How should we measure bank capital adequacy? A (simple) proposal Lucy Chernykh Clemson University, Clemson, SC [email protected] Rebel A. Cole * DePaul University, Chicago, IL [email protected]

Abstract: In this study, we test the predictive power of several alternative measures of bank capital adequacy in identifying U.S. bank failures during the recent crisis period. We find that an unconventional ratio—the non-performing asset coverage ratio—significantly outperforms Baselbased ratios including the Tier 1 ratio, the Total Capital Ratio, and the Leverage ratio— throughout the crisis period. It also outperforms in predicting failures among “well-capitalized” banks (as defined by the current Prompt Corrective Action guidelines). Based on our results, we argue that NPACR outperforms other ratios in at least five aspects: (i) it aligns capital and credit risks—the two primary risks of bank failures—in one measure; (ii) it is easier to calculate than the Tier 1 and Total Capital ratios, as it requires calculation of no complex risk weights; (iii) it allows one to account for various time period and cross-country provisioning rules and regimes, including episodes of regulatory forbearance and cross-country differences; (iv) it removes the incentives of both banks and regulators to mask capital deficiencies by creating/requiring insufficient loan-loss reserves; and (v) it outperforms all other commonly used capital ratios in predicting bank failures. We believe that all the above features of proposed measure promise its effective use in the prompt corrective actions by bank regulators. We also expect that this single and informative measure of bank risk can be efficiently used in empirical banking studies. The results of this study also shed light on regulatory forbearance during the recent banking crisis. Keywords: bank, bank failure, capital, capital adequacy, capital ratio, forbearance, prompt corrective action JEL Classifications: G21, G28, G32, G33

DRAFT: January 15, 2014 * Corresponding author: 1 E. Jackson Blvd. Room 5531, Chicago, IL 60604 Office: 1-312-362-6887

1

How should we measure bank capital adequacy? A (simple) proposal

1. Introduction How should we measure bank capital adequacy? The answer to this seemingly simple question is the central piece of the prudential regulation in modern banking and the core element of ongoing regulatory, practitioners’ and academic debates. Regulators need a simple and timely measure that is a reliable and robust indicator of a bank insolvency risk and that can trigger prompt corrective actions. Practitioners need a simple, intuitive, and robust measure that is easy to calculate and monitor. Scholars want all of the above—plus the ability to calculate a consistent measure over long time series and across countries whose bank regulators collect only rudimentary supervisory data. Amid the evolving Basel accords, regulators around the world have used increasingly complex measures of bank capital adequacy. As these measures become more complex and tedious to calculate, they account for a broader range of potential bank risk factors, but differ substantially among banks of different size, scope of operations, and from different national regulatory regimes. Haldane (2011) notes that, under Basel I, only a few calculations would produce a representative large bank’s regulatory capital ratio; under Basel II, closer to 200 million calculations are needed. The latest iteration of capital adequacy rules under Basel III does little to reverse this mind-numbing complexity. Haldane (2012) argues that “the type of complex regulation developed over recent decades might not just be costly and cumbersome but suboptimal for crisis control.”

1

In this study, we follow Haldane’s advice by offering a very simple, timely, and robust measure of capital adequacy that we argue is superior to Basel regulatory capital ratios. We support our claim with macro- and bank-level evidence from the U.S. banking system that documents early warning performance for our proposed capital adequacy measure that is superior to the Basel regulatory capital ratios. Our proposed capital adequacy ratio, which we call the Nonperforming Assets Coverage Ratio (NPACR), explicitly accounts for the capital-constrained banks’ reluctance (or inability) to build up adequate reserves for anticipated future loan losses, and for regulators’ forbearance in enforcing loan-loss reserving requirements. More specifically, our proposed simple formula for the NPACR ratio is as follows: total equity capital plus loan-loss reserves less nonperforming assets, all divided by total assets (all in book values). Each component of this formula is readily available from a representative bank’s regulatory filings. The intuitive interpretation of the NPACR as a capital adequacy measure also is straightforward: it is the ratio of equity to assets when every bank is forced to adequately provision against its non-performing assets. Insufficient loan-loss reserves against accumulated nonperforming bank assets should effectively reduce bank capital adequacy while excessive loan-loss reserves should strengthen the bank’s capital position. We empirically test the NPACR ability to detect problem banks and to predict bank failures using the large sample of U.S. depository institutions around the recent financial crisis. When compared to other commonly used regulatory capital ratios, we find that the NPACR is more sensitive as an early warning indicator of bank solvency problems. Using U.S. bank-level data for the 2007 – 2012 period, we also show that the NPACR reveals severe regulatory forbearance and delays in closing banks extreme levels of non-performing assets and grossly 2

insufficient loan-loss reserves, but with “adequate” traditional capital ratios. This forbearance enabled U.S. bank regulators to skirt the prompt corrective action (PCA) regulations put into place after the last banking crisis, which saw more than 1,000 bank closures during 1984 – 1992.1 Finally, we document that even in the face of the regulatory forbearance, NPACR can predict bank failure outcomes at least as good as the more computationally intensive measures like Basel 2 risk-weighted capital ratios. By relying on the macro- and bank-level evidence provided in this study, we argue that NPACR has at least five important strengths compared to all other commonly used capital adequacy ratios. First, it allows us to account for the two primary banking risks—capital adequacy and asset quality—in one simple measure. Second, it is informative, intuitive, and easy to calculate; as such, it simplifies bank monitoring by regulators and market participants. Third, it eliminates bank management incentives to mask capital deficiencies by failing to create sufficient loss reserves and/or failing to write off non-performing assets as required by regulations. Fourth, the incorporation of NPACR information into the formal bank supervision and prudential regulation is a low-cost solution that should constrain regulators’ discretion in addressing shortcomings of loss recognition and provisioning by problem banks. Last, but not

1

Section 131 of the FDIC Improvement Act of 1991 included a set of regulations designed to eliminate regulatory forbearance by requiring U.S. bank regulators to place into conservatorship within 90 days any depository institution whose tangible equity capital fell below two percent of assets. By failing to require that troubled banks write off non-performing assets, regulators were able to avoid this requirement during the 2009-2013 period. At the end of 2013, the FDIC reported that more than 500 banks remained on its “problem bank” list, while only 22 such banks were closed during all of 2013. Hence, it appears that forbearance is alive and well in the U.S. 3

the least, the NPACR can be successfully used as a capital adequacy measure across countries, and, within countries, across time.2 At the same time, on a less optimistic note, we presume that the NPACR-based monitoring would be adamantly opposed not only by the banking industry, but also by bank regulators, as it closes a number of loopholes for the regulatory arbitrage in bank capital management and limits the discretion (i.e., “regulatory forbearance”) of bank supervision in enforcing prompt corrective actions. However, from the banking system stability and the market discipline enhancement points of view, the NPACR measure offers easy-to calculate but informative indicator that brings to the surface the banks’ actual capital position and singles out seemingly well-capitalized banks with massive nonperforming assets and insufficient cushion for these anticipated losses. Our study makes three important contributions to the literature on financial institutions. First, we contribute to the literature on bank capital adequacy (See, e.g., Pettway, 1976; Sharpe, 1978; Buser, Chen, and Kane, 1981; Kashyap, Rajan, and Stein, 2008; Allen, Fulghieri, and Mehran, 2011; Admati et al., 2013; Rosengren, 2013). We demonstrate that the simple and intuitive NPACR outperforms the complex Basel regulatory capital ratios in forecasting bank failures in the U.S. Because the NPACR is simple to calculate across time and across countries, it holds great promise for researchers looking to analyze capital adequacy, but limited by the availability of supervisory data needed to calculate Basel regulatory capital ratios.

