ABSTRACT
The purpose of this study is to examine the boundary conditions of transformational leadership, follower psychological capital, and their effects on follower mental health outcomes. Specifically, we utilize archival, multi-wave data from a military sample to examine whether the negative relationship between transformational leadership and adverse follower stress outcomes increases as the context shifts from a relatively safe environment to one in which follower lives are at risk. Additionally, psychological capital, a constellation of personal psychological resources, is also assessed to account for individual buffers against extreme stressors. Findings from the current study suggest that the negative relationship between transformational leadership and follower stress increases significantly when the context shifts to a high-risk, mortality-salient environment.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Data availability statement
The data that support the findings of this study are available from the corresponding author, Paul B. Lester, upon reasonable request.
Notes
1. During the review process, it was suggested that we test this difference using the Fisher z-transformation (Cohen et al., Citation2003). This test assumes independent samples for the two correlations, which in this particular analysis is violated because 185 participants appeared in both samples. Therefore, we urge readers to interpret the results of this statistical test with caution. The z-transformed correlations are −.17 (T1) and −.24 (T2), with a difference of .07, which is not statistically significant (z = .95). Exploring differences in the subsample that appeared in both groups (N = 185), the difference is .11 (ns, z = 1.01). Exploring differences in the subsamples that only appeared in one group (NT1 = 186, NT2 = 132), the difference is <.01 (ns, z = .03).
2. Understanding that the 185 participants who appear in both samples may have exerted some influence on these results, we next compared means and inter-scale correlation matrices separately for participants appearing in only one sample and participants appearing in both. At T1 (N = 186 unique participants, 185 who also appeared in the T2 data), there was a statistical difference in means on transformational leadership (unique group 0.20 scale points higher, t = 2.28, p < .05), but no statistical difference between groups on psychological capital (t = 0.29, p > .05) or PTSD symptomology (t = 0.63, p > .05). At T2 (N = 132 unique participants, 185 who also appeared in the T2 data), there was no statistical difference between groups on transformational leadership (t = 1.92, p > .05), psychological capital (t = 1.49, p > .05) or PTSD symptomology (t = −1.66, p > .05). To compare inter-scale correlation coefficients, we used the SRMR statistic (Bentler, Citation1995). While this technique is typically used to compare correlation matrices in structural equation modeling, it is also useful for comparing identical correlation matrices in other contexts as well (DeSimone, Citation2015), as it is simply an estimate of the average absolute deviation in correlation. It is noteworthy that the direction of all correlations was identical for the unique and common samples at both T1 and T2. The SRMR statistic was .04 at T1 and .07 at T2. Both values are below the empirical cutoff of .08 recommended by Hu and Bentler (Citation1999), indicating similarity between correlation matrices.
3. It is impossible to perfectly adjust Type I error rates in response to artificially doubling sample size because the effects on power also depend on the magnitude of the effect size (Cohen, Citation1988). Therefore, we opted for a conservative approach where we divided the acceptable Type I error rate by five in exchange for multiplying the sample size by two.