The impact of relative position on attributes of an estimated performance distribution

Abstract

Decision theoretic equations have been used to estimate the relative utility of personnel programs for years. These equations; however, have been used sparingly in I/O psychology. A number of reasons have been propsed to explain this dearth of the use. Some psychologists believe validity to be situation specific. Others believe the data required by the equations do not satisfy the assumptions of the general linear model, on which the utility equations are based. the most cited reason for the infrequent use; however, is the difficulty in obtaining one of the required parameters namely the standard deviation of job performance in the utility metric, usually dollars (SDy). Recent studies have developed techniques for estimating this parameter. One such technique is the global estimation method proposed by Schmidt, Hunter, McKensie and Muldrow (1979). The method is based on the assumption that if job performance is normally distributed then estimates of a workers value at the 85th percentile minus that of the value of the work performed by an average employee should yield an estimate of the SDy. The technique depends on whether an individual can provide estimates that adhere to the normal distribution.

The comparison revealed little similarity in the estimates produced by the different methods. These results point to some ineresting aspects of utility analysis and measurement in general. First it appears that the best judge of a position may be the person functioning in that position. Second, the assumption of normality is not supported by these data. (Abstract shortened)