Government size—the degree of participation by governments in the economy—has been considered a potential breeding ground for corruption. However, heterogeneity in reported findings— reflecting different viewpoints on the role of large governments—makes it difficult to assess the size of the effect and, consequently, the design of anti‐corruption measures. To address this issue, we perform a meta‐regression analysis (MRA) of the literature on government size and corruption, examining 450 empirical estimates retrieved from 44 primary papers published from 1998 to 2022. We find that the considerable heterogeneity in results depends mainly on whether or not the paper is published, if it accounts for endogeneity, and whether it uses panel or cross‐sectional data. There is evidence of bias in favor of publishing studies reporting a positive effect‐size estimate. However, after controlling for publication selection bias, a negative or zero mean effect remains, which overturns the conventional wisdom. Moreover, the type of measures of corruption has a significant impact on the sign of the relationship with government size. Finally, several robustness checks confirm our main results.