Measurements and Data Analysis

We sent self-reported questionnaires to both team leaders and followers. The questionnaire for the leaders included a work engagement measurement, while the questionnaire for the followers contained both a subjective career success scale and a work engagement scale. To reduce the participants' concerns and to protect their privacy, all the questionnaires and test compensation were enclosed and sealed in envelopes in advance. On the cover of the questionnaires, the following statement was included: “To protect your privacy, please return the completed questionnaire in the envelope and seal it well.” The participants were assigned a random number for identification and matching.

Regarding the questionnaires for the followers, we adopted Harman's one-factor test to examine common method variance (Malhotra et al., 2006). The goodness-of-fit index of the one-factor model and four-factor model are as follows, respectively: χ² = 888.99, df = 152, RMSEA = 0.118, TLI = 0.78, CFI = 0.80, and SRMR = 0.068 (one-factor); χ² = 517.82, df = 146, RMSEA = 0.086, TLI = 0.88, CFI = 0.90, and SRMR=0.056 (four-factor). The result that the four-factor model fit better than the one-factor model revealed that common method variance was possible in the present research; however, its influence could be small.

SPSS 21.0 was used in this research to perform descriptive statistics, correlation analyses, exploratory factor analysis, and reliability analysis. We used Mplus 7.4 (Muthén and Muthén, 1998-2015) to conduct confirmatory factor analysis and multilevel structural equation modeling (MSEM). Missing data accounted for <5%, and we adopted the EM algorithm to account for it.