Key points are not available for this paper at this time.
Studies using cross-lagged panel models (CLPM) have shown prospective effects of initial scores on various predictors, e.g., job control and depressive symptoms, on subsequent scores on work engagement when adjusting for initial work engagement. However, it is well established that cross-lagged effects may be spurious due to correlations with residuals and regression to the mean. Here, we reanalyzed data simulated to resemble data in twelve studies using CLPM, suggesting increasing/decreasing prospective effects on work engagement. Through triangulation, we show that all of these twelve suggested prospective effects appear to have been spurious rather than truly increasing/decreasing. Researchers need to be aware that correlations, including effects in cross-lagged panel models, do not prove causality. For improved causal inference, we recommend researchers to triangulate and scrutinize cross-lagged effects by fitting complementary models to their data.
Sorjonen et al. (Fri,) studied this question.