Abstract
In occupational psychology, the Job Demands–Resources (JDR) model is considered as a compelling model to explain burnout and work engagement. Despite its robustness, it can be addressed two main criticisms, namely a lack of three-wave longitudinal studies and an exclusive focus on work-related predictors of well-being. The aims of our study are (1) to test the JDR model using a three-wave longitudinal design, and (2) to test the JDR model using predictors reflecting intergroup relationships within the work context. Structural equation modelling analyses were performed on data collected in a Belgian public institution (N = 473). Results indicate that burnout and work engagement are respectively predicted by perceived stigma against one's occupational group and by group identification. Moreover, group identification moderates the relation between perceived stigma and work engagement. Results are discussed in terms of the role of group identification as a coping strategy, as well as with regards to potential effects of what has been called “dirty work”.
Notes
1We also tested an alternative model. Reverse causation was defined. Exhaustion at Time 1 predicted Time 1–Time 2 changes in time pressure and in perceived stigma, whereas work engagement at Time 1 predicted Time 1–Time 2 changes in availability of finances and equipment and changes in group identification. Crossovers between the two processes were also added. Time1–Time 2 changes in time pressure and in perceived stigma predicted engagement at Time 2, whereas Time 1–Time 2 changes in availability of finances and equipment and changes in group identification predicted burnout at Time 2. Finally, exhaustion at Time 2 predicted intention to leave at Time 3, whereas engagement at Time 2 predicted health complaints at Time 3. Fit indices for this alternative model were satisfactory: χ2(12) = 250.42; SRMR = .08; NFI = .91; CFI = .92; GFI = .92. This was, however, lower than Model 3, Δχ2 = 56.27, Δdf = 124, p < .01, and lower than Model 4, Δχ2 = 56.42, Δdf = 10, p < .01. Adding reverse causation and cross-relations does not improve model fit in our sample.