Therefore, any evidence on how the CEO pay gap in banks affects bank risk is politically relevant. We provide two competing theories to link the CEO pay gap to bank risk-taking: tournament theory and CEO power. For example, previous non-bank studies relate larger CEO pay gap to higher risks (Kini and Williams, 2012), reduced firm performance (Bebchuk, Cremers, and Peyer, 2011), more fraud (Haß, Müller, and Vergauwe, 2015), but more returns to acquirers (Nguyen, Phan and Tran, 2018) and increased innovation (Shen and Zhang, 2018).
The negative correlation between the pay gap and bank risk mainly stems from incentive-based pay gaps. Indeed, our results confirm that the CEO pay gap is significantly related to CEO proxies.
PSM Estimator
Thus, there is an economically plausible link between each of these instruments and the wage gap (satisfied significance condition), but, intuitively, no plausible explanation as to why they can directly affect bank risk (satisfied exogeneity condition ). Therefore, bank year observations are matched so that a pair is otherwise similar except for the CEO pay gap. Comparing the treatment effect in a matched sample allows for a more robust econometric estimate of the effect of the wage gap on different risk measures (Armstrong et al., 2010).
Analyzing Business Policy Channels
Therefore, in the matched sample of bank-year observations, banks with a high wage gap are associated with lower total and idiosyncratic risk, confirming the results of our previous tests. Typically, MBS are risky and are believed to have contributed to the GFC of 2007-2009; therefore, we expect a negative association with the wage gap. Thus, banks with a large difference in CEO pay increase their use of MBS to spread risk.
Analyzing Bank Disclosure Quality
The last corporate policy measure, revenue from derivatives trading, is associated with increased risk and vulnerability of banks (DeYoung et al. 2013). As a result, a negative relationship between derivatives trading and the wage gap is expected. Contrary to our expectation, the significant positive coefficient in column (3) indicates that banks with a larger wage gap have more MBSs in their portfolio. In summary, we provide evidence of corporate policy channels through which the wage gap is negatively associated with bank risk: greater reliance on deposit financing and less derivatives trading, but greater exposure to MBSs.
Panel A of Table 8 presents the bank FE and 2SLS-IV results for the effect of wage gap on DLLP proxies. The significant negative coefficient on pay gap across all four columns shows that, regardless of our LLP model and estimation method, DLLP decreases with the pay gap. Based on the observed order flow, the PIN is a bank-specific estimate of the probability that a trade originates from a privately informed investor.
Following Brown and Hillegeist (2007), annual PIN is estimated using an extended version of the popular market microstructure model of Easley, Kiefer, and O'Hara (1997). The wage gap is expected to limit private information collection because improved financial reporting quality combined with the wage gap effectively reduces information gaps between traders (Verrecchia, 2001), thereby reducing speculative trading by informed traders (Diamond, 1985) and investors are discouraged from pursuing expensive private transactions. information (Diamond, 1985; Verrecchia, 2001). The significant coefficient on the pay gap in all four columns provides strong evidence that banks with a larger CEO pay gap reduce the PIN, improve the readability and comparability of financial statements, and voluntarily increase disclosure.
In short, a larger pay gap for CEOs improves the quality of banks' disclosures, which helps limit bank risk-taking.
Analyzing Analyst Forecasts and Audit Fees
Next, we assess the effect on the comparability of banks' annual accounts because it lowers the cost of information gathering (Franco, Kothari and Verdi, 2011). The comparability index developed by Franco et al. 2011) is based on the premise that two companies have comparable accounting systems if they prepare similar accounts for a given set of economic events. Panel B of Table 8 presents the results for the other four proxies for bank information quality.
In sum, a larger CEO pay gap improves the quality of banks' disclosure, which helps limit banks' risk-taking. et al., 2010). Excess total audit fees is a dummy variable equal to one if the residuals from the above audit pricing model, as shown in Table OA.2 in the online supplement, are positive and zero otherwise. The last column of Panel A of Table 9 shows the results of the logit estimation of the effects on excess fees paid to bank auditors.
Pay Gap and CEO Power
Our main variables of interest are High_PayGap, an indicator variable that takes a value of one if a bank has a pay gap greater than the average pay gap of the sample banks in a year and zero otherwise, and the interaction term, High_PayGap×Perform, capturing incremental differences in CEO turnover–performance sensitivity in High_PayGap banks relative to that in Low_PayGap banks. We expect the coefficient on performance to be significantly negative and that on the interaction term to be either significantly positive or insignificant to support our conjecture that CEO turnover is either insensitive or less sensitive to performance for banks with gaps higher wages. Columns (1) and (2) of Panel B in Table 10 present the regression results from equation (5), omitting the two terms—High_PayGap and High_PayGap×Perform, separately for the samples of Low_PayGap and High_PayGap banks.
