The effect of bank capital on risk accumulation and value : analysis of a complete crisis cycle

(1)

University of Amsterdam

Amsterdam Business School

MSc. Business Economics – Finance Track

Master Thesis

The Effect of Bank Capital on Risk Accumulation and Value:

Analysis of a Complete Crisis Cycle

Dario Pabst 11144769 Supervisor: Dr. R. Vlahu

(2)

STATEMENT OF ORIGINALITY

This document is written by Dario L. Pabst, who declares to take full responsibility for the contents of the document.

I declare that the text and work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

ABSTRACT

This thesis empirically analyses how bank capital affects value and risk-taking over a full crisis cycle. Based on pre-crisis measures of bank characteristics, credit risk estimated during the build-up of the recent subprime crisis and market values during the crash period are examined. The main findings show that capital increases the losses during the downturn due to the heightened incentive for risk acquisition materializing during times of rapid expansion. Multiple robustness checks are performed, while results are shown to hold when adjusting the dependent or independent variables, the sample and the crisis timeline.

(4)

TABLE OF TABLES

TABLE 1 - DESCRIPTIVE STATISTICS 24

TABLE 2 - SELECTED ADDITIONAL SUMMARY STATISTICS 25

TABLE 3 - MAIN REGRESSION RESULTS 26

TABLE 4 - ROBUSTNESS TEST I 30

TABLE 5 - ROBUSTNESS TEST II 32

TABLE 6 - ROBUSTNESS TEST III 35

TABLE 7 - ROBUSTNESS TEST IV 44

TABLE 8 - ROBUSTNESS TEST V 45

TABLE 9 - AUGMENTED ENDOGENEITY TEST 46

TABLE OF FIGURES

FIGURE 1 – CRISIS TIMELINE 8

FIGURE 2 – UNITED STATES GDP 16

FIGURE 3 – BUSINESS AND FINANCIAL CYCLES IN THE UNITED STATES 36

(6)

1. INTRODUCTION

Retrospective analysis of the recent financial crisis has shown that there were many different forces driving the downturn. Of major influence were the excessive developments in securitized banking, effectively creating an opaque environment of contagion and concentration in the banking industry, a network of assets tied up in an asymmetrical bet associated with real estate values (Acharya et al., 2010; 2009). Specifically, Jorda et al. (2016), show that especially a bank’s exposure to household debt in form of mortgages has escalated rapidly in the first decade of the 21st_{century. In the same vein, Cole and White (2012) provide evidence of bank failures following}

the crash driven by real estate holdings, amongst other factors. Fuelled by lax monetary policy determining interest rates considered too low for too long, as well as meagre regulation of financial intermediaries, the financial market slid into a crisis after the initial crash (Maddaloni and Peydro 2011). Losses and panic were allowed to spread into various sectors previously unlinked to real estate and mortgages through the short-term credit market, triggering the infestation of the global stock markets (Gorton and Metrick, 2012a; 2012b). Thus, the crash in 2007 matured into a severe recession in part due to the coinciding downturn of the financial cycle, and the preceding tightening of lending standards (Lown and Morgan, 2006). The financial cycle acted as a catalyst, causing the debt levels in the financial system to increase rapidly during the upswing, and then halted investment levels due to debt overhang after the bust (Drehmann, 2012; Borio, 2014). Many banks were stressed on both fronts, with credit lines being drawn by borrowers while creditors decreased short term funding and new lending to large borrowers diminished (Ivashina and Scharfstein, 2010). In the end, the massive collapse caused multiple bailouts in the financial sector and Lehman Brothers Holding Inc. dissolved.