2

Data needed to calculate Basel regulatory capital ratios simply does not exist prior to 1991. However, Estrella, Park, and Peristiani (2000) estimate regulatory capital ratios for U.S. banks in 1988-1990 based on the Capital Adequacy Guidelines published by the Board of Governors of the Federal Reserve System 4

Second, we contribute to the literature on regulatory forbearance and prompt corrective action (see, e.g., Dahl and Spivey, 1995; Jones and Kuester-King, 1995; Aggarwal and Jacques, 2001). We provide convincing new evidence that U.S. bank regulators engaged in a massive scheme of forbearance during 2008 – 2013 that subverted the prompt corrective action provisions of the FDIC Improvement Act of 1991. Our evidence points to the need for new laws and/or regulations designed to limit the discretion of regulators in enforcing laws duly passed by the U.S. Congress. Third, we contribute to the literature on bank failures (see, e.g., Sinkey, 1975; Bovenzi, Marino, and McFadden, 1982; Lane, Looney, and Wansley, 1986; Thomson, 1992; Cole and Gunther, 1995, 1998; Estrella, Park, and Peristiani, 2000; Cole and White, 2012; Berger and Bouwman, 2013). We offer a new measure of capital adequacy, and then demonstrate that it is superior to Basel regulatory capital ratios in predicting bank failures. Because of its simplicity, the NPACR holds great promise for researchers seeking to analyze bank failures across time and/or across countries. The rest of the paper is organized as follows. In Section 2, we provide a short review of the literatures on capital adequacy, bank failure, and prompt corrective action. In Section 3, we briefly discuss the evolution of the regulatory capital ratios in the U.S. banking system and the recent state of the regulatory and academic debate on the reliable bank capital measures. Section 3 also introduces the NPACR measure by illustrating its relevance in detecting industry- and bank-level capital deficiencies during the recent financial crisis. Section 4 describes our dataset and alternative capital ratio measures used in this study. Our main empirical evidence, including descriptive and regression analyses results, is presented in Section 5. Section 6 concludes the study by discussing the regulatory and practical implications of our findings. 5

2. Literature Review

2.1. Capital Adequacy

{IN PROCESS}

2.2. Bank Failures In this section, we will not try to provide a complete literature review on the causes of bank failures because recent papers by DeYoung and Torna (2013) and Demyanyk and Hasan (2009) contain extensive reviews; we refer interested readers to those studies for further depth. Instead, we will summarize papers that have econometrically explored the causes of recent bank failures.3 Ng and Roychowdhury (2010) use a Cox proportional-hazard model to find that bank failures in 2008 and 2009 were positively related to additions to loan loss reserves in 2007, after controlling for other bank characteristics. However, unlike our study, Ng and Roychowdhury do not examine the influence of bank characteristics for years that were earlier than 2007 on their sample of bank failures; they do not include as failures those banks that were so financially weak

3

We exclude from this category the extensive, and still growing, literature on the failures of the subprime-based residential mortgage-backed securities (RMBS). For examples of such analyses, see Gorton (2008), Acharya and Richardson (2009), Brunnermeier (2009), Coval et al. (2009), Mayer et al. (2009), Demyanyk and Van Hemert (2010), and Krishnamurthy (2010). It is striking that, in the literature reviews provided by Torna (2010) and Demyanyk and Hasan (2009), there are no cites to econometric efforts to explain recent bank failures (except with respect specifically to RMBS failure issues). 6

that they were likely to fail after 2009; and they fail to distinguish between loans for residential real estate and for commercial real estate – a distinction that we find to be quite important. Aubuchon and Wheelock (2010) examine bank and thrift failures between January 1, 2007 and March 31, 2010. Unlike our study, however, Aubuchon and Wheelock focus mostly on the regional economic characteristics that are associated with bank failures rather than on the detailed characteristics of the banks themselves. Cebulla, Koch and Fenili (2011) investigate factors that explain bank failures over the 1970-2009 period. They find that passage of the FDIC Improvement Act of 1991 was associated with lower failure rates whereas the Riegle-Neal Interstate Banking and Branching Efficiency Act of 1994 was associated with higher failure rates. Jin, Kanagaretnam, and Lobo (2011) analyze whether the quality of accounting and audit explains bank failures. They find that banks audited by Big Four auditors and by bank specialists have lower probabilities of default than do other banks. Cole and White (2012) analyze determinants of bank failures during 2009. They find that the determinants of failure are largely the same as the determinants of failures in the 1980s as documented by Cole and Gunther (1995, 1998) and others. Specifically, banks with more capital, higher earnings, fewer non-performing assets, and fewer investments in commercial real estate mortgages and construction & development loans have lower probabilities of failure. Shaffer (2012) analyzes banks that failed during a two year window beginning at yearend 1984, 1989, and 2008. In contrast to Cole and White (2012), he finds that the linkages between traditional risk ratios and the probability of failure are different in the recent crisis as compared with the 1980s crisis.

7

Cole and Dahl (2013) analyze whether the incidence, rather than the quality, of audit explains bank failures. They find that audited banks were three times more likely to fail during the crisis years, but that this difference is fully explained by the factors identified by Cole and White (2012). DeYoung and Torna (2013) analyze the impact of nontraditional activities on the probability of bank failure. They find that greater involvement in asset-based nontraditional activities, such as venture capital, investment banking, an asset securitization, is associated with higher probabilities of failure, while greater involvement in fee-based activities, such as securities brokerage and insurance sales, is associated with lower probabilities of failure. Lu and Whidbee (2013) examine the impact of charter type, holding company structure and measures of fragility on the probability of bank failure. They find that fragile banks, newly chartered banks, and banks that had grown aggressively had higher probabilities of failure. Kerstein and Kozberg (2013) use accounting proxies for CAMELS components to predict bank failures. They find that proxies for each of the six components are significant in explaining 204 bank failures that occurred during 2008 – early 2010. Berger and Bouwman (2013) examine how capital affects the probability of bank failure and how this effect varies across normal and crisis times. They find that more capital reduces the probability of failure for small banks during all times, but for large banks only during crisis times.

2.3. Prompt Corrective Action and Forbearance

{IN PROCESS} 8

3. Background 3.1. Evolution of the regulatory capital ratios in the U.S. banking industry: Past, present and future. Given the central role of the bank capital requirements in the modern prudential regulation it is hard to believe that the explicit capital adequacy rules and in the US banking system did not exist until the early 1980s. Indeed, the capital adequacy supervision in the 1960s and 1970s relied heavily on the bank-specific evaluations, regulators’ subjective judgments, peer comparisons, and case-by-case recommendations prescribed without any uniform enforcement rules and penalties. Not surprisingly, as it is well documented in the literature now, the lax capital regulation during this period lead to a gradual decline in the capital ratios in the US banking system. However, during the last three decades, U.S. regulators have set a number of different rules and minimal capital level thresholds. The first step along this way is associated with the introduction of the primary capital ratio in 1981. The primary regulatory capital was defined as the sum of the common equity, preferred equity, perpetual debt, convertible debt, and reserves for loan losses; the on-balance sheet assets served as a denominator to construct the capital ratio. By 1985, the minimal primary capital requirement ratio for all U.S. banks, large and small, was set at a uniform level of 5.5%. The next and more elaborated stage of the strengthening capital adequacy requirements involved the Basel 1 recommendations that rely on the Tier 1 and Tier 2 capital definitions, four distinct credit risk categories and their corresponding weights, and the risk-weighted assets that account for total, balance sheet and off-balance sheet items (see Table 1 for the summary of 9

these ratios’ definitions). The Tier 1 capital risk-based ratio and the Total capital Basel 1 capital ratios were gradually introduced in the US banking system starting from 1989 and became mandatory since the year-end of 1992. At about the same time, in 1991, in addition to the Basel 1 risk-based ratios, the US regulators have also introduced the Tier 1 Leverage ratio, defined as Tier 1 capital to total balance sheet assets. [Insert Table 1 here.] Collectively, the three regulatory ratios described above served as a framework for the Prompt Corrective Action (PCA) guidelines that became effective in 1991, following the passage of the Federal Deposit Insurance Corporation Improvement Act (FDICIA). The major dual intents of the PCA provisions were to empower the FDIC to close critically undercapitalized banks and to mitigate the regulatory discretion and forbearance in the termination of high capital risk FDIC-insured institutions. In Appendix Table 1, we summarize the definitions of the capital adequacy categories under the PCA standards. The follow-up Basel 2 Accord recommendations, initially published in 2004, brought to life the three regulatory pillars framework and long delays in its implementation in the US. In terms of the capital requirements, Basel II attempted to account for a combination of credit, market and operational risks and put a special emphasis on the role of the market discipline in preventing banks from excessive risk-taking. Overall, the recent crisis has revealed a number of weaknesses in the Basel 2 framework as its capital adequacy recommendations not only failed to provide an adequate cushion for the materialized risks but, according to some critiques, have also magnified the banks’ counter-cyclical exposure to capital risks. The current and the widely debated stage of the capital regulation development has emerged under the Basel 3 standards issued in 2010 in response to the recent crisis and the Basel 10

2 criticism and failure. The new Basel accord follows the previously established pillars’ structure. There are, however, notable and major changes in the Pillar 1 recommendations that redefine and enhance banks’ capital cushions and buffers. The revised Pillar 1 put more emphasis on the role of the common equity, countercyclical buffers, and liquidity risks. In line with more focus on the capital quality Basel 3 also introduces a new ratio - Common Equity Tier 1 risk-based capital ratio (or CET1). There are also several changes and adjustments in the definition of the prior three regulatory capital ratios that substantially increase their detailization and the computational complexity. 4 In mid 2013, U.S. regulators approved a new regulatory capital framework that incorporates both the Basel 3 standards and the related components of the Dodd-Frank Act. The gradual introduction of these more complex capital regulations is set up as a multi-step process and is expected to be fully in place by 2019. As a part of this large-scale regulatory reform, the PCA guidelines are also modified to incorporate revised ratios and definitions. Although it is still early to speculate about the long-term anticipated and unanticipated effects of the new capital requirements on banks and their willingness and ability to extend credit to the economy, the commonly shared concern is that the complicated, multi-layered rules may open up a wide array of the new regulatory arbitrage opportunities and/or undermine the effectiveness of the market discipline in tracking banks’ capital positions.

