We investigate the impact of CEO network centrality, a central position in a social network, on bank risk. Using a sample of 471 bank holding companies (BHCs) in the US from 1999 to 2018, we find that CEO network centrality is negatively related to bank risk and this is due to CEOs with higher levels of network centrality implementing less risky policies. Additionally, we find that information flow and CEO power are two channels through which CEO network centrality reduces bank risk. Our empirical results still hold when we employ a battery of methods to mitigate endogeneity issues. Our research further extends a new strand of studies focusing on the specific position/hierarchy of executives/directors in a social network.

We study the impact of the recent government bailout, called Trouble Asset Relief Program (TARP), on bank accounting quality. By adopting a difference-in-difference (DID) method, we find a significantly positive impact of TARP on earnings management of recipient banks, compared with their non-recipient peers. Further, we observe that TARP-recipient banks engage more in earnings-decreasing manipulation rather than earnings-increasing manipulation. This behavior is more obvious for those banks that voluntarily request for TARP funds. Also, participant banks change their accounting strategy to manipulate earnings upwards after TARP funds are paid back. Our findings confirm our hypothesis that TARP-recipient banks are motivated to manipulate downwards (or hide some earnings) to obtain further favorable treatment by the program administrators.

We examine the impact of social networks between independent directors and the CEO on firm risk. Employing the deaths and retirements of socially connected independent directors and the passage of the 2002 Sarbanes-Oxley Act for two identifications, we find that board-CEO social networks have a positive impact on firm risk. Specifically, CEOs who are socially connected to their independent directors are motivated to adopt riskier investment, operating and financing strategies. This positive influence is more pronounced for prior under-performing firms and for CEOs with low power or overconfidence, indicating that board-CEO social networks act as career insurance and a power-enhancing mechanism to encourage managerial risk-taking.

We examine how women on boards influence bank earnings management. Using the likelihood of a board appointing women directors based on a Blau index of gender diversity in each director's total employment connections outside our sample banks for identification, we find an inverted U-shaped relation between women on boards and bank earnings management. Specifically, when there exists only a marginal number of women directors, banks are more likely to manipulate earnings. But, when the number of women directors reaches three or more, bank earnings management declines. This inverted U-shaped impact is intensified if women sit on audit or nomination committees, is moderated if women directors have higher education levels and more board experience, and is unchanged during the 2007–2009 financial crisis. Our results hold when we use alternative measures of bank earnings management, employ GMM estimations and test for alternative hypotheses for the inverted U-shaped relation.