After the dust had settled, the global markets had withstood the most severe recession since the 1930’s in the wake of the subprime mortgage crisis of 2007. Historically, the inherent prevalent problems of the banking industry have plagued the markets time and time again, urging along the creation of new financial regulation. One regulatory body has enjoyed particular scholarly interest: The Basel Committee on Banking Supervision (BCBS) released the first Basel Capital Accord in 1988, establishing international minimum capital requirements and leveling the playing field for U.S. banks. Recently in 2010, Basel III was announced as the third extension of the Basel accords, and has new regulatory frameworks such as liquidity ratios, in addition to refined capital requirements. Today, one of the most important pillars of banking supervision is based on capital requirements. In the traditional view of financial regulation, equity capital will have various positive effects on a banking institution. Generally, it acts as a mechanical barrier,

(7)

absorbing incurred losses and preventing costly failures, while additionally mitigating moral hazard (Admati et al., 2013).1_{Recent academic discussion highlights the positive aspects of Basel}

III and enforced capital ratios, but also mentions many shortcomings. Basel III will uphold new standards for banks that are specifically designed to avoid a repetition of the events in 2007. One measure, for example, is a counter-cyclical buffer as part of Basel III’s capital requirements, which should avoid increased leverage accumulation in times of high growth (Repullo and Saurina, 2011). Blundell-Wignall and Atkinson (2010), argue that while the new regulations are a step in the right direction, they are not capturing the essence of the problem. Risk-weighted assets are still being distorted via credit insurance, while capital requirements favoring equity and specific types of debt will shift the banks portfolios into imperfect positions.

As a response, several empirical studies have been conducted to estimate the effect of capital on various economical measures of a bank’s health. On the one hand, the International Institute of Finance (2010) claims that banks will have to issue $ 0.7 trillion of common equity and $ 5.3 trillion of long-term debt by 2015, with return on equity (ROE) falling roughly two percent annually for American banks. On the other hand, a recent study by Banque de France investigates the relation of capital structure and ROE and finds that an increase in common equity appears to have a positive effect on profitability (de Bandt, Camara and Pessarossi, 2014). In the same vein, capital can be shown to improve market value (Mehran and Thakor, 2011), and increase a bank’s liquitiy creation (Berger and Bouwman, 2009). Notably, both empirical and theoretical literature in this field shows noteworthy discrepancy. However, most work on capital requirements merely includes the effect of capital on bank behavior in the derivation of their hypotheses, but fails to consider it’s empirical influence on the incentives to accumulate risk. Furthermore, the existing contradictions in studies on bank capital show the essence of additional empirical mediation that could be provided by this incentive based channel. In theory, capital could encourage a bank to seek higher risk exposure, which will be elaborated more carefully in section 3. This thesis aims to establish a framework for the empirical mediation of the effect of bank capital on market value through its influence on investment behavior. Considering the recent subprime mortgage crisis, first, the effect of pre-crisis capital on market value during the downturn is assessed in order to show the effect of holding increased amounts of equity during the build-up of the mortgage bubble (Figure 1). Then, in order to capture the mechanism through which risk accumulation is affected, the relationship between pre-crisis capital and

1_{It should be mentioned here that there is a strict definition and separation on behalf of the Bank for International}

Settlements regarding capital and equity, but the two terms are used interchangeably in this thesis. Even though capital includes a broader spectrum of subordinated debt, it is not the ultimate aspiration to investigate the differences between these types of capital.

(8)

different measures of on- and off-balance-sheet risk during the period of the market upswing is measured.

Notes: The timeline shows the three crisis periods used to analyze the impact of capital on bank health. Capital is measured during the pre-crisis period and linked to measures of risk taking during the build-up and to market value during the downturn using the empirical approach elaborated in section 4.1.

Figure 1- Crisis Timeline

2. LITERATURE REVIEW

While the vast field of financial regulations has gained many interesting additions in the last decade, there has been ongoing discussion of bank capital holdings. This section will develop the discourse leading up to the modern view of equity capital, specifically for financial institutions, in order to place the conducted empirical tests in a broader context. First it will be highlighted why banks generally operate with large amounts of leverage, based on the role they fulfill in the modern financial system. Next, the theoretical importance and practical implications of capital regulation for establishing a more stable financial market are explained2_.

2.1 Incentives for High Leverage

Josef Ackermann, CEO Deutsche Bank, said in an interview as recent as 20th November 2009: “More equity

might increase the stability of banks. At the same time, however, it would restrict their ability to provide loans to the rest of the economy. This reduces growth and has negative effects for all.”