5

4

See the U.S. Federal Reserve Board’s webpage for the Basel Regulatory Framework at: http://www.federalreserve.gov/bankinforeg/basel/USImplementation.htm#baseIII) 5

See ibid. 11

3.2. Nonperforming assets, loan-loss reserves, and capital-adequacy ratios: Two decades of macro-level evidence In Figures 1 and 2, we present macro-level evidence for U.S. banks capital and loss reserves management and the regulatory forbearance during the recent financial crisis. Using the publicly available year-end FDIC statistics for the last two decades (1992 to 2012), we track the industry-level trends in the accumulation of the non-performing assets and the corresponding loan-loss reserves against anticipated future losses. We also construct a number of key ratios, including the coverage ratio of loan-loss reserves to NPLs, three commonly used regulatory capital ratios, and our proposed NPACR ratio. The shaded area on both figures—2007 to 2012— indicate the recent crisis and immediate post-crisis years. [Insert Figures 1 and 2 here.] As expected, the industry-level volume of nonperforming assets that we calculated conservatively as: NPA

= 20% *Assets Past Due 30 to 89 days + 50% *Assets Past Due 90 or more days + Assets in Nonaccrual Status + Real Estate Owned Assets]

[1]

exhibits a more than six-fold jump during the peak of the recent crisis period, from $55.0 billion in 2006 to $363.6 billion in 2009. However, the increase in the loan-loss reserves during the same period is relatively modest—from $69.1 billion in 2007 to only $213.8 billion in 2009. Notably, the wide gap between the NPA and LLR remains relatively constant during the subsequent period from 2010 – 2012, and reveals inadequate provisioning for anticipated loan losses. The pattern of the Coverage ratio further confirms inadequate provisioning and regulatory forbearance during the crisis. As the crisis starts to unfold, the Coverage ratio drops 12

from its peak of 154.2% in 2004 to below 100% in 2007, then quickly sinks to 58.8% in 2009, and stays at about 60 to 65% during 2010 – 2012. These numbers reveal that, during and after the recent financial crisis, the banking industry failed to create sufficient reserves for at least 1/3 of anticipated future loan losses. Although we do not have complete data series for the savingsand-loan crisis of the 1980s and 1990s, the 1992 situation (the starting point of the current FDIC public disclosures) seems to be very similar to the one described above: NPA levels visibly exceed LLR levels and, thus, push the Coverage ratio well below 100% (or 59.2% to be precise). In Figure 2, we further explore the macro-level evidence by looking at the levels of key regulatory capital ratios. Using FDIC data, we were able to reconstruct three commonly used regulatory capital ratios: Leverage ratio (E/TA), Tier 1 Capital to Risk-Weighted Assets (T1/RWA) and Total Capital (i.e. Tier 1 plus Tier 2) to Risk-Weighted Assets (TOT/RWA). Since none of these three ratios are materially affected by loan-loss reserves or NPLs, all three exhibit relatively good performance during the crisis.6 Moreover, the Basel 2 risk-weighted ratios—T1/RWA and TOT/RWA—show pronounced increases following onset of the crisis. In contrast, the NPACR (which explicitly accounts for insufficient reserving against/write-offs of non-performing assets) drops from 10.34% in 2006 to 8.86% in 2008. It also remains about 300 to 400 basis points below the E/TA ratio during all subsequent years—signaling insufficient provisioning for/write-offs of nonperforming assets on/from the banks’ balance sheets. Consistent with our prior observations in Figure 1, the 1992 scenario is again similar to the most recent crisis: the NPACR ratio, being at only 6.44% and well below all other regulatory capital ratios, clearly detects weak capital position in the banking industry in this last S&L crisis year.

6

Loan-loss reserves and NPLs do enter into the calculation of risk-weighted assets, but not in a material way. 13

3.3. Bank-level evidence: Some recent examples of the regulatory forbearance In the previous subsection, we showed that NPACR delivers a promise to detect capital (in) adequacy in the presence of regulatory forbearance in the form of insufficient loan loss provisioning and write-offs. In this subsection, we further highlight the ability of the NPACR to detect insufficient capital adequacy and regulatory forbearance practices by providing some notable bank-level examples. Appendix Table 2 presents the 20 banks with the worst NPACRs as of year-end 2010. Third on the list is First State Bank of Stockbridge, GA, with NPACR equal to -26.0%. This bank also reported foreclosed real estate equal to 14.0% of assets, nonaccrual loans equal to 15.4% of assets and total non-performing assets equal to 35.0% of assets; yet it reported loanloss reserves equal to only 1.9% of assets and total equity equal to 2.7% of assets—just above the 2.0% PCA threshold for critical undercapitalization. Appendix Table 3 presents the 20 banks with the worst NPACRs as of year-end 2009. First State Bank of Stockbridge, GA shows up at number 20, with NPACR equal to -18.6%, nonperforming assets equal to 28.6% of assets, but loan-loss reserves of only 2.1% of assets and equity equal to 3.8% of assets. Going back another full year to year-end 2008, First State Bank of Stockbridge shows up with the 59th worst NPACR at -4.8%; it also reports nonperforming assets equal to 20.6% of assets, but loan-loss reserves equal to only 1.2% of assets and equity equal to a relatively healthy 10.1% of assets. First State Bank of Stockbridge, GA is not an outlier; there are scores, if not hundreds, of other troubled banks with similar histories. Douglas City Bank is second on the list in Appendix Table 2 with a NPACR equal to -27.4% and nonperforming assets equal to 36.0% of assets, but loan-loss reserves of only 0.9% of assets and equity equal to 4.7% of assets. As of year-end 14

2009, Douglas City bank had the 24th worst NPACR at -17.5% and nonperforming assets equal to 29.2% of assets, but loan-loss reserves of only 1.3% of assets and equity equal to 6.8% of assets. Going back another full year, Douglas had the 86th worst NPACR at -2.5% and nonperforming assets equal to 16.5% of assets, but loan-loss reserves of only1.5% of assets and equity equal to 9.2% of assets. Fourth on the list in Appendix Table 2 is Bank Commerce of Wood Dale, IL with an NPACR of -26.0% and nonperforming assets equal to 34.8% of assets. A year earlier, Bank Commerce had the 59th worst NPACR at -11.1% and nonperforming assets of 19.9%. Two years earlier, Bank Commerce had the 232nd worst NPACR at 1.7% and nonperforming assets equal to 8.8% of assets. These are just three cases picked from the worst five NPACRs shown in Appendix Table 2. If we were to start from year-end 2011 or 2012, we would find similar stories as we tracked back in time. Forbearance is alive and well among U.S. bank regulators. As of September 30, 2013, the FDIC reports that 515 banks remain on its list of “problem banks,” five full years after the financial crisis began in late 2008.7 Only 22 banks were closed by regulators during all of 2013.

7

See the FDIC Quarterly Banking Profile, available at: http://www2.fdic.gov/qbp/2013sep/qbp.pdf. 15

4. Data To construct various capital ratios for all U.S. commercial banks, we use the data from the Federal Financial Institutions Examination Council (FFIEC), which provides quarterly financial data for each FDIC-insured bank.8 In Table 1, we have detailed the construction and definitions of four commonly used regulatory capital ratios—the Leverage Ratio (E/TA), the Tangible Equity Ratio (TE/TAA), the Tier 1 Risk-based Capital Ratio (T1/RWA), and the Total Risk-Based Capital Ratio (TOT/RWA). We also describe our proposed ratio—the Nonperforming-Assets Coverage Ratio (NPACR), which we calculated as: NPACR = (E + LLR – NPA) / TA,

[2]

where: E is the book equity capital; LLR is the loan loss reserves; NPA is non-performing assets, calculated as the sum of 20% of loans past due 30 – 89 days, 50% of loans past due 90 – 180 days, and 100% of nonaccrual loans and OREO;9 TA is the year-end bank assets. Our sample period covers 2007 to 2012, including banks’ year-end capital ratios for the 2007 to 2010 period and the corresponding two-year window survival outcomes for the 2009 to

8

More specifically, we obtain our data from the website of the Federal Reserve Bank of Chicago, which provides quarterly FFIEC data from 1980 through 2010. See: http://www.chicagofed.org/webpages/banking/financial_institution_reports/commercial_bank_data.cfm. 9

These percentages correspond to the loan-loss reserves required for loans deemed to be substandard, doubtful, and loss during onsite examinations by U.S. regulators. We use past-due status a proxy for examination classifications. 16

2012 period. Bank failure data come from the FDIC’s official list of closed banks.10 As reported in Table 2, the total number of bank-year observations in our sample is 29,148, including:    

7,603 banks in 2007, 7,439 banks in 2008, 7,211 banks in 2009, and 6,895 banks 2010.

