Abstract
Extant research on climate strength has frequently invoked situational strength in passing. Given this, it is surprising that the fundamental prediction from the situational strength literature—namely, that strong situations attenuate personality–behaviour relationships—has thus far never been tested empirically using climate strength as an operationalization of situational strength. Consequently, in the present study, we tested this fundamental prediction by positing that organizational safety climate strength moderates the relationships between employee conscientiousness and two forms of employee safety behaviour, such that those relationships are attenuated in strong climates. Using a multilevel design consisting of 964 Korean employees nested within 17 manufacturing organizations, we found support for these cross-level interaction hypotheses. These findings legitimize the frequent invocation of situational strength by climate strength researchers, contribute to the theoretical foundation underlying climate strength, and yield important implications for future research and practice.
This work was supported by the Korea Occupational Safety and Health Agency.
Notes
1 We re-ran the analyses with the organization size and the number of respondents per organization as control variables. Because the substantive conclusions for focal predictor variables remained the same, we report the analyses without these control variables.
2 We employed Mathieu et al.’s (Citation2012) method because it is the only one available for cross-level interaction effects in multilevel analyses at this time. However, the method has some limitations when applied to our models. First, whereas our models include two Level 1 predictors (conscientiousness and age) and two Level 2 predictors (organizational climate level and strength), the method is limited to models with one Level 1 predictor and one Level 2 predictor/moderator. Second, the method uses within-group centring and reintroduces the between-group variance in the Level 1 predictor as a Level 2 predictor. Thus, the interaction term in the model used in this method is an interaction between the Level 2 moderator and the within group variance of the Level 1 predictor; consequently, it is not perfectly comparable to the interaction term in our model.