Assessment centres : a comparison of trait ratings and task ratings for predicting managerial performance

Abstract

In teenstelling met die takseersentrumteorie het navorsing getoon dat takseersentrum trektellings nie oor verskillende situasies stabiel is nie. Hierdie bevindinge mag 'n aanduiding wees van 'n lae konstrukgeldigheid van takseersentrums. Navorsers het tot die gevolgtrekking gekom dat navorsingsresultate nie die aanname ondersteun dat takseersentrums dimensietellings genereer wat trekke weerspieel nie. Dit het navorsers laat bespiegel of take (d.w.s. oefeninge), eerder as trekke (d.w.s. dimensies) as 'n basis gebruik behoort te word om takseersentrumbeoordelings te maak. Die doel van die studie was om te bepaal of takseersentrum trektellings beter voorspellers van werksprestasie van bestuurders is as taaktellings. Die studie sluit 97 blanke middelvlakbestuurders van 'n Suid Afrikaanse diensorganisasie in. Geen verskille is tussen die voorspellingsgeldighede van trekgebaseerde en taakgebaseerde tellings verkry nie (r = 0,087 vir beide tellings). Oorsake vir die gebrek aan verskille tussen die voorspellingsgeldighede is bespreek.

In contrast to assessment centre theory, research has shown that assessment centre trait ratings are not stable across different situations. This may be an indication of the relatively low construct validity of assessment centres. Researchers concluded that research results did not support the assumption that assessment centres generate dimensional ratings that represent traits and speculated whether tasks (i.e. exercises) should be used as a basis for making assessment centre ratings, rather than traits (i.e. dimensions). The aim of the study was to determine whether assessment centre task ratings are better predictors of managerial job performance than trait ratings. The study included 97 white middle-level managers of a South African service organisation. No difference was obtained between the predictive validities of traits ratings and task ratings (r = 0,087 for both ratings). Causes for the lack of differences between the predictive validities were discussed.