The corresponding failure rates over the two-year window are:    

150 banks (or 1.97%) of banks that were active in 2007 but failed during 2008 – 2009; 264 banks (or 3.55%) of banks that were active in 2008 but failed during 2009 – 2010; 225 banks (or 3.12%) of banks that were active in 2009 but failed during 2010 – 2011, and 128 banks (or 3.55%) of banks that were active in 2010 but failed during the 2011 – 2012.

[Insert Table 2 here.]

In Table 2, we also report the distribution of capital adequacy for our sample banks based on the FDICIA Prompt Corrective Action (PCA) guidelines for the FDIC insured US banks. Appendix Table 3 defines the five capital categories that we use to classify sample banks across PCA groups. The standardized PCA capital-adequacy definitions rely on the leverage ratio, the Tier 1 risk-based capital ratio, the total risk-based capital ratio, and the tangible-equity ratio. Based on these regulatory capital categories, a bank is classified as well-capitalized one if it meets all of the following:   

a leverage ratio of at least 5%, a Tier 1 risk-based capital ratio of at least 6%, and a Total risk-based capital ratio of at least 10%.

10

More specifically, we obtained our data from the website of the Federal Deposit Insurance Corporation. See: http://www.fdic.gov/bank/individual/failed/banklist.html. 17

At the other extreme, a bank is classified as “critically undercapitalized” if it has a ratio of tangible equity to total assets equal to, or less than, 2%. The frequency distribution of sample banks across PCA groups reveals three notable patterns. First, as expected, the combined ratio of banks in the lowest three PCA groups (Undercapitalized, Critically Undercapitalized, and Significantly Undercapitalized) increases rapidly as the financial crisis unfolds, from only 6 out of 7,603 banks (or 0.08%) in 2007 to 171 out of 6,895 banks (or 2.48%) in 2010. Second, especially at the beginning of the sample period, there appears to be a very weak relation between bank capital adequacy position, as defined by the PCA guidelines, and its survival outcomes. For example, 143 out of 150 banks that were active in 2007 and failed during the 2008-2009 period, were classified as well-capitalized based on the standard PCA group classification. As the crisis developed, however, we observe more consistency in the PCA group classification and bank failure outcomes: in 2010, only 17 out of 128 failed banks were classified as well-capitalized. Finally, in the survived banks subsample, we detect a non-negligible number of undercapitalized, critically undercapitalized, and significantly undercapitalized banks that managed to survive. For example, in 2010 as many as 76 banks (or in these inadequate capital categories have survived the follow-up two-year period. This stylized fact suggests increased regulatory forbearance at the later stage of the crisis.

18

5. Results 5.1. Descriptive Statistics for the Non-performing asset coverage ratio vs. other capital ratios In Table 3, we report descriptive statistics for five capital ratios, including NPACR, for all sample banks (Panel A) and by failed and surviving banks’ subsamples (Panels B and C, respectively). To mitigate the effects of extreme outliers, we winsorize each constructed capital ratio at the 1st and 99th percentiles of its distribution.

[Insert Table 3 here.]

In the full sample, the TOT/RWA ratio has the highest value, with the mean of 17.0% and the median of 14.4%, followed by the T1/RWA ratio with a mean value of 15.9% and the median of 13.2%. As expected, each capital ratio has a mean and median that is higher for surviving banks (Panel C) than for failing banks (Panel B). However, the NPACR ratio sends the distinctly strong and alarming early warning signal: NPACR is consistently negative for the mean (-1.5%) and the median (-0.4%) failed bank. None of the other four capital ratios signal the same degree of the extreme capital inadequacy, as their mean/median values all are greater than of 6%--well in excess of the 2% PCA level for critical undercapitalization. In Table 4, we perform a series of univariate comparison test for the mean and median capital ratios between failed and survived sample banks by breaking them down by the PCA capital adequacy groups. In the well-capitalized banks’ group by the PCA guidelines, NPACR performs slightly better than the T1/RWA and TOT/RWA ratios in discriminating failed and survived banks. For an average survived bank, the NPACR ratio is 10.5% versus only 4.4% 19

NPACR ratio for an average failed bank, resulting in a highly significant 6.1% difference. For the T1/RWA and TOT/RWA ratios this difference is 5.1% and 4.9%, correspondingly.

[Insert Table 4 here.]

For all other PCA capital adequacy groups, however, the NPACR is an unambiguous leader in distinguishing banks that will fail in the two-year window. For example, for the adequately capitalized banks, the NPACR is, on average 5.0% lower for failed banks. For comparison, all other four ratios exhibit only marginal, less than 0.5% average and median differences, between failed and survived banks. This result and the magnitude of the capital ratios’ differences between failed and survived banks hold for all other lower level PCA groups. Moreover, except well-capitalized banks’ subsample, the mean and median NPACR ratio is consistently negative for failed banks. It is also worth mentioning that in all three undercapitalized PCA bank groups (Undercapitalized, Significantly Undercapitalized, and Critically Undercapitalized), the mean and the median NPACR ratio for surviving banks is also negative. The latter result is consistent with the regulatory forbearance.

5.2. Which capital ratio is the best single predictor of bank failure? Results from logistic regression

In this section, we attempt to identify which capital ratio has the best ability to predict bank failures using a logistic regression framework. In the logistic regression models, our dependent variable is a binomial representation of bank survival outcome over a two-year 20

horizon, where bank failure is coded as a one and bank survival is coded as a zero. To test the early warning signals coming from alternative capital ratios, we run a series of separate univariate logistic regressions where we sequentially use each of the five capital ratios as the sole explanatory variable. We calculate bank capital ratios using year-end data for 2007, 2008, 2009, and 2010, respectively. We match these capital ratios with corresponding two-year survival outcomes for the 2008 – 2009, 20090 – 2010, 2010 – 2011, and 2011 – 2012, respectively. For this analysis, we pool data from the four overlapping periods. Results for the full sample appear in Panel A of Table 5. We supplement our main results in Panel A by splitting the full sample into two subsamples by PCA capital-adequacy criteria–well-capitalized banks (Panel B of Table 5) and less than well-capitalized banks (Panel C of Table 5). Each of the three panels contains five models, i.e., one model for each of the five capital ratios used in this study. To facilitate interpretation of the coefficients, we report all estimates as the marginal effects. We assess the predictive power of each capital ratio using Pseudo R-square values, which are essentially identical to the Chi-square in a univariate logistic regression setting.

[Insert Table 5 here.]

Overall, logistic regression results for the full sample shown in Panel A provide evidence of superior predictive power for the NPACR as compared to the other four commonly used ratios. The value of the pseudo-R-Square for the NPACR is 0.348, as compared with values of 0.323 for T1/RWA, 0.315 for TOT/RWA, 0.245 for TE/TAA, and only 0.232 for E/TA. The marginal effects are sizeable and statistically significant at 1% level across all five capital ratios, 21

but are largest for the models with the least predictive power: -0.384 for E/TA; -0.375 for TE/TAA; -0.239 for NPACR; -0.091 for TOT/RWA and -0.086 for T1/RWA. In Panel B of Table 5 are the logistic regression results for the subsample of banks that were classified as “well-capitalized” in each year according to the PCA guidelines. These are the banks of most interest to regulators from an early warning perspective, as they had not already been flagged as troubled by the PCA guidelines. The value of pseudo-R-square for NPACR is 0.138, as compared with values of 0.095 for T1/RWA; 0.086 for TOT/RWA; 0.033 for TE/TAA; and 0.026 for E/TA. Again, the marginal effects are largest for the models with the least explanatory power: -0.221 for E/TA; -0.257 for TE/TAA; -0.177 for NPACR; -0.141 for TOT/RWA and -0.132 for T1/RWA. In Panel C of Table 5 are the logistic regression results for the subsample of banks that were classified as less than “well-capitalized” in each year according to the PCA guidelines. These banks are of less interest to regulators from an early warning perspective, as they already had been flagged as troubled by the PCA guidelines. The value of pseudo-R-square for NPACR is 0.222, as compared with values of 0.200 for T1/RWA; 0.192 for TOT/RWA; 0.151 for TE/TAA; and 0.160 for E/TA. One severe identification limitation of our logistic regression results, however, is that they do not show evidence of the regulatory forbearance that we detect in the descriptive part of our empirical analysis. In other words, regulators’ reluctance to close banks with severely insufficient loss provisioning and write-offs, and, therefore, low NPACR indicator, should impair the ability of the NPACR to identify closed banks.

