Charter value, risk-taking and systemic risk in banking before and after the global financial crisis of 2007-2008

ABSTRACT

We investigate how bank charter value affects risk for a sample of OECD banks by using standalone and systemic risk measures before, during, and after the global financial crisis of 2007–2008. Prior to the crisis, bank charter value is positively associated with risk-taking and systemic risk for very large ‘too-big-too-fail’ banks and large U.S. and European banks but such a relationship is inverted during and after the crisis. A deeper investigation shows that such a behaviour before the crisis is mostly relevant for very large banks and large banks with high growth strategies. Banks’ business models also influence this relationship. We find that for banks following a focus strategy, higher charter value amplifies both standalone and systemic risk for large U.S. and European banks. Our findings have important policy implications and cast doubts on the relevance of the uniform more stringent capital requirements introduced by Basel III.

KEYWORDS:

• Systemic risk
• standalone risk
• charter value
• global financial crisis
• bank regulation

JEL CLASSIFICATIONS:

• G21
• G28
• G32

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

¹ Jones, Miller, and Yeager (2011) emphasize three factors to explain the increase in charter value during the 1988–2008 period: a rise in banks’ noninterest income, a run-up in the stock market, potentially ‘irrational exuberance’, and a strong economic growth.

² M&A operations have significantly reduced the degree of competition and have positively affected prices and margins. They were achieved for strategic reasons, such as improving market share, profitability, or efficiency (Jones, Miller, and Yeager 2011; De Jonghe and Vander Vennet 2008).

³ Laeven and Levine (2007) argue that SIFIs engaged in multiple activities (charter-gain-enhancing) suffer from increased agency problems and poor corporate governance that could be reflected in systemic risk. Demirgüç-Kunt and Huizinga (2013) find that banks that rely to a larger extent on non-deposit funding and non-interest income are more profitable but also riskier.

⁴ Banks in these three geographical areas have very different business models and operate in differently organized banking systems. U.S. and European banks are more market-oriented; whereas, Australian, Canadian and Japanese banks are more reliant on traditional intermediation activities. Haq et al. (2019) argue that Australian and Canadian banks appear to pursue safer policies, even before the GFC (1995–2006), hence preserving financial stability.

⁵ Stock returns are computed in local currency terms. Annual income statement and balance sheet data are converted into U.S. dollars.

⁶ From 988 banks, we end up with 853 banks due to our data cleaning process as well as the data availability that varies depending on the combination of variables used in regressions. Our sample consists of 22 European countries, three Americas countries (U.S., Canada and Mexico) and three Asian-Pacific countries (Japan, South Korea, Australia). Iceland and New Zealand were dropped because of insufficient liquid stocks (see Table 1).

⁷ Community banks are known for their focus on traditional banking activities, i.e. lending and deposit activities within a limited market area. Hence, we exclude community banks, i.e. those with total assets less than $500 million, ratio of total loans to total assets above 33%, and the ratio of total deposits to total assets above 50%.

⁸ We use rolling regressions of a bank’s daily stock returns on market returns, as a return generating process. We estimate risk measures for each bank using a moving window of 252 daily observations.

⁹ Economically, the term ‘marginal’ refers to the bank’s capital shortfall stemming from each unit variation in the equity value ${\mathrm{M}\mathrm{E}\mathrm{S}}_{\mathrm{i},\mathrm{t}}^{\mathrm{q}}$. The MES measures the increase in systemic risk induced by a marginal increase in the exposure of bank i to the system.

¹⁰ To estimate risk measures, we either employ the financial sector index for the most developed financial market or the broad market index.

¹¹ As MES, CoVaR is a conditional VaR computed at time t given the information available at time t-1 based on the financial system Expected Shortfall.

¹² According to the CVH, regulation promotes bank franchise value through more entry restrictions and more market concentration enhancing profit opportunities. By contrast, deregulatory efforts that increase financial service competition may erode charter value and thereby increase risk-taking incentives (Anginer, Demirguc-Kunt, and Zhu 2014; Allen and Gale 2004; Hellmann, Murdock, and Stiglitz 2000).

