Lagged relationships between a multilevel model of safety climate and employee safety outcomes

Abstract

In Zohar’s Multilevel Model of Safety Climate (Zohar, 2000; 2003; Zohar & Luria, 2005), workers’ perceptions of their supervisor and manager’s commitment to safety are separately assessed and aggregated to the group and organisational level respectively. Treating safety climate as a multilevel construct has a number of conceptual and methodological advantages over the traditional single-level approach; however few researchers have adopted this practice and examined the cross-level or the lagged effects of safety climate.In the current thesis, the cross-level and lagged relationships safety climate has with safety outcomes was investigated through the use of a recently developed multilevel safety climate survey in the Australian oil and gas context. Data was collected over a two year period in a single organisation. The survey consisted of three scales, separately examining manager, supervisor, and co-worker commitment to safety. Safety outcomes were operationalized as self-reported near misses and injuries.The assessment of cross-level and lagged relationships involved a number of analyses, distributed across five objectives. In Objective One the factorial validity of the scales was assessed using confirmatory factor analysis. In Objective Two the cross-sectional criterion validity of the scales was examined using multilevel logistic regression in order to determine whether commitment to safety at each level of the organisation was associated with safety outcomes. In Objective Three, the predictive validity of each scale was assessed using multilevel Poisson regression in order to determine whether safety climate scores in Year One were predictive of safety outcomes in Year Two. In Objective Four, a series of path models were compared using path analysis in order to examine cross-level relationships between manager, supervisor, and co-worker commitment to safety. This analysis allowed the replication of Zohar and Luria’s (2005) findings, and could determine whether Zohar’s model could be extended to include perceptions of co-workers. In Objective Five, comparisons were made between safety climate operationalized at the individual and aggregate level, given the inconsistent labelling of safety climate at both levels of analysis despite the possibility that they may be distinct constructs.While all three scales demonstrated acceptable factorial validity, only the supervisor and manager scales provided evidence of criterion and predictive validity through significant associations with self-reported near misses. Path model comparisons provided support for Zohar’s Multilevel Model of Safety Climate and further suggested that co-workers were of lesser importance in promoting safety compared to supervisors and managers.Comparisons between safety climate operationalized at the individual and aggregate level demonstrated that level of analysis did affect that pattern of relationships between safety climate and self-reported near misses. While aggregated co-worker safety climate was the weakest predictor of self-reported near misses, individual level co-worker safety climate was the strongest predictor even after controlling for higher level variance. Analyses further indicated that individual level co-worker safety climate mediated the relationship between aggregated supervisor safety climate and individual level self-reported near misses.While replication of these findings is necessary, the results overall supported Zohar’s Multilevel Model of Safety Climate and suggested that the model could be extended to include individual level perceptions of co-workers. Results also indicated that level-of-analysis has a potentially important effect on the pattern of relationships between safety climate and safety outcomes.