22

5.3. Which capital ratio is the best single predictor of bank failure? Receiver-Operating-Characteristics (ROC) curves

Although we cannot bypass the regulatory forbearance limitation in our analysis, we take an alternative look at the evaluation and interpretation of our empirical results and construct a series of the Receiver Operating Characteristics (ROC) curves to assess predictive ability of alternative capital ratios in subsample of well-capitalized banks by the PCA guidelines definition. ROC curve is a graph that is commonly used in classification problems and represents a plot of Sensitivity versus (1 – Specificity) for different cut-offs. For a model with no predictive power, the ROC curve takes the shape of a 45-degree line; for a model with a good predictive power, the ROC curve has concave shape and lies above the diagonal. In the context of our classification task, the sensitivity is the probability of correctly predicting troubled banks’ failure, which should be the primary bank regulators’ concern. The complement of this probability is the false negative rate. Another way to interpret ROC curve results is from the prism of Type I ("false alarm" or “false positive”) versus Type II ("missed detection" or “false negative”) error rates. Because the cost of a Type II error is orders of magnitude than the cost of a Type I error, it is important to focus on the left portion of the ROC curve, typically at 1%, 5% and 10%. The primary rationale for this stage of analysis is that from the regulatory point of view, the consequences of misclassifying a bank that subsequently fails are pronouncedly more adverse than the consequences of misclassifying a bank that does not fail. We are also specifically

23

interested in subsample of well-capitalized banks as the most desirable feature of the early warning indicator is to predict failures in this seemingly safe group of banks. Figure 3 presents the ROC curves for the bank failure models estimated using only wellcapitalized banks, and using each of the five capital ratios as a single explanatory variable. Panels A – D present results based upon capital adequacy data from year-ends 2007, 2008, 2009 and 2010, respectively; and corresponding two-year survival outcomes for the 2008 – 2009, 20090 – 2010, 2010 – 2011, and 2011 – 2012, respectively. Consistent with our task at hand, we focus our analysis on the left-hand tail of the ROC graphs (denoted as a shaded area up to 25% level in all Panels of Figure 3).

[Insert Figure 3 here.]

The ROC curves in Panels B, C, and D of Figure 3 show that the NPACR dominates all of the other capital ratios from 2008, 2009, and 2010 across all values of Type 1 errors. For example, based on the 2008 year-end capital ratios, Panel B indicates that the sensitivity for the NPACR is about 55% when (1 – Specificity) is 5%. For the second-best predictor for 2008— T1/RWA ratio—the sensitivity is less than 40% at the 5% for the (1 – Specificity). In 2009 for the same (1-Specificity) 5% level, NPACR sensitivity is as high as 60% while the sensitivity of the second-best predictors, the two risk-weighted capital ratios, does not exceed 40%. For 2007 (Panel A in Figure 3), where capital ratios are based upon pre-crisis levels, the NPACR is more accurate in the critical far-left portion of the ROC curve, but the two RWA capital ratios are more accurate for values of (1 – Specificity) of greater than 0.20. This is most

24

likely due to the risk weights capturing the importance of commercial real-estate loans, especially construction & development loans, in explaining subsequent failures.11 Collectively, the evidence in Figure 3 reveals that, in terms of correct classification of failed banks, NPACR outperforms the other four capital ratios during crisis years when capitaladequacy forbearance was available from regulators in the form of insufficient write-offs and provisioning. The limited pre-crisis evidence suggests that the NPACR is at least as accurate as the risk-based capital ratios.

6. Summary and conclusions In this study, we test the predictive power of several alternative measures of bank capital adequacy in identifying U.S. bank failures during the recent crisis period. We find that an unconventional ratio—the non-performing asset coverage ratio—outperforms Basel-based ratios including the Tier 1 ratio, the Total Capital Ratio, and the Leverage ratio—throughout the crisis period in identifying bank failures. It also outperforms in predicting failures among “wellcapitalized” banks as defined by the current Prompt Corrective Action guidelines. From an early warning perspective, these banks are of most concern to regulators because they have not been identified as troubled by the PCA guidelines. Based on our results, we argue that NPACR outperforms other ratios in at least five aspects: (i) it aligns capital and credit risks—the two primary risks of bank failures—in one measure; (ii) it is easier to calculate than the Tier 1 and Total Capital ratios, as it requires 11

Cole and White (2012) find that construction & development loans and other commercial real estate loans are as important as the equity-to-asset ratio in explaining bank failures during 2009 based upon year-end 2007 financial data. Because these two types of loans carry 100% risk weights, these banks would have relatively low values of T1/RWA and TOT/RWA. The values of the other capital ratios would not be affected by these investments. 25

calculation of no complex risk weights; (iii) it allows one to account for various time period and cross-country provisioning rules and regimes, including episodes of regulatory forbearance and cross-country differences; (iv) it removes the incentives of both banks and regulators to mask capital deficiencies by creating/requiring insufficient loan-loss reserves; and (v) it outperforms all other commonly used capital ratios in predicting bank failures. We believe that all of the above features of the NPACR argue in favor of its integration into the prompt corrective action guidelines. We also expect that this single and informative measure of bank risk can be efficiently used in empirical banking studies across countries and across time. The results of this study also shed new light on capital-adequacy forbearance offered by U.S. bank regulators during the recent financial crisis. Our study makes three important contributions to the literature on financial institutions. First, we contribute to the literature on bank capital adequacy (See, e.g., Pettway, 1976; Sharpe, 1978; Buser, Chen, and Kane, 1981; Kashyap, Rajan, and Stein, 2008; Allen, Fulghieri, and Mehran, 2011; Admati et al., 2013; Rosengren, 2013). We demonstrate that the simple and intuitive NPACR outperforms the complex Basel regulatory capital ratios in forecasting bank failures in the U.S. Because the NPACR is simple to calculate across time and across countries, it holds great promise for researchers looking to analyze capital adequacy, but limited by the availability of supervisory data needed to calculate Basel regulatory capital ratios. Second, we contribute to the literature on regulatory forbearance and prompt corrective action (see, e.g., Dahl and Spivey, 1995; Jones and Kuester-King, 1995; Aggarwal and Jacques, 2001). We provide convincing new evidence that U.S. bank regulators engaged in a massive scheme of forbearance during 2008 – 2013 that subverted the prompt corrective action provisions of the FDIC Improvement Act of 1991. Our evidence points to the need for new laws 26

and/or regulations designed to limit the discretion of regulators in enforcing laws duly passed by the U.S. Congress. Third, we contribute to the literature on bank failures (see, e.g., Sinkey, 1975; Bovenzi, Marino, and McFadden, 1982; Lane, Looney, and Wansley, 1986; Thomson, 1992; Cole and Gunther, 1995, 1998; Estrella, Park, and Peristiani, 2000; Cole and White, 2012; Berger and Bouwman, 2013). We offer a new measure of capital adequacy, and then demonstrate that it is superior to Basel regulatory capital ratios in predicting bank failures. Because of its simplicity, the NPACR holds great promise for researchers seeking to analyze bank failures across time and/or across countries.

27

References Acharya, V., Engle, R., and Richardson, M. (2012). Capital shortfall: A new approach to ranking and regulating systemic risks. American Economic Review 102, 59-64. Admati, A.R, DeMarzo, P.M., Hellwig, M.F., and Pfleiderer, P. (2013). Fallacies, irrelevant facts, and myths in the discussion of capital regulation: Why bank equity is not socially expensive. Rock Center for Corporate Governance Working Paper Series No. 161 Aggarwal, R, and Jacques, K. (2001). The impact of FDICIA and prompt corrective action on bank capital and risk: Estimates using a simultaneous equations model. Journal of Banking & Finance 25, 1139-1160. Allen, F., Fulghieri, P., and Mehran, H. (2011). The value of bank capital and the structure of the banking industry. Review of Financial Studies 24, 971-982. Aubuchon, C.P., and Wheelock, D.C. (2010). The geographic distribution and characteristics of U.S. bank failures, 2007-2010: Do bank failures still reflect local economic conditions? Federal Reserve Bank of St. Louis Review (Sep.-Oct.) 395-416. Baer, H. L., and McElravey, J. (1992). Capital adequacy and the growth of U.S. banks. Federal Reserve Bank of Chicago No. 92-11 Berger, A. N., and Bouwman, C. H. (2013). How does capital affect bank performance during financial crises? Journal of Financial Economics 109, 146-176. Bovenzi, J.F., Marino, J.A., and McFadden, F.E. 1983. Commercial bank failure prediction models. Economic Review, Federal Reserve Bank of Atlanta (November), 14-26. Buser, S.A., Chen, A.H., and Kane, E.J. (1981). Federal deposit insurance, regulatory policy, and optimal bank capital. The Journal of Finance 35, 51-60. Cebulla, R.J., Koch, J.V., and Fenili, R.N. (2011). The bank failure rate, economic conditions,and banking statutes in the U.S., 1970-2009. Atlantic Economic Journal 39, 39-46. Cihák, M., Demirgüç-Kunt, A., Martinez Peria, M. S., and Mohseni-Cheraghlou, A. (2012). Bank regulation and supervision around the world: A crisis update. World Bank Policy Research Working Paper No. 6286. Cole, R.A. and D. Dahl. (2013). Bank audits and failures: Which causes which? SSRN Cole, R.A. and Gunther, J.W. (1995). Separating the likelihood and timing of bank failure. Journal of Banking & Finance 19, 1073-1089. 28

