A Computational and Robustness Reproduction of "Ramadan Fasting Increases Leniency in Judges from Pakistan and India"

Mehmood et al. (2023) estimate the effect of Ramadan fasting hours on judicial decisions using case-level data from Pakistan and India. Using exogenous variation in fasting intensity due to the Islamic lunar calendar and latitude, the authors find that Muslim judges are significantly more likely to issue acquittals with longer fasting hours. Their main result, reported in Table 1, shows that each additional hour of fasting beyond the baseline minimum increases acquittal rates by about 10%. I successfully computationally reproduced this main result using the original authors' data and code: I found no coding errors or discrepancies in the replication package, and the point estimates and p-values in my reproduction match those reported in the published article. I then evaluated robustness for the Pakistan sample of judges using three alternative specifications. First, the result is robust to alternative inclusion of control variables: It remains stable and statistically significant whether controlling for case-level covariates, judge-level covariates, both, or none. Second, the effect persists across a different fixed effects specification that includes only district fixed effects. Finally, the result is robust to clustering standard errors at the judge level, although clustering at the month level increases the standard error and renders the estimate statistically insignificant. Overall, the authors' main finding is both computationally reproducible and robust.