We conduct a field experiment in a large corporation to investigate effects of supportive leadership behaviors on worker satisfaction, engagement and performance. The CEO encourages all employees to participate in team discussions to provide input on a question of strategic importance. Employees indicate availability in the week of implementation in the baseline survey. The discussions take place on four consecutive workdays. The employees with no personnel responsibility are randomly matched to employees with personnel responsibility (leader) to form groups of three or four members (accounting for availability), all from different companies within the corporation to maximize the likelihood of them not knowing each other. Under the guidance of the leader, each group spends about 30 minutes discussing the question of strategic importance and summarizes their recommendation for the CEO on a power point slide. Using block randomization, the leaders are randomized into treatment and control. Prior to the group discussions all leaders receive a brief guideline on how to guide the discussion using Microsoft Teams. The instructions are identical, but for the treated leaders it also contains a brief leadership-training promoting more supportive leadership behaviors. We investigate treatment impact on work satisfaction, engagement and performance in the group task.
When estimating treatment impacts, we use several model specifications investigating robustness to adding sets of control variables. All models control for blocks. Additional control variables are: leader sex, leader tenure (years working at the company), leader coaching behavior at baseline (Heslin et al., 2006), leader mindset at baseline (Chiu et al., 1997), group size, group member average tenure, group member average creative mindset (Karwowksi, 2014) at baseline.
We investigate differential treatment effects across leader’s sex, leader’s tenure (median split), leader mindset at baseline (measured using the 8-item scale of Chiu et al. (1997), median split), and coaching behavior at baseline (measured using a 6-item instrument developed by Heslin et al. (2006) median split).