Cole, R.A. and Gunther, J.W. (1998). Predicting bank failures: A comparison of on- and off-site monitoring systems. Journal of Financial Services Research 13, 103-117. Cole, R. A., and White, L. J. (2012). Déjà Vu all over again: The causes of US commercial bank failures this time around. Journal of Financial Services Research 42, 5-29. DeYoung, R., and Torna, G. (2013). Nontraditional activities and bank failures during the financial crisis. Journal of Financial Intermediation 22, 397-421. Estrella, A, Park. S., and Peristiani, S. (2000). Capital ratios as predictors of bank failures. Economic Review. Federal Reserve Bank of New York (July), 33-52. Flannery, M. J., and Rangan, K. P. (2008). What caused the bank capital build-up of the 1990s? Review of Finance 12, 391-429. Gajewski, G.R. (1989). Assessing the risk of bank failure. Proceedings of a Conference on Bank Structure and Competition, Federal Reserve Bank of Chicago, 432-456. Goodhart, C. (2013). Ratio controls need reconsideration. Journal of Financial Stability 9, 444450. Haldane, A.G. (2011). Capital discipline. Speech to the American Economics Association. Denver, CO. Jan. 9, 2011. Haldane, A.G. (2012). The dog and the frisbee. Speech at the Federal Reserve Bank of Kansas City’s 366th economic policy symposium “The Changing Policy Landscape.” Jackson Hole, WY, August 31. Huizinga, H., and Laeven, L. (2012). Bank valuation and accounting discretion during a financial crisis. Journal of Financial Economics 106, 614-634. Jarrow, R. (2013). A leverage ratio rule for capital adequacy. Journal of Banking & Finance 39, 973-976 Jin, J., Kanagaretnam, K., and Lobo, G. 2011. Ability of accounting and audit quality variables to predict bank failure during the financial crisis. Journal of Banking & Finance 35, 2811-2819. Jones, D.S. and Kuester-King, K. (1995). The implementation of Prompt Corrective Action: An assessment. Journal of Banking & Finance 19, 491-510. Kashyap, A., Rajan, R., and Stein, J. (2008). Rethinking capital regulation. In Federal Reserve Bank of Kansas City Symposium at Jackson Hole. Keeley, M.C., and F.T. Furlong. (1990). A re-examination of mean-variance analysis of bank capital regulation. Journal of Banking & Finance 14, 69-84. 29

Kerstein, J., and Kozberg, A. (2013) Using accounting proxies of proprietary FDIC ratings to predict bank failures and enforcement actions during the recent financial crisis. Journal of Accounting, Auditing and Finance 18, 128-151 Kocherlakota, N.R. and I. Shim. (2007). Forbearance and prompt corrective action. Journal of Money, Credit & Banking 39, 1107-1129. Ng, J., and Roychowdhury, S. (2012). Do loan-loss reserves behave like capital? Evidence from recent bank failures. SSRN Working Paper No. 1646928. Pettway, R.H. (1976). Market tests of capital adequacy of large banks. The Journal of Finance 31, 865-875. Rosengren, E. S. (2013). Bank capital: lessons from the US financial crisis. Federal Reserve Bank of Boston Speech, (Feb 25). Sharpe, W.F. (1978). Bank capital adequacy, deposit insurance, and security values. Journal of Financial & Quantitative Analysis 13, 701-718. Shaffer, S. (2012). Bank failure risk: Different now? Economic Letters 116, 613-616. Sinkey, J. F. (1975). A multivariate statistical analysis of the characteristics of problem banks, The Journal of Finance 30, 21-36. Thomson, J.B. (1992). Modeling the regulator's closure option: A two-step logit regression approach. Journal of Financial Services Research 6, 5-23. Vyas, D. (2011). The timeliness of accounting write-downs by U.S. financial institutions during the Financial Crisis of 2007-2008. Journal of Accounting Research 49, 823-860.

30

Appendix Table 2 Banks with worst Nonperforming Asset Coverage Ratio (NPACR) As of Year-End 2010 FDIC Cert. is the FDIC certificate number that uniquely identifies each FDIC-insured bank. E_TA is the ratio of total equity to total assets. LLR is the ratio of loan-loss reserves to total assets. REO is the ratio of foreclosed real estate to total assets, NonAccr is the ratio of nonaccrual loans to total assets, PD90 is the ratio of loans past due 90 or more days and still accruing interest to total assets. PD30 is the ratio of loans past due 30 – 89 days and still accruing interest to total assets. NPA is the sum of REO, NonAccr, PD90 and PD30. NPACR is equal to the sum of E_TA and LLR, less the sum of 20% of PD30, 50% of PD90, 100% of NonAccr and 100% of REO. FDIC Obs Cert. Bank Name City ST NPACR NPA E_TA LLR REO NonAccr PD90 PD30 1 34613 BUILDERS BK CHICAGO IL -0.293 0.397 0.075 0.013 0.178 0.200 0.000 0.019 2 21649 DOUGLAS CTY BK DOUGLASVILLE GA -0.274 0.360 0.047 0.009 0.178 0.144 0.000 0.038 3 19252 FIRST ST BK STOCKBRIDGE GA -0.260 0.350 0.027 0.019 0.140 0.154 0.002 0.053 4 34292 BANK COMMERCE WOOD DALE IL -0.257 0.348 0.004 0.049 0.077 0.224 0.000 0.047 5 35299 SECURITY EXCH BK MARIETTA GA -0.255 0.333 0.036 0.017 0.254 0.048 0.000 0.031 6 19554 HIGH TRUST BK STOCKBRIDGE GA -0.248 0.372 0.033 0.025 0.119 0.168 0.009 0.076 7 19758 ENTERPRISE BKG CO MCDONOUGH GA -0.248 0.305 0.013 0.015 0.056 0.212 0.000 0.038 8 35242 NORTH GA BK WATKINSVILLE GA -0.245 0.287 0.004 0.007 0.080 0.161 0.017 0.029 9 57440 OGLETHORPE BK BRUNSWICK GA -0.244 0.289 -0.027 0.021 0.072 0.142 0.033 0.042 10 58273 PATRIOT BK OF GA CUMMING GA -0.241 0.318 0.023 0.018 0.133 0.137 0.009 0.039 11 58539 FIRST CHOICE CMNTY BK DALLAS GA -0.224 0.274 -0.025 0.051 0.010 0.234 0.004 0.026 12 57646 FIRSTIER BK LOUISVILLE CO -0.220 0.274 -0.003 0.046 0.081 0.179 0.000 0.014 13 57432 AMERICAN TR BK ROSWELL GA -0.212 0.261 0.006 0.013 0.084 0.133 0.018 0.025 14 151 HABERSHAM BK CLARKESVILLE GA -0.212 0.261 0.007 0.017 0.105 0.123 0.004 0.029 15 4744 FIRST NB OF OLATHE OLATHE KS -0.208 0.299 0.029 0.041 0.126 0.146 0.002 0.024 16 57256 PIEDMONT CMNTY BK GRAY GA -0.203 0.288 0.043 0.024 0.135 0.130 0.000 0.023 17 25155 FIRST NB OF FL MILTON FL -0.195 0.295 0.067 0.030 0.099 0.194 0.000 0.003 18 19498 MONTGOMERY B&TC AILEY GA -0.192 0.304 0.073 0.014 0.114 0.159 0.000 0.032 19 19237 MCINTOSH ST BK JACKSON GA -0.189 0.246 0.022 0.028 0.075 0.160 0.005 0.007 20 34965 FIRST CMRL BK OF FL ORLANDO FL -0.188 0.236 -0.023 0.063 0.026 0.200 0.000 0.009