The significant coefficient on the performance proxy in column (1) suggests that CEOs are more likely to be fired following poor stock performance at banks with a lower pay gap, while this is not true for banks with a larger pay gap. In particular, the significant coefficient on the performance and the insignificant coefficient on the interaction term High_PayGap×Perform is evidence that, after poor performance, banks with a small pay gap is 1.07%. In sum, our results in both panels in Table 10 lend credence to the view that a wider pay gap is associated with increased CEO power.
Other Robustness Tests
Pay Gap and Bank Performance
We also investigate whether and to what extent CEO pay differential is related to bank performance because the banking literature provides no guidance on whether ex ante CEO conservatism reduces bank performance. To this end, contrary to the non-banking findings of Bebchuk et al. 2011), both our FE and 2SLS-IV bank estimates provide new evidence in Panel A of Table 11 that bank performance, such as a proxy for return on equity (ROE), return on assets (ROA), stock returns, and adjusted purchase size - and-hold (Adj-BHAR) returns, significantly increased by the difference in payment. Therefore, we find that a larger difference in CEO pay is negatively related to bank risk and positively related to bank performance.
Alternative Bank Risk and Pay Gap Proxies
Subsample Analysis
Our main findings remain robust with the inclusion of these additional covariates, as shown in row (17). In addition, columns (2) and (4) of Table OA.3 of the Online Appendix provide some useful insights. For example, the significant coefficients for CEO_Female in column (2) and CEO_Overconfidence in column (4) provide some evidence that banking risk decreases when the CEO is female, while it increases when the CEO is overconfident.
Exclusion of big banks: There is an implicit idea that the biggest banks are TBTF. Due to the potential economic impact of the failure of a major financial institution, the government is expected to intervene to prevent these banks from collapsing. Therefore, CEOs in these so-called TBTF banks may have different risk incentives than the CEOs in other banks.
To ensure that these banks do not influence the results reported in Table 3, the basic regression equation (1) is re-estimated, excluding the top 10 banks in each year. The significant negative coefficient of the total risk pay difference in row (18) proves that our findings are robust except for the top 10 banks. Finally, our main finding also holds for a balanced subsample of banks, with each bank having at least five observations, as shown in row (19).
Concluding Remarks
The finding that the pay gap is associated with lower bank risk has implications for the response and interpretation of the employee ratio to be reported in 2018 and beyond. Idiosyncratic The standard deviation of the residuals from a single index market model, estimated each year for each bank. Tail The marginal expected shortfall measured as the negative of the bank's average return over the 5% worst performing days of the CRSP value-weighted market index return.
ComAcct_10 Average of the 10 highest CompAcct values for bank i, where ComAcct is the absolute value of the difference between the predicted value of the regression of bank i's earnings on bank i's return using the estimated coefficients for bank i and j, respectively. Zer Index Based on the December 2009 and 2011 Financial Stability Reports from the Bank of England, Zer's (2015) disclosure index is the results of principal component analysis of 14 sub-indices that capture four main categories: liquidity risk, risk positions of key group affiliates and subgroups, intra-annual information and spillover risk information. Forecast spread The standard deviation of all analysts' earnings forecast for bank i during the fiscal year divided by the stock price for bank i at the beginning of the fiscal year.
Forecast accuracy The absolute value of the bank's forecasted average earnings over the fiscal year minus the bank's actual earnings divided by the stock price at the beginning of the fiscal year. Revenue Growth The percentage increase in total revenue from the beginning of each year to the beginning of the next year. Equity measured by the equity ratio, calculated as the ratio of the book value of equity to the book value of total assets.
Institutional Ownership Four-quarters average of the proportion of total outstanding shares owned by all institutional shareholders. Adj-BHAR Size-adjusted buy-hold return is calculated as the fiscal year buy-hold return over the buy-hold return on a value-weighted portfolio based on size deciles of NYSE/AMEX banks. In panel D, for each of the subsamples/specifications, we have presented the results for total risk only to save space, since the interpretation of coefficients for other bank risk proxies is similar.
Firstly, Table OA.1 contains the pooled OLS estimates of two variants of the LLP model, i.e. this table presents the pooled OLS estimates of two variants of the following LLP model, namely