Criterion and construct validation of an assessment centre

Abstract

This study is an attempt in assessment centre research to apply both criterion and construct validation strategies to a single sample and examine, simultaneously, a relatively comprehensive set of variables including assessor ratings, psychological test measures, supervisory ratings of job performance and actual promotions, hence allowing more direct comparisons of a variety of validities and explorations of previously unexamined issues. Results showed a lack of both internal construct validity as demonstrated by multitrait–multimethod analyses and factor analysis and external construct validity when placed in a nomological network of constructs independent of the centre. Assessment centre ratings were found to be predictive of subsequent promotion (r = .59, p < .001) but not of concurrent supervisory ratings of performance (r = .06, n.s.). Logistic regression analyses showed that assessment centre ratings produced a significant and substantial increment in validity in predicting promotion over and above current supervisory ratings of job performance (Δχ2(1) = 20.06, p < .001), which is an important relationship that has not been previously examined. Implications of these findings for the nature of constructs in assessment centres and future research are discussed in the context of Klimoski & Brickner's (1987) ‘performance consistency’ explanation and ‘subtle’ criterion contamination explanation as to why assessment centres work.