Inclusive Work Practices Actually Work: Reconsidering Turnover Intention Among LGBT Employees

Authors

• J. Tobin Grant, Southern Illinois University CarbondaleFollow
• Michelle N. Fletcher, Michelle N. Fletcher

Abstract

Testing the reproducibility of research reinforces the scientific principles and builds cumulative knowledge. This study tests the reproducibility of Sabharwal, Levine, D'Agostino, and Nguyen's (2019) analysis of turnover intention among lesbian, gay, bisexual, and transgender (LGBT) federal employees. Sabharwal et al. partially rejected two of their hypotheses. They reported surprising, counter-theoretical findings such as that open and supportive work practices increase turnover intention among LGBT employees. We employ a reproducibility methodology that duplicates, verifies, and corrects the original empirical tests. We reconstruct the original data. We duplicate most of the original results. We find misreported coefficients, including a reversal of the signs for the effects of inclusive measures in their final model. Verification reveals that none of the original analysis used probability weighting that is necessary for unbiased inferences under the sampling design. There was also a substantial coding error of agency type and tenure. Inferential reproducibility tests find that the original hypothesis tests cannot be trusted given the biased standard errors, incorrect coefficients, and misinterpretation of interaction terms. The corrected models provide a sound basis for hypothesis tests. We find that the original hypotheses are, in fact, supported by the data. Inclusive workplace practices are associated with lower turnover intention among LGBT employees; empowering employees is the most effective practice. Type of agency and gender are also associated with the decision to leave. The revised conclusions advance our understanding of LGBT employee turnover intention and encourage public administrators to continue diversity and inclusion initiatives.