Robustness analysis assesses the fragility or stability of empirical findings by testing whether research findings remain stable across alternative, justifiable analytical choices. Understanding the robustness of empirical results is important both for management scholars to build reliable theories and for management practitioners to derive appropriate practical implications for decision-making. Yet, the management field currently faces three problems in designing and reporting robustness analysis: terminological confusion, fragmented recommendations across isolated topics without integration, and a lack of systematic evidence on actual reporting practices. This paper clarifies the concept of robustness analysis and distinguishes it from related credibility-enhancing practices. Using a framework encompassing six key dimensions of robustness (i.e., methods, measurements, covariates, preprocessing, subsampling, and statistical specifications), we review and synthesize the currently fragmented recommendations for each dimension. Next, to systematically review the reporting of robustness practices, we develop a hybrid Generative AI-assisted coding approach enabling the analysis of a larger volume of articles than feasible with human coding alone. In a systematic review of 1,706 articles containing 2,770 quantitative sub-studies published in seven leading management journals (2020–2025), we find that reporting of robustness analyses is often done narrowly and inconsistently, scattered across multiple article sections, and using different terms. Our findings reveal gaps between recommended and documented practices, between macro and micro domains, and across time periods. Based on the review results, we offer recommendations for conducting and reporting robustness analyses.
Yuan et al. (Fri,) studied this question.