|
Article Excerpt
Few job stress studies have directly compared competing causal models using rigorous methodologies as recommended in the literature (Frese & Zapf, 1988; Williams & Podsakoff, 1989; Zapf, Dormann, & Frese, 1996). This replication study investigated whether De Jonge et al.'s (2001) 2-wave panel study could be cross-validated in a second sample of health care workers.
De Jonge et al. (2001) structural equation modelling (SEM) analyses compared three cross-lagged causal models of the relationship between job characteristics (e.g. job demands, job autonomy, workplace social support) and employee well-being (e.g. job satisfaction, work motivation, emotional exhaustion): a regular causation model (job characteristics at Time 1 [T1] influence well-being at Time 2 [T2]); a reverse causation model (T1 well-being influences T2 job characteristics); and a reciprocal causation model (combining regular and reverse causation). Their model comparisons supported regular causation as the best account (despite weaker evidence for reverse causation), consistent with a principal assumption of Karasek's demand-control-support (DCS) model: favourable (or unfavourable) job characteristics lead over time to increased (or reduced) employee well-being (e.g. Karasek & Theorell, 1990; see De Lange, Taris, Kompier, Houtman, & Bongers, 2003; Van der Doef & Maes, 1999a). Their results showed significant longitudinal paths from T1 job demands ([beta] : -0.18, p < .05) and T1 workplace social support ([beta] = 0.34,p < .05) to T2 job satisfaction.
As De Jonge et al. (2001) have argued, there are theoretical arguments and some empirical evidence supporting all three causal models. Nevertheless, in line with their findings and the DCS model, our hypothesis for this study was that regular causation would again provide the best account. Prior research has linked all three DCS job characteristics--demands, autonomy and social support--to job satisfaction and well-being later in time (see De Lange et al., 2003), suggesting that such effects might be expected in the present study. The DCS model also predicts that job autonomy and social support should buffer negative effects of high job demands; however, such interactive effects have been observed only rarely (De Lange et al., 2003; Van der Doef & Maes, 1999a) and were not evaluated in either the original or the present study.
Method
Design and procedure
A 2-wave full panel design was used, with a time lag of 2 years (vs. 1 year in the original study). In April 1999 (T1) and April 2001 (T2), questionnaires were distributed to all employees at two Dutch residential health care institutions, followed by reminder letters several weeks later.
Participants
Of 715 questionnaires distributed at the first measurement, 328 (46%) were returned. Of these respondents at T1,100 (30%) had left their institution by T2. Of the remaining 228 respondents, 137 (60%) returned a second questionnaire at T2. Hence, the panel group comprised 19% of the original group at T1 (vs. 57% in the original study). Most respondents in the panel group were women (82%), as in the original study. The majority (60%) were aged between 36 and 65 years (vs. 40% aged between 18 and 35 years); 43% had received university-level education.
The panel group (N = 137; vs. N = 261 in the original study) was compared with the group lost to turnover at T2 (N = 100) and to non-responders at T2 (N = 93) in one-way ANOVAs with post hoc comparisons, on all control variables, job characteristics and well-being indicators at T1. Significantly higher demands (p < .05), higher emotional...
|