(Thakor, 2014)

Financial institutions are an integral part of the world’s economy, performing a multitude of important tasks. At the basic level, banks engage in qualitative asset transformation by holding deposits, and consequently providing companies with funds. In other words, financial

2_{Two recent reviews of the academic literature pertaining to capital requirements additionally provide a relatively}

straightforward summary: Santos (2001) and Van Hoose (2007). The latter review places particular emphasis on the ambiguous findings of researchers, and questions the practicality of extending capital regulation in the absence of clear evidence for positive effects.

1998-2003 Pre-Crisis Crisis Build-up 2004- June 2007 July 2007-2009 Crisis Downturn

Crisis Timeline

(9)

intermediaries are vital because they enable a coalition of borrowers and investors to engage in mutually beneficial lending that would otherwise be unfeasible (Bhattacharya and Thakor, 1993). In addition, depository intermediaries provide demandable monetary claims and fulfill monitoring duties (Greenabum, Thakor and Boot, 2015, Part III, IV; Diamond, 1984; Ramakrishnan and Thakor, 1984; Brealy, Leland and Pyle, 1977). By issuing retail deposits banks provide liquidity insurance to individual households facing unknown idiosyncratic risks, creating significant social value. Due to the importance of a bank’s debt and their structure of value creation, as well as the tax benefits of debt, intermediaries have exhibited naturally high leverage. Jayaraman and Thakor (2013) analyze banks in various countries around the globe and state that their average capital structure is comprised of 8% equity and 75% deposits. Banking professionals argue that an increase in equity would decrease lending, leading to a negative impact on investment activity. They also maintain that highly levered banks experience a disciplinary effect of debt. Depository institutions are exposed to bank runs, since they provide demandable claims as their main source of funding (Diamond and Dybvig, 1983). In order to afford liquidity to creditors, intermediaries must hold less fixed assets than liquid deposits. In such an environment, a creditor’s expected value is related to the sequential service constraint and they might attempt to withdraw as soon as possible if there are indications of poor performance. Hence, the threat of bank runs induces management to be more effective in their lending (Jacklin and Bhattacharya, 1988).

Nevertheless, besides the nature of a bank’s value creation there are other reasons why it may be beneficial to hold large amounts of debt, in the face of ever increasing pressure for higher levels of equity. Retail deposits are insured by government guarantees, which creates an agency rigidity related to limited liability, labeled risk shifting. This asset-substitution moral hazard in banking due to deposit insurance might cause shareholders to pursue suboptimal projects, given high enough leverage (Cooperstein, Pennacchi and Redburn, 1995; Gropp, Gruendl and Guetter, 2013). Merton (1977) as well as Jensen and Meckling (1976) argue that shareholders equity can be viewed as an option-like security with a claim on the total profits of the bank. A bank’s management will increase the volatility of the underlying by choosing riskier investments to maximize the payoff. Moreover, deposit insurance relieves the threat of debt and its disciplinary effect, increasing the incentive to choose imperfect portfolio allocations. Another concern is the common compensation scheme at financial institutions, which is based on return on equity (ROE). Less capital on the balance sheet translates into a higher ROE, ceteris paribus, and hence encourages high leverage. Additionally, Mehran and Thakor (2011) hold that managers are likely to underestimate the probability of a financial crisis occurring, specifically in a prosperous time.

(10)

Thus, banks might set their capital structure in a way that is not socially effective, and financial regulators must intervene with capital requirements. The amount of equity capital held by a financial intermediate is of exceptional interest, since it reflects the ability to withstand shocks created by the particular threats of the banking sector.

There is also some empirical evidence for the implementation of capital requirements being quite costly and restrictive to economic growth. These expenses arise from the need to raise additional capital and due to the tightening of lending standards that will have occurred beforehand. As elaborated by Slovik and Cornéde (2011) the implementation of Basel III will cause the European gross domestic product (GDP) growth to slow down by 0.08 percentage points per year, whereas the growth of the United States GDP will decrease by 0.02 percentage points. In a similar vein, the Institute for International Finance (2010) argues that the ROE of U.S. banks will drop by two percent per year due to fundraising costs. However, operating at a higher level of capital is in itself not expected to have negative effects after the roll-out period (Admati and Hellwig, 2014), as discussed in the next section.