31

Appendix Table 3 Banks with worst Nonperforming Asset Coverage Ratio (NPACR) As of Year-End 2009 FDIC Cert. is the FDIC certificate number that uniquely identifies each FDIC-insured bank. E_TA is the ratio of total equity to total assets. LLR is the ratio of loan-loss reserves to total assets. REO is the ratio of foreclosed real estate to total assets, NonAccr is the ratio of nonaccrual loans to total assets, PD90 is the ratio of loans past due 90 or more days and still accruing interest to total assets. PD30 is the ratio of loans past due 30 – 89 days and still accruing interest to total assets. NPA is the sum of REO, NonAccr, PD90 and PD30. NPACR is equal to the sum of E_TA and LLR, less the sum of 20% of PD30, 50% of PD90, 100% of NonAccr and 100% of REO. FDIC Obs Cert. Bank Name City ST NPACR NPA TE_TA LLR REO NonAccr PD90 PD30 1 5672 FLORIDA CMNTY BK IMMOKALEE FL -0.310 0.398 0.002 0.065 0.104 0.268 0.001 0.025 2 58429 WHEATLAND BK NAPERVILLE IL -0.305 0.393 -0.014 0.072 0.003 0.352 0.000 0.037 3 21521 CITY BK LYNNWOOD WA -0.281 0.341 0.032 0.010 0.151 0.167 0.000 0.023 4 57147 PREMIER AMER BK MIAMI FL -0.280 0.388 0.000 0.045 0.018 0.292 0.000 0.077 5 57448 FIRST CMRC CMNTY BK DOUGLASVILLE GA -0.280 0.384 0.059 0.016 0.146 0.202 0.000 0.036 6 29952 GEORGE WA SVG BK ORLAND PARK IL -0.277 0.450 0.033 0.052 0.045 0.296 0.000 0.110 7 57399 MCINTOSH CMRL BK CARROLLTON GA -0.268 0.363 0.023 0.013 0.132 0.143 0.039 0.050 8 35279 HIGH DESERT ST BK ALBUQUERQUE NM -0.259 0.346 0.035 0.041 0.112 0.220 0.000 0.014 9 58072 USA BK PORTCHESTER NY -0.245 0.297 0.030 0.018 0.009 0.282 0.000 0.006 10 59021 INDEPENDENT BKR BK SPRINGFIELD IL -0.234 0.000 -0.234 0.000 0.000 0.000 0.000 0.000 11 58104 CENTURY SCTY BK DULUTH GA -0.219 0.304 0.024 0.011 0.174 0.061 0.012 0.057 12 33904 GORDON BK GORDON GA -0.210 0.332 0.064 0.029 0.112 0.184 0.000 0.036 13 1252 BARNES BKG CO KAYSVILLE UT -0.197 0.299 0.009 0.081 0.083 0.201 0.000 0.015 14 57697 TOWNE BK OF AZ MESA AZ -0.194 0.313 0.056 0.034 0.195 0.081 0.000 0.037 15 5702 COMMUNITY B&TC CORNELIA GA -0.193 0.296 0.016 0.030 0.036 0.188 0.000 0.072 16 33989 APPALACHIAN CMNTY BK ELLIJAY GA -0.192 0.269 0.018 0.023 0.129 0.094 0.004 0.043 17 34658 CITIZENS B&TC CHICAGO IL -0.190 0.340 0.031 0.024 0.033 0.160 0.074 0.073 18 34613 BUILDERS BK CHICAGO IL -0.189 0.265 0.062 0.014 0.106 0.160 0.000 0.000 19 35299 SECURITY EXCH BK MARIETTA GA -0.187 0.312 0.054 0.025 0.124 0.120 0.028 0.039 20 19252 FIRST ST BK STOCKBRIDGE GA -0.186 0.286 0.038 0.021 0.077 0.157 0.001 0.051

32

Appendix Table 1: Definitions of capital adequacy Under the U.S. Prompt Corrective Action (PCA) framework.

PCA capital categories: Well capitalized Adequately capitalized Undercapitalized Significantly undercapitalized Critically undercapitalized

Total risk-based capital ratio

Tier1 risk-based capital ratio

≥ 10% ≥ 8% < 8%

and and or

and and or

≥ 5% ≥ 4%a < 4%a

< 6%

or < 3% or Tangible Equity ratio < 2%

< 3%

a

≥ 6% ≥ 4% < 4%

Tier1 leverage ratio

For banks with CAMEL rating equal to one and no expectations for subsequent growth, the corresponding thresholds for the minimal tier 1 leverage ratio are set at 3% instead of 4%.

33

Figure 1: Nonperforming assets, loan loss reserves and coverage ratios in the U.S. banking system: Aggregate-level evidence, all insured depository institutions 1992 – 2012 This industry-level graph shows that the pronounced increase in the nonperforming assets in the U.S. banking industry during the recent financial crisis comes in hand with increased but insufficient loan loss provisioning. In response to these trends, the coverage ratio, defined as the ratio of nonperforming assets to loan loss provisions (right-hand scale), gradually drops from 154.2 % in 2004 to only 58.8% in 2009. If loan loss provisioning is inadequate, the traditional capital ratios, considered in isolation with provisioning regime, will tend to underestimate bank insolvency risk and to overestimate bank capacity to absorb losses. The shaded area on this graph denotes the sample period. The nonperforming assets are calculated as the sum of 20% of loans past due 30-89 days, 50% of loans past due 90-180 days, and 100% of nonaccrual loans and OREO. Data source: FFIEC Reports of Condition and Income via Federal Reserve Bank of Chicago website. 400,000,000

160%

350,000,000

140%

300,000,000

120%

250,000,000

100%

200,000,000

80%

150,000,000

60%

100,000,000

40%

50,000,000

20%

NPA ($000) LLR ($000) Coverage ratio (LLR/NPA, %)

34

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

0%

1992

0

Figure 2: Key capital ratios in the U.S. banking system: Aggregate-level evidence, all insured depository institutions 1992 – 2012 This industry-level graph shows a two-decade pattern for the four key capital ratios used in this study. The definitions of capital ratios are provided in Table 1. The shaded area on this graph denotes the sample period. Data source: FFIEC Reports of Condition and Income via Federal Reserve Bank of Chicago website. 16% 14% 12% 10% 8% 6%

E/TA T1/RWA TOT/RWA NPACR

35

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

1992

4%

Figure 3: Receiver Operating Characteristic Curves Two-year window survival outcomes for well-capitalized banks (PCA group = 1) By sample years Panel A. 2007 (2008-2009 outcomes: 143 failures and 7,383 survivals)

Panel B. 2008 (2009-2010 outcomes: 174 failures and 7,081 survivals)

36

Figure 3: (cont.) Receiver Operating Characteristic Curves Two-year window survival outcomes for well-capitalized banks (PCA group = 1) by sample years. Panel C. 2009 (2010-2011 outcomes: 52 failures and 6,821 survivals)

Panel D. 2010(2011-2012 outcomes: 17 failures and 6,570 survivals)

37

Table 1: Definition of bank capital ratios used in this study Capital ratio

Notation

Regulatory capital ratios: Leverage ratio E/TA Tangible equity TE/TAA ratio Tier 1 risk-based T1/RWA capital ratio Total risk-based TOT/RWA capital ratio Proposed ratio: Nonperforming assets coverage ratio

NPACR

Definition

Book equity capital divided by year-end assets Tangible equity a divided by average total assets The Tier 1b capital divided by risk-weighted d assets. The sum of Tier 1 and Tier 2 c capital divided by riskweighted assets.

Book equity capital plus loan loss reserves less nonperforming assets, all divided by year-end assets: NPACR = (E + LLR – NPA) / TA, where the nonperforming assets are calculated as the sum of 20% of loans past due 30-89 days, 50% of loans past due 90-180 days, and 100% of nonaccrual loans and OREO

a

Tangible equity capital is equal to bank equity less goodwill, other intangible assets and the value of the bank's perpetual preferred stock and related surplus. b

Tier 1 capital is the sum of core capital components, including capital stock, surplus, undivided profits, qualifying non-cumulative perpetual preferred stock and minority interest, excluding goodwill and other intangible assets. c

Tier 2 capital is the sum of the subordinated debt, perpetual and intermediate term preferred stock not qualifying as Tier 1 capital and loan loss allowance limited to 1.25 % of RWA. d

Risk-weighted assets are calculated by assigning one of the four risk categories (0%, 20%, 50% or 100%) for bank assets and off-balance sheet items.

