Using unique hand-collected data for 336 large banks across 48 countries, together with values of national culture, our empirical analysis uncovers three new robust findings. First, variations of bank risk-taking across national culture and CEO power are more pronounced when cultural values and CEO power indicators are high. Second, while the individualism dimension of national culture has a moderating influence, the uncertainty avoidance dimension has a reinforcing effect, on the relationship between CEO power and bank risk-taking. In more detail, the results for the average marginal effect of CEO power on risk for different cultural values show that CEO power has a negative (positive) or insignificant impact on bank risk-taking when the value of individualism (uncertainty avoidance) is low; however, the impact becomes positive (negative) and statistically significant as the value of individualism (uncertainty avoidance) increases. Third, intra-cultural diversity matters: ‘tight’ cultures (e.g., strong social norms) are more pronounced than ‘loose’ cultures (e.g., heterogeneous values) in influencing bank risk.

We examine the likelihood that the US Internal Revenue Service (IRS), in its enforcement role, will accord particular attention to firms that are managed by CEOs who exhibit over-confidence, given that such CEOs may be more aggressive in their tax policies and strategies. Using data from 7757 firms, we find that this is indeed the case. Such attention is even more pronounced in the instance of overconfident CEOs whose firms are financially constrained and/or financially distressed. We also find that the IRS has augmented its audit processes to give more attention to overconfident CEOs during and post financial crisis. This may be due to the increased vulnerability of their firms to external shocks, which consequently increases the incentives to embark on tax avoidance strategies, value-destroying investments, and/or highly biased financial reporting (and forecasting responses) to tax authorities. Our results are robust after accounting for the possibility of endogeneity and using a wide range of specifications, measures, and econometric models.