How robust is our cumulative knowledge on turnover?

Although systematic reviews are considered the primary means for generating cumulative knowledge and their results are often used to inform evidence-based practice, the robustness of their meta-analytic summary estimates is rarely investigated. Consequently, the results of published systematic reviews and, by extension, our cumulative knowledge have come under scrutiny. Using a comprehensive approach to sensitivity analysis, we examined the impact of outliers and publication bias, as well as their combined effect, on meta-analytic results on employee turnover. Our analysis of 205 distributions from seven recently published meta-analyses revealed that meta-analytic results on turnover are often affected by publication bias and, less frequently, outliers. Moreover, we observed that 33% of the recommendations for practice provided in the original systematic reviews on turnover were not robust to outliers and/or publication bias, which, if implemented by practitioners, could yield unexpected consequences and, thus, widen the science-practice gap. We argue that practitioners should be skeptical about implementing practices recommended by meta-analytic studies that do not include sensitivity analyses. To improve sensitivity analysis reporting rates and, thus, the transparency of meta-analytic findings and recommendations for practice, we introduce an open-access software ( metasen.shinyapps.io/gen1/ ) that conducts all analyses performed in the current study. We provide guidelines and recommendations for future turnover studies and sensitivity analyses in the meta-analytic context.