2.2 The Benefits of Capital

Given the circumstances that encourage bank shareholders and management to engage in excessive risk taking, financial regulation is actively trying to limit negative repercussions of the banking sector through different measures, like liquidity ratios and investment guidelines (Barth, Li and Lu, 2010). Particularly capital requirements are considered one of the most important pillars of micro- and macro-prudential regulation for multiple reasons. Regulators argue that the provision of demandable liquidity also exposes the institution to panic-induced bank runs, as opposed to information-based bank runs (Chari and Jagannathan, 1988; Calomiris and Gorton, 1991). Panic-induced runs have no positive effect in terms of discipline for intermediaries, since they are irrational and not caused by poor performance. These runs are socially costly even at the individual bank level, because they lead to premature liquidation of profitable assets. In extreme cases, both irrational and information-based runs can create contagion in the financial sector resulting in system-wide distress or failure (Aghion, Bolton and Dewatripont, 2000). As mentioned earlier, equity resembles the first line of defense, and will partially absorb any economic shock caused by a market downturn or banking crisis (Greenabum, Thakor and Boot, 2015, Chapter 13). Proponents of high leverage also acknowledge this argument, since they imply a trade-off between stability, and lending/ liquidity creation. However, Admati et. al (2013), Thakor (2014), as well as Miller (1995), each provide a dissertation that challenges most

(11)

arguments put forth by practitioners, and concludes that bank equity can in fact be value enhancing. For example, they argue that an increase of capital will only decrease deposits available to the public if the size of the financial institution is held fixed, because otherwise the balance sheet can just be extended to arrive at the necessary amount of deposits. More generally, the authors seemingly refute the banker’s claim that the Modigliani and Miller propositions (Modigliani and Miller, 1958; 1963) do not apply to the financial sector. So at the very least bank capital provides a buffer to banks that will increase survivability and prevent costly bailouts or failures. Moreover, the cushion can prevent a financial institution to succumb to the pressure of fire sales in a contagious crisis environment, reducing liquidity risk (Shleifer and Vishny 2010, Acharya, Mehran and Thakor 2015).

Another well-known line of reasoning is concerned with overcoming the phenomenon of risk shifting, mentioned above, through increasing the amount of equity in a bank. By inflating a banker’s “skin in the game” capital will alleviate the incentive to choose riskier assets at the expense of the general public (Hellman et al., 2000; Repullo 2004; Repullo and Suarez, 2004). Additionally, Holmstrom and Tirole (1997) as well as Mehran and Thakor (2011) have reassured the notion that increased capital would positively affect the behavior of banks in another fashion. The aut hors argue that holding additional equity encourages increased monitoring of borrowers and aligns incentives with financial stability, due to the higher costs of failure. Strengthened monitoring effort enables better financing conditions for valuable borrowers and enhances the benefits of survival that come with capital in the long run. Finally, a bank also experiences agency problems common to most firms with public investors, compounded by the fact that depositors are mostly laymen (Dewatripont and Tirole, 1993a; 1993b). These unsophisticated investors face prohibitive monitoring costs and have little incentive to engage due to their relatively small individual claims in the entity’s welfare. Hence, regulators are required to intervene by enforcing legislature like capital standards or deposit insurance. However, these methods can have costs themselves, just as governmental guarantees for deposits can induce moral hazard.

There is also considerable empirical support for the positive effect of capital on measures of bank health. An early study by Peek and Rosengren (1997) reports a positive relationship between equity shocks at international Asian banks and the lending activities of their American subsidies. Increased profitability has been associated with higher capital in European banks (Staikouras and Wood 2011), and French banks in particular (de Bandt, Camara and Pessarossi, 2014). Another recent study on Asian banks has found a significant effect of bank capital on profitability, and finds that riskier investment banks actually have a stronger positive effect on ROE caused by

(12)

holding additional capital, than safer commercial banks (Lee and Hsieh, 2013). Mehran and Thakor (2011) show a significantly positive relation between the market value of a bank and its equity capitalization, arguable due to a stronger incentive to monitor borrowers. Capital also positively affects liquidity creation (Berger and Bouwman 2009), while a bailout produces neither a positive nor a negative effect (Berger et al. 2014). Hence, while bailouts have often been compared to preemptive capital holdings, they seem to be inefficient measures in terms of liquidity creation. Finally, Berger and Bouwman (2013) show that higher capitalization increases banks survival probability and market share during financial crises.

