Abstract
Negative spillover from work into private life has often been found to be a mediator between work characteristics and health complaints. This study examined whether this mediator role can also be found when work characteristics are measured independently of job incumbents, i.e., when self-reported bias can be ruled out as a likely alternative explanation of the findings. Multiple mediator models in a heterogeneous sample of 517 employees from three occupational sectors were tested. Job demand and job control, both self-reported by job incumbents, and assessed by job analysis experts via a standardized observational interview, were measured as job characteristics. As hypothesized, the relations between expert-rated job characteristics and vital exhaustion are mediated by both, self-reported job characteristics and negative spillover. However, whereas expert-rated job demands have an impact on negative spillover and vital exhaustion without being perceived, the perception of job control plays a crucial role for the relations between expert-rated job control, negative spillover, and vital exhaustion. Altogether, we demonstrate that the strong relations between job characteristics, negative spillover, and vital exhaustion are not merely a consequence of biased subjective perception, but are rooted in the “real” job environment.
Notes
1The question arises, whether negative spillover and vital exhaustion are really distinct constructs because of the relative high correlation between both variables (r = .650, p < .001). Therefore, we additionally calculated an exploratory factor analysis including the items that measure negative spillover, and the items that measure vital exhaustion. As a result, the items that measure negative spillover and the items that measure vital exhaustion loaded on different factors. We ran further confirmatory factor analyses to test whether a model with all items loading on one factor, or a model with the items that measure vital exhaustion loading on one factor and the items that measure negative spillover loading on a second factor would be superior. The model with two factors had a much better fit (chi-square = 1091.34, df = 371, RMSEA = .065, CFI = .88, SRMR = .058) than the model with only one factor (chi-square = 1370.49, df = 370, RMSEA = .077, CFI = .83, SRMR = .065). Thus, we conclude that the items do indeed measure distinct constructs.