Catalytic Variables for Improving Personnel Classification and Assignment

Abstract

Organizations have a fundamental problem of placing personnel Into jobs to maximize expected performance. Whether or not placing people in specific jobs really makes a difference in overall expected system performance depends on the interaction of people characteristics with jobs. It is desirable to increase the interaction of the people characteristics, as measured by predictor tests, with the jobs.

The purpose of this effort is to suggest a procedure for using one set of performance predictor variables in a simple noninteractive way to enhance the differential classification potential (person-job interaction) of a set of operational predictor variables. The noninteractive variables are required only in determination of the regression coefficients for the operational predictors, but are not required for operational use in future differential classification actions.

Separate equations are developed to predict performance on each job. The equations are determined so that the weights for the operational predictors are allowed (if necessary) to vary across the various jobs. However, one set of predictors (the potential catalytic variables) is required to have the same regression weights across all jobs (noninteractive). If this noninteractive set of predictors can increase the amount of person-Job interaction in the new predicted performance values, then the potential for improved assignment has been increased. These noninteractive variables are called catalytic.

Since catalytic variables are used in prediction systems in a noninteractive way, they are not required for future use in the classification system. Therefore, this procedure will allow personnel classification system developers to use a set of catalytic predictors to enhance the differential classification potential of a set of operational (interactive) prediction, but not require these catalytic predictors for future classification. If catalytic variables can be found, savings in time and money might be possible with little loss in classification effectiveness of the operational predictors.

This approach should be applied to prediction situations in which data are already available and it is desirable to enhance the classification effectiveness of a set of operational predictors without requiring the operational use of the catalytic variables.