An evaluation of the construct validity of situational judgment tests

Abstract

Situational judgment tests are analogous to earlier forms of "high fidelity" simulations such that an ostensible paradox emerges in the consistent finding of criterion-referenced validity but almost complete lack of construct validity evidence. The present study evaluates the extent to which SJT's can demonstrate convergent and discriminant validity by analyzing a SJT from a multitrait-multimethod perspective. A series of hierarchically nested confirmatory factor models were tested. Results indicate that the SJT demonstrates convergent and discriminant validity but also contains non-trivial amounts of construct-irrelevant method variance. Wide variability in the content and validation methods of SJT's are discussed as the reason previous attempts to find construct validity have failed.