It seems then, that capital has a positive influence on many crucial empirical metrics as opposed to the common interpretation of practitioners. Nevertheless, there are discrepancies in the discussion as well as in the empirical results, relating the need for additional review. Specifically, while a host of studies confirm the positive influence of capital on a bank’s risk taking behavior as mentioned earlier (eg. Repullo, 2004), VanHoose (2007) points out that this conclusion only holds if capital holdings and returns of a bank are exogenous. However, the empirical evidence suggests either a decrease in returns during the introduction or increase of capital requirements (e.g. Institute of International Finance, 2010), or an increase in profitability when operating at a higher level of capital in the long run (e.g. de Bandt, Camara and Pessarossi, 2014). Moreover, since the real benefits of tax-exempt debt matter for a bank’s health, the broad conclusion is at best ambiguous. There is another less prominent string of theory that states the possible negative consequences of increased capital based on an incentive to accumulate risk. The next section will elaborate the mechanism behind these theories and highlight the critique on the established perceptions of bank capital.

3. HYPOTHESIS DERIVATION

A majority of academic work considers equity holdings of financial institutions to have positive effects on the stability of the financial sector. However, similar to the drawbacks arising from deposit insurance, capital requirements might create hidden moral hazards. Acharya, Mehran and Thakor (2015) argue that capital requirements can make bank debt too safe and mitigate market discipline. This may lead to a situation where banks over-lend and choose socially inefficient portfolio allocations, expecting ex-ante governmental intervention and support. An early study by Koehn and Santomero (1980) states that if a bank is forced to hold more capital its efficient investment frontier will decrease because the bank cannot achieve exactly same leveraging capabilities. In other words, a bank can achieve the same profitability in the presence of binding

(13)

capital requirements, but it is inclined to choose slightly riskier strategies after altering the capital structure. This is especially true considering the cycle of a financial crisis3_{. During the build-up of}

a crisis there is an abundance of enticing investment potential, with unknown risk implications. In prosperous times of rapid expansion of bank credit, adjusting the structure of a bank’s balance sheet to adhere to capital requirements is likely to induce an unconscious accumulation of risk. Moreover, a bank’s balance sheet is so heavily based on government-secured deposits that any capital requirements will not outweigh the incentives of limited liability (Perotti, Ratnovski, Vlahu, 2011). To a certain extent this could mean that increased capital will not cause the bank to refrain from excessive risk taking. The accumulation of additional capital above and beyond the requirements, allows a financial institution to invest more heavily into securities carrying tail risk, without alerting supervisors and regulators through an insufficient capital ratio. And indeed, Berger et al. (2008) report much higher capital holdings than would be necessary based on current regulations. Moreover, the financial crisis exposed multiple valuable banks with relatively high capital ratio holding a wealth of assets with systemic tail risk, causing them to fail.

Martynova, Ratnovski and Vlahu (2014) develop a theoretical framework of the possible negative effects of capital. The authors provide a model where intermediaries with large amounts of capital or a high franchise value could accumulate new risk very quickly, if systemic threats are misinterpreted. They argue that banks expose themselves to risk by levering up due to shareholders preference for debt, and invest the liabilities into risky market-based instruments. Bank capital and franchise value act as guarantees and will enable financial institutions to borrow additional funds in a more rapid fashion. This could prove problematic in an opaque environment where risk is practically unobservable in good times, because the exposure accumulated by engaging in market-based instruments can outweigh the buffers held back by capital requirements in the long run. Protection of creditor rights in the form of deposit insurance will also strengthen the incentive to invest into riskier securities, particularly in a landscape of dangers contained in tail events. In other words, if more capital is held, the chaotic environment of a leverage bubble will encourage an unobserved increase of tail risk on a bank’s balance sheet. Shleifer and Vishny (2010 b) agree, claiming that “liquid markets created tremendous opportunities for banks, and profit maximization pushed banks to take them.” (see p.39). Another factor providing incentives for increased risk can be found in lax monetary policy.

