| The current literature on test score banding in personnel selection suggests that it can only be used in static hiring situations where employee selection is based on a fixed population of applicants. One reason for this view is that banding relies on test reliability, and so the cutoffs for bands cannot be set until a sample is drawn. In applied settings where hiring is continuous (i.e. the applicant population is ever changing), it may not be practical to delay hiring decisions until an adequate sample is drawn for computing reliability. Based on a population of 14,848 applicant test scores, the present study filled this void by (1) determining the smallest sample size needed to accurately approximate a test's reliability in a large population, (2) describing a bootstrapping method for estimating test reliability in a large population using a small sample and (3) demonstrating that when using the method described, sample characteristics did not impact the calculated reliability of the test. The results of the study are significant for measurement theory by describing empirically the relationship between sample size and reliability, as well as for test developers who now have a practical method for using banding in continuous hiring settings. |
