ESTIMATES OF THE DOLLAR VALUE OF EMPLOYEE OUTPUT IN UTILITY ANALYSES - AN EMPIRICAL-TEST OF 2 THEORIES

This study examined distributions of estimates of the dollar value of performance in studies employing Schmidt, Hunter, McKenzie, and Muldrow's (1979) method for estimating the standard deviation of job performance (SD(y)) and found evidence that (a) the mean 50th percentile estimate is biased downward, (b) estimates of SD(y) appear to be a constant percentage of the 50th percentile estimate, and (c) estimates of SD, as a percentage of the 50th percentile value (SD(p)) are quite similar to empirical SD(p) values based on actual employee output. These findings suggest that the downward bias in the mean estimate of the 50th percentile causes the mean estimate of SD(y) to be similarly biased downward, but does not bias the estimates of SD(p). Finally, an objective method for estimating the value of average employee output is described. We conclude that the product of this value and the mean supervisory estimate of SD(p) yields an unbiased estimate of SD(y).