Computational Reproducibility of Heiserman & Simpson (2023)

We report a computational reproducibility assessment of Heiserman and Simpson (2023), who examined the causal effects of workplace discrimination on employees' work effort across five experiments (total analytic N = 1,041) and a manager survey (N = 107). Using the authors' publicly available data and reproducing the published analyses in R (the original used Stata), we closely reproduce the reported reward-expectation results, the main work-effort pattern, the mediation decomposition, and the manager-survey results. We identify seven minor reporting discrepancies between the manuscript, supplement, and reproduced values, none of which changes the substantive conclusions. Extension analyses meta-analytically pool count-model results that the original supplement reports only per-study, conduct additional meta-analytic robustness checks (leave-one-study-out, exclusion sensitivity), and assess pre-registration adherence using Lakens' severity framework. Overall, we find that conclusions weaken under some alternative analytical choices, but that the main claim is robust.