How Does Information Disclosure Affect Bank Systemic Risk in the Presence of a Deposit Insurance System?

ABSTRACT

This article establishes a dynamic game with incomplete information to theoretically analyze the influence mechanism of information disclosure on systemic risk in the presence of a deposit insurance system. To verify the mechanism, we use panel data on 247 global banks in 41 countries during the period 2006 to 2015 in an empirical analysis. Our article finds that a high degree of information disclosure can reduce deposit insurance premiums and weaken the negative incentive from a bailout by regulatory authorities. Moreover, the effect of deposit insurance on financial stability is not apparent, but the synergistic effect of deposit insurance and information disclosure reduces bank systemic risk. Furthermore, different deposit insurance designs affect bank behavior, so it is crucial for bank supervisors to create proper deposit insurance systems, which are helpful in strengthening market discipline and preventing moral hazard thus contributing to a stable financial environment. Therefore, under the deposit insurance system, regulatory authorities should strive to improve the standard of information disclosure to ensure systemic stability.

KEY WORDS:

• deposit insurance system
• dynamic game with incomplete information
• information disclosure
• systemic risk

JEL CLASSIFICATION:

• G21
• G28

Notes

1. If a bank invests in both two industries, the information costs will be high.

2. According to Acharya and Yorulmazer (2007), f(x) = ax, where a > 0 and x stands for fiscal funds.

3. Because there is only one state at time 0, the state i of banks can be omitted.

4. According to Acharya and Yorulmazer (2007), in such a case, the optimal choice for the regulator is that the surviving bank buys the failed bank. Given the degree of information disclosure, banks can take only s proportion of all the deposits at the interest rate $r_1^{SS}$because only s percentage of depositors have the informational advantage, and the rest of the deposits will be taken at the rate $r_1^{SF}$.

5. In the state FF, (1 – s) of the depositors have informational disadvantages so they will exit the market.

6. Detailed mathematical calculations are shown in the Appendix.

7. Because our research concerns the impact of bank regulation on systemic risk, we focus only on banks that are large enough to destabilize the financial system at the global level. The Dodd-Frank Wall Street Reform and Consumer Protection Act in 2010 considers a bank systemically important if its total assets exceed $50 billion. Based on this criterion, our final sample consists of 2,570 bank-year observations of 257 banks in 41 countries over the period 2006 to 2015. An overview of the number of banks included in our sample sorted by country is given in Figure 1.

8. Because of space limitations, the Spearman correlation coefficient matrix among variables is not reported in the text.

9. We select the daily closing prices of 257 listed banks from January 4, 2006, to December 30, 2015, as our sample data and calculate the systemic risk of each bank. Then, we obtain the national annual systemic risk by averaging the values of financial institutions in the country.

10. The proxy variable for the regulatory environment is the regulator rights index, the proxy variable for historical information on crises in the banking system is whether the country has suffered a banking crisis, and the specific construction of variables and data come from Barth et al. (2013).

11. To overcome the shortcomings of the Wald test, we test the differences among groups by bootstrapping. There are four steps. The first step is to obtain the sample containing n observations, which covers n1 banks that experience a banking crisis and n2 banks that do not experience a banking crisis. In the second step, n1 and n2 samples are randomly selected from n samples and assigned to the group with a banking crisis or the group without a banking crisis. The third step is estimating the coefficients of the interaction terms in each group and recording the difference of the coefficients. The last step mainly repeats the second and third step 300 times and then calculates the percentage of the sampling difference and the actual difference, that is, the p-value.

Additional information

Funding

This research was supported by the National Natural Science Fund Project of China [71673225], the China Postdoctoral Science Foundation [2018M632273], and the Fundamental Research Funds for the Central Universities of China from Southwestern University of Finance and Economics [JBK1902022].