Criterion dynamism and growth mixture modeling
MetadataShow full item record
There is growing consensus among organizational researchers that job performance changes over time, especially in jobs requiring skill acquisition. The nearly ubiquitous finding of declining correlations between selection assessments and performance as time increases calls into question the utility of selection tools. Studying performance cross-sectionally or at the mean level cannot address why predictive validity decreases over time. Latent growth models assume that growth trajectories come from a single population with normal variance around the parameters, but this is likely a naïve assumption. Growth mixture modeling (GMM) does not make this assumption and was used to identify classes of performance change that would be considered desirable by organizations' stakeholders. Objective job performance data for three metrics were collected over nine months for a sample of 203 call center agents. Class membership probabilities were calculated using Mplus (Muthén & Muthén, 2007) for a model in which individual performance trajectories were estimated with assessments of cognitive ability, emotional resilience, sales ability, and conscientiousness as covariates. Across the three metrics, GMM identified three to four classes of growth. Though the assessments did not predict membership in "desirable" trajectories, hypothesized predictor-criterion relationships were supported in some classes and not others. This suggests that GMM is appropriate for identifying subpopulations in which the predictor-criterion relationships do not behave as expected.