when hiring a preestablished proportion of employees from a finite sample of applicants. To correct for this deficiency, the present article shows how an exact estimate of the expected payoff can be obtained. In addition, it indicates how the new approach can be used to assess the sampling variability of the utility estimates.

Using order statistics to assess the sampling variability of personnel selection utility estimates

Virtually all models for the utility of personnel selection are based on the average criterion score of the predictor selected applicants. This paper indicates how standard results from the theory on order statistics can be used to determine the expected value, the standard error and the sampling distribution of the average criterion score statistic when a finite number of employees is selected. Exact as well as approximate results are derived and it is shown how these results can be used to construct intervals that will contain, with a given probability 1 - f , the average criterion score associated with a particular implementation of the personnel selection. These interval estimates are particularly helpful to the selection practitioner because they can be used to state the confidence level with which the selection payoff will be above a specific value. In addition, for most realistic selection scenarios, it is found that the corresponding utility interval estimate is quite large. For situations in which multiple selections are performed over time, the utility intervals are, however, smaller.Virtually all models for the utility of personnel selection are based on the average criterion score of the predictor selected applicants. This paper indicates how standard results from the theory on order statistics can be used to determine the expected value, the standard error and the sampling distribution of the average criterion score statistic when a finite number of employees is selected. Exact as well as approximate results are derived and it is shown how these results can be used to construct intervals that will contain, with a given probability 1 - f , the average criterion score associated with a particular implementation of the personnel selection. These interval estimates are particularly helpful to the selection practitioner because they can be used to state the confidence level with which the selection payoff will be above a specific value. In addition, for most realistic selection scenarios, it is found that the corresponding utility interval estimate is quite large. For situations in which multiple selections are performed over time, the utility intervals are, however, smaller.

Estimating the classification efficiency of a test battery

Classification decisions relate to situations in which a battery of predictor tests is used to allocate subjects to a number of different assignment positions according to some prespecified quota proportions. In the case that the performance on the assignment criteria is continuously measurable, the efficiency of the predictor-based classification can be equated to the expected criterion score of the optimally assigned individuals: the allocation average. It is shown how a theorem proven by Brogden can be used to estimate the allocation average assuming that the predictor intercorrelations and validities are known and that the predictor variables have a joint multivariate normal distribution. Compared to previous analytical procedures to estimate the efficiency of a classification, the proposed solution is much more generally applicable in that it is no longer assumed that the predictor-based criterion estimates are both equally valid and equi-correlated and that all quota proportions are equal.Classification decisions relate to situations in which a battery of predictor tests is used to allocate subjects to a number of different assignment positions according to some prespecified quota proportions. In the case that the performance on the assignment criteria is continuously measurable, the efficiency of the predictor-based classification can be equated to the expected criterion score of the optimally assigned individuals: the allocation average. It is shown how a theorem proven by Brogden can be used to estimate the allocation average assuming that the predictor intercorrelations and validities are known and that the predictor variables have a joint multivariate normal distribution. Compared to previous analytical procedures to estimate the efficiency of a classification, the proposed solution is much more generally applicable in that it is no longer assumed that the predictor-based criterion estimates are both equally valid and equi-correlated and that all quota proportions are equal.

An analytic procedure to estimate the benefits of a particular type of personnel classification decision

The article presents an analytic procedure to estimate the expected benefits of personnel classification decisions for which it is assumed that the available criterion estimates are both equi-correlated and equally valid, that equal quotas have to be filled for the jobs, and that all jobs are equally important. The method outperforms some already existing procedures in that an exact benefit estimate is obtained even when the criterion estimates correlate substantially. To perform the estimation, an efficient numerical procedure is developed and subsequently applied to study the effect of the number of assignment categories, the applicant rejection rate, the validity of the criterion estimates, and the intercorrelation among the estimates. Among other things, the results indicate that the expected benefits of a classification decision are considerably higher than those obtained for the corresponding selection decision, even when the criterion estimates correlate substantially.The article presents an analytic procedure to estimate the expected benefits of personnel classification decisions for which it is assumed that the available criterion estimates are both equi-correlated and equally valid, that equal quotas have to be filled for the jobs, and that all jobs are equally important. The method outperforms some already existing procedures in that an exact benefit estimate is obtained even when the criterion estimates correlate substantially. To perform the estimation, an efficient numerical procedure is developed and subsequently applied to study the effect of the number of assignment categories, the applicant rejection rate, the validity of the criterion estimates, and the intercorrelation among the estimates. Among other things, the results indicate that the expected benefits of a classification decision are considerably higher than those obtained for the corresponding selection decision, even when the criterion estimates correlate substantially.