Patterns of assessment center construct and criterion-related validity as a function of the conceptual width of dimensions
Sabatini, Kenneth Alan
MetadataShow full item record
This study proposes that all assessment center (AC) dimensions can be categorized along a continuum of conceptual width, makes a theoretical and research-based connection between conceptual width and patterns of construct and criterion validity, and examines these effects using data from an operational AC particularly suited for such an assessment. It is argued that ACs measure narrow inference dimensions largely as a function of their observation and rating approaches and that these narrow dimensions are virtually assured to yield situationally specific constructs. Broader dimensions, carefully identified using sets of broad level within-dimension rating items, are believed to provide a truer test of ACs' validity potential. Construct validity is assessed using traditional multitrait-multimethod analysis and confirmatory factor analysis. This study's results indicate that measuring AC dimensions more broadly does nothing to improve validity and instead is related to losses in discriminant validity, increases in exercise effects, and weakened criterion-related validity. Although the conceptual width concept did not provide a solution to ACs' validity limitations, it demonstrates that AC dimensions can be meaningfully categorized along an underlying continuum of width and that this categorization is meaningfully related to patterns of construct and criterion-related validity. Patterns found in past AC construct validity research are interpreted using this study's levels of inference concept.