¹³ Anginer, Demirguc-Kunt, and Zhu (2014) and Allen and Gale (2004) argue that in highly competitive markets, banks earn lower rents, which also reduces their incentives for monitoring.

¹⁴ Because we analyse a global sample of banks, the data we use might be subject to country specific reporting standards. Consequently, we control for this possible bias by estimating our baseline regressions using country-level controls.

¹⁵ The differences in the number of observations is due to missing accounting and market data for some banks.

¹⁶ We compute the variance inflation factor (VIF) for each model estimates. The VIF statistics are always higher than 10, suggesting the absence of major multicolinearity issues.

¹⁷ In earlier models, charter value is usually assumed to be exogenous (e.g. Keeley 1990; Gropp and Vesala 2004).

¹⁸ Banks with higher default risk could have a higher market-to-book asset ratio if deposit insurance were underpriced and its value were capitalized on the market (but not on the book). Riskier banks could be over valuated, because risk-shifting increases the option value of equity (Keeley 1990).

¹⁹ Although core deposits are regarded as important to explain charter value (Jones, Miller, and Yeager 2011), we do not introduce them in the regressions because of insufficient observations for banks in countries other than the U.S. Similarly, we do not use the entry denied index as an instrument of charter value, such as in (Laeven and Levine 2009), because the index is not available for almost all the countries, including the U.S., during the 2008–2012 period. Instead, we use a proxy of market power.

²⁰ We follow Keeley (1990), Gropp and Vesala (2004) and González (2005) who use the same model specification. Keeley (1990) and Gropp and Vesala (2004) consider the potential endogeneity of Tobin’s Q and apply a two-stage procedure to analyse its possible influence on bank risk-taking.

²¹ To confirm the validity of the IV, we report the KP rank F-statistics. Under identification test is also assessed by the KP Cragg-Donald Wald F-statistics of the first stage (the null hypothesis of weak instruments is rejected if F-statistic is greater than the Stock-Yogo’s critical value (Stock and Yogo 2005; Cragg and Donald 1993)).

²² Overall, the KP rank LM rejects the null hypothesis at the 1% level, indicating that the models are well identified. The Partial F-statistic, of the KP rank Wald F-test, from the first stage rejects this null hypothesis that the instruments are weak at the 1% level. Hansen’s j tests (p-values) for overidentification of instruments show that the instruments are valid. Unreported first stage regression results and tests show that these instruments are both relevant and exogenous.

²³ [0.17*(−0.65)]/1.55 = −7% and [0.17*(−0.17)]/0.53 = −5%. This is also associated with 11% and 3% standard deviation reduction in the individual bank’s systemic risk exposure (the MES) and volatility risk (systematic risk), respectively.

²⁴ Studies using similar definitions include Saheruddin (2014), Berger and Bouwman (2013) and Temesvary (2014), among others.

²⁵ Based on the standard deviations of the charter value and the mean values of the MES over the pre- and post-crisis periods, respectively.

²⁶ Considering sub-samples over the three sub-periods instead of the model in Equation (11)(11) $Ris{k}_{i,t}=({\beta }_1+{\beta }_{2}Factor)\times {\widehat{Charter}}_{i,t}+{\beta }_3{X}_{i,t-1}+{\beta }_5{C}_{i,t}+{\lambda }_t+{\mu }_i+{\epsilon }_{2i,t}$(11) with interaction terms yields similar conclusions (see Table A1 in Appendix).

²⁷ The diversification ratio is defined as noninterest income over total income.

²⁸ Growth strategy (business model) variation is computed as the change over the pre-GFC period (between 2000 and 2006) in total assets (diversification ratio) over the average total assets (diversification ratio) (see descriptive statistics, Table 3).

²⁹ Table 3. contains the definitions of growth strategy and business model (activity-mix).

³⁰ We use the ratio of non-interest income to total income as the diversification ratio. Alternately, we consider the ratio of non-interest income to operating income and obtain similar results.

³¹ To save space, Table 8 does not report the results obtained for alternative risk measures─ Tail beta and Specific risk.

³² We calculate the standardized market value added MVA as (current market capitalization – total equity) divided by total equity.