38

Table 2: Descriptive statistics Distribution of sample banks by year, PCA capital adequacy group and the twoyear survival outcomes. This table reports numbers of sample banks across three dimensions: sample years, PCA capital adequacy group (as defined in Appendix 1) and survival outcomes in the next two years after the bank capital level observation. The total number of bank-year observations is 29,148, including 767 failures and 28,381 survivals. The sample includes all insured depository institutions. The year-end bank capital positions are for the period 2007-2010 period; the corresponding survival outcomes are for the 2009-2012 and cover a two-year horizon for each study sample observations. Number of banks 2007 2008 2009 2010 All years All banks including:

7,603

7,439

7,211

6,895

29,148

Well-Capitalized Adequately Capitalized Undercapitalized Critically Undercapitalized Significantly Undercapitalized

7,526 71 3 3

7,255 131 29 15

6,873 167 83 52

6,587 137 75 63

28,241 506 190 133

0

9

36

33

78

Failed over a two-year window including: Well-Capitalized Adequately Capitalized Undercapitalized Critically Undercapitalized Significantly Undercapitalized

150

264

225

128

767

143 4 1 2

174 51 20 12

52 43 57 41

17 16 26 39

386 114 104 94

0

7

32

30

69

Survived over a two-year horizon including: Well-Capitalized Adequately Capitalized Undercapitalized Critically Undercapitalized Significantly Undercapitalized

7453

7,175

6,986

6,767

28,381

7,383 67 2 1

7,081 80 9 3

6,821 124 26 11

6,570 121 49 24

27,855 392 86 39

0

2

4

3

9

39

Table 3: Descriptive statistics: U.S. banks capital ratios (2007-2010, pooled data) All constructed ratios are winsorized at the 1st and the 99th percentiles. The definition and the notation for each variable are detailed in Table 1. The survival outcomes are over two-year horizons. Mean

Median

SD

Min

Max

Panel A. All banks (N obs. = 29,148) E/TA TE/TAA T1/RWA TOT/RWA NPACR

0.111 0.107 0.159 0.170 0.104

0.099 0.095 0.132 0.144 0.096

0.051 0.052 0.091 0.091 0.065

-0.053 -0.050 -0.165 -0.165 -0.310

0.957 0.964 0.997 0.997 0.970

Panel B. Failed banks (N of obs. = 767) E/TA TE/TAA T1/RWA TOT/RWA NPACR

0.067 0.064 0.079 0.092 -0.015

0.068 0.067 0.084 0.100 -0.004

0.040 0.035 0.045 0.046 0.096

-0.053 -0.050 -0.165 -0.165 -0.310

0.397 0.338 0.354 0.384 0.470

Panel C. Survived banks (N of obs. = 28,381) E/TA TE/TAA T1/RWA TOT/RWA NPACR

0.112 0.107 0.160 0.171 0.103

0.100 0.095 0.133 0.145 0.095

0.049 0.050 0.089 0.088 0.061

0.000 0.000 0.000 0.000 -0.296

0.952 0.964 0.981 0.992 0.970

40

Table 4: Univariate comparison tests: Failed and survived banks' capital ratios by PCA capital adequacy groups This table presents comparison tests for the differences in mean and median capital ratios between failed and survived US banks across capital adequacy groups for 29,148 bank-years over the 2007 – 2010 period. The survival outcomes are over the subsequent two-year window for each observation. All bank capital ratios used in this study are defined in Table 1. Appendix 1 outlines the PCA bank capitalization groups’ criteria and the distribution of sample banks across the PCA groups is reported in Table 2. Difference tests are t-test for means and Wilcoxon rank-sum tests for medians. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. Well-Capitalized Failed Survived

Diff.

Adequately Capitalized Failed Survived

Diff.

Undercapitalized Failed Survived

Diff.

Critically Undercapitalized Failed Survived

Diff.

Signif. Undercapitalized Failed Survived

Diff.

E/TA Mean

0.094

0.113

-0.019*** 0.064

0.068

-0.004**

0.045

0.049

-0.004**

0.031

0.032

-0.001

0.006

0.023

-0.017**

Median

0.086

0.100

-0.014*** 0.063

0.066

-0.003*

0.044

0.047

-0.003**

0.029

0.029

-0.001

0.010

0.013

-0.003

Mean

0.089

0.108

-0.019*** 0.063

0.064

-0.001

0.044

0.047

-0.003**

0.030

0.030

0.000

0.006

0.011

-0.005

Median

0.083

0.095

-0.012*** 0.062

0.065

-0.003

0.043

0.046

-0.002*

0.028

0.028

0.000

0.009

0.012

-0.004

Mean

0.110

0.161

-0.051*** 0.076

0.080

-0.004*** 0.056

0.057

-0.001

0.035

0.039

-0.004**

0.006

0.034

-0.028**

Median

0.099

0.134

-0.035*** 0.076

0.081

-0.005*** 0.056

0.058

-0.002

0.036

0.040

-0.004**

0.012

0.019

-0.006*

Mean

0.123

0.173

-0.049*** 0.089

0.092

-0.003*** 0.069

0.071

-0.001

0.048

0.052

-0.004**

0.014

0.045

-0.031**

Median

0.111

0.146

-0.034*** 0.088

0.093

-0.004*** 0.069

0.071

-0.002

0.049

0.054

-0.004**

0.025

0.036

-0.011*

Mean

0.044

0.105

-0.061*** -0.030

0.021

-0.050*** -0.067

-0.028

-0.039*** -0.108

-0.053

-0.056*** -0.122

-0.048

-0.074***

Median

0.055

0.097

-0.043*** -0.020

0.033

-0.053*** -0.062

-0.024

-0.037*** -0.105

-0.038

-0.067*** -0.114

-0.079

-0.034**

TE/TAA

T1/RWA

TOT/RWA

NPACR

41

Table 5: Capital ratios and bank survival: Univariate logistic regressions This table reports estimated marginal effects of the bank capital ratios on the bank survival outcomes for a full sample (Panel A, N = 29,148 bank-year observations) and for subsamples of well-capitalized (Panel B, N = 28,241 bank-years, PCA group = 1) and less than well-capitalized banks (Panel C, N = 907 bank-years, PCA group = 2 to 5). Bank capital ratios are for the 2007 – 2010 period while the survival outcomes are for the subsequent two-year window, i.e. 2009 – 2012. Robust standard errors are shown in parenthesis below each estimate. The dependent variable is (1 = bank failure over the two-year window, 0 = bank survival over the two-year window). All alternative capital ratios are defined in Table 1. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. Panel A: Full sample E/TA

Model 1 -0.384*** (0.020)

TE/TAA

Model 2

Model 3

Model 4

-0.375*** (0.020)

T1/RWA

-0.086*** (0.008)

TOT/RWA

-0.091*** (0.008)

NPACR Intercept

N of observations N of failed banks N of surviving banks Pseudo R-Square ROC Area Under Curve

Model 5

0.011*** (0.001)

0.012*** (0.001)

0.004*** (0.000)

0.006*** (0.000)

-0.239*** (0.011) -0.016*** (0.001)

29,148 767 28,381 0.232

0.245

0.323

0.315

0.348

0.821

0.825

0.893

0.887

0.888

42

Table 5: (cont’d) This table reports estimated marginal effects of the bank capital ratios on the bank survival outcomes for a full sample (Panel A, N = 29,148 bank-year observations) and for subsamples of well-capitalized (Panel B, N = 28,241 bank-years, PCA group = 1) and less than well-capitalized banks (Panel C, N = 907 bank-years, PCA group = 2 to 5). Bank capital ratios are for the 2007 – 2010 period while the survival outcomes are for the subsequent two-year window, i.e. 2009 – 2012. Robust standard errors are shown in parenthesis below each estimate. The dependent variable is (1 = bank failure over the two-year window, 0 = bank survival over the two-year window). All alternative capital ratios are defined in Table 1. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. Panel B: Well-capitalized banks (PCA group 1) E/TA

Model 1 -0.221*** (0.020)

TE/TAA

Model 2

Model 3

Model 4

-0.257*** (0.019)

T1/RWA

-0.132*** (0.010)

TOT/RWA

-0.141*** (0.010)

NPACR Intercept N of observations N of failed banks N of survived banks Pseudo R-square ROC Area Under Curve

Model 5

-0.024*** (0.003) 28,241 386 27,885 0.026

-0.017*** (0.003)

0.033

0.095

0.086

0.138

0.674

0.681

0.805

0.795

0.799

43

0.001 (0.001)

0.002 (0.001)

-0.177*** (0.009) -0.018*** (0.001)

Table 5: (cont’d) This table reports estimated marginal effects of the bank capital ratios on the bank survival outcomes for a full sample (Panel A, N = 29,148 bank-year observations) and for subsamples of well-capitalized (Panel B, N = 28,241 bank-years, PCA group = 1) and less than well-capitalized banks (Panel C, N = 907 bank-years, PCA group = 2 to 5). Bank capital ratios are for the 2007 – 2010 period while the survival outcomes are for the subsequent two-year window, i.e. 2009 – 2012. Robust standard errors are shown in parenthesis below each estimate. The dependent variable is (1 = bank failure over the two-year window, 0 = bank survival over the two-year window). All alternative capital ratios are defined in Table 1. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. Panel C: Less than well-capitalized (PCA groups 2 to 5) E/TA

Model 1 -11.717*** (0.992)

TE/TAA

Model 2

Model 3

Model 4

-12.216*** (1.037)

T1/RWA

-13.272*** (1.052)

TOT/RWA

-13.135*** (1.077)

NPACR Intercept N of observations N of failed banks N of survived banks Pseudo R-square ROC Area Under Curve

Model 5

0.520*** (0.056) 907 381 526 0.160

0.526*** (0.057)

0.747*** (0.071)

0.905*** (0.086)

-4.337*** (0.330) -0.212*** (0.019)

0.151

0.200

0.192

0.222

0.758

0.749

0.788

0.783

0.804

44