Control Variables in Leadership Research: A Qualitative and Quantitative Review

Abstract

Statistical control of extraneous (i.e., third) variables is a common analytic tool among leadership researchers. While such a strategy is typically assumed to prove beneficial, it can actually introduce various complications that are underestimated or even ignored. This study investigates and summarizes the current state of control variable usage in leadership research by qualitatively and quantitatively examining the use of statistical control variables in 10 highly regarded management and applied psychology journals. Despite available “best practices,” our results indicate that control variable usage in existing leadership studies is rarely grounded in theory but instead frequently relies on outdated misconceptions. Moreover, a meta-analysis of the relationships between popular control variables and leadership constructs finds nearly universal weak effect sizes, suggesting that many studies may not only be losing valuable degrees of freedom but also making inferences based on biased parameter estimates. To address these issues, we put forth a number of recommendations to assist leadership scholars with determining whether potential third variables should be controlled for in their leadership research.

Supervisors' emotional exhaustion and abusive supervision: The moderating roles of perceived subordinate performance and supervisor self-monitoring

“…Notably, the pattern and interpretation of the results remained virtually unchanged when including or excluding all of the control variables. Hence, we did not incorporate these variables when testing the study hypotheses (Becker, ; Bernerth et al, in press).…”

Are followers satisfied with conscientious leaders? The moderating influence of leader role authenticity

“…Consequently, we excluded these two factors from further analyses, not only to provide maximum power for the following statistical tests, but because analyses that include unnecessary control variables may yield biased parameter estimates (Becker, ). Nevertheless, we followed best practice recommendations (e.g., Becker et al, ; Bernerth, Cole, Taylor, & Walker, ) and compared results with and without control variables, observing minimal differences between the two sets of results (i.e., no study conclusions were significantly impacted).…”

Strong body, clear mind: Physical activity diminishes the effects of supervisor interpersonal injustice

“…It is important to note that the hypothesized interaction maintains the same direction but falls out of traditional standards of significance without modeling the control variables (p = .04 to .17). We indicate that including control variables that are theoretically relevant is of great importance (Becker et al, 2016;Bernerth & Aguinis, 2016;Bernerth et al, 2018;Carlson & Wu, 2012). In the current context, there are theoretical explanations for why our results without controls would not hold.…”

“…Due to the nonexperimental nature of our study design, it was important to control for theoretically relevant factors (Becker et al, 2016;Bernerth & Aguinis, 2016;Bernerth, Cole, Taylor, & Walker, 2018;Carlson & Wu, 2012). In this study, we postulate physical activity as a Level 2 variable that moderates the relationship of interpersonal injustice and self-regulation depletion modeled at Level 1.…”