3_{Arguments in a similar vein have been criticized in that increased equity only has implied negative consequences if}

the banks balance sheet is fixed in size (Admati et al., 2013; Miller, 1995). However, these critics do not account for a volatile crisis environment. Koehn and Santomero (1980) also show that banks holding relatively volatile portfolios before the implementation of new capital requirements choose to restructure using even riskier investments than their safer counterparts.

(14)

Maddaloni and Peydro (2011) argue that banks will need to find other sources of profit due to the decrease in lending spreads and net interest margins, caused by interest rates that are considered too low for too long. The successive softening of lending standards due to a decrease in profitability as well as tempting liquid markets have caused banks to accumulate additional credit risk, conceivably reinforcing the negative effect of increased capital. Additionally, multiple researchers and professionals pointed out that the Basel III regulations are simply not enough, as incentives have not shifted and the creation of an opaque financial market via distortion of risk weights is not substantially hindered (Blundell-Wignall and Atkinson 2010, Mariathasan and Merrouche 2014). DeAngelo and Stulz (2013) elaborate that the regulation of equity holdings and the coercion into a more conservative capital structure cause banks to loose competitiveness with regards to shadow banks. These unregulated entities, able to hold an optimal amount of leverage, will threaten the vital commercial banks and cause a reduction in bank efficiency. In light of these arguments, one can conclude that capital regulation might create moral hazards, especially during turbulent times.

It seems as if the cyclicality of bank financing creates problems concerning perceptions of risk and value, influencing the impact of capital on bank incentives during certain periods of rapid expansion. To test the developed theory, the employed empirical framework will examine the effect of pre-crisis capital on risk accumulation during the boom, and market valuation during the bust. As can be seen in the hypotheses reported below, pre-crisis bank equity is expected to increase the risk on a bank’s balance sheet during the build-up, which causes a sharper decline in value during the downturn of a crisis. A distinction of different crisis and non-crisis periods is an important development in the methodology of bank capital studies, and might illuminate why there is considerable ambiguity in the academic literature. As opposed to the divergent expectations about the effects of capital, there might be distinct conclusions concerning the capital holdings going into a crisis. Theoretically, it is substantial to analyze how capital affects bank behavior during the development of a crisis since it is unknown when a disaster will occur. The following two hypotheses are formulated based on these iterations, while the methodology utilized to test them is elaborated in the next section.

Hypothesis 1: Pre-crisis bank capital has a negative effect on the relative market value during the crisis-downturn.

Hypothesis 2: Pre-crisis bank capital has a positive effect on the accumulation of risk during the build-up of a crisis.

(15)

4. DATA AND EMPIRICAL APPROACH

The following sections will elaborate the methodology and variables, describe the origin of the data and finally show descriptive statistics. Beforehand, the significance of the applied empirical approach is addressed, since it is vital to consider the effect of increased capital holdings at the crisis inception. The use of lagged variables, measured before financial turbulence could be foreseen, shows whether a high capital ratio at the beginning of a crisis leads to a beneficial outcome during the downturn. This approach might create results that help guide regulators with insights as to how increased equity affects bank behavior during the build-up of a crisis. Furthermore, the literature review establishes that academics expect a link from capital to performance, while theoretically there should be no such link in the opposite direction due to banks capital ratios being heavily regulated. However, because of intertemporal preferences affecting bank decision making, performance might also influence the capital choice of a bank. This endogeneity issue is addressed by the structure of the employed analysis, since the market to book ratio during the crash period is less likely to have a causal link with pre-crisis capital as a strongly lagged determinant.

The empirical approach is similar to a study conducted by Berger and Bouwman (2013), which estimates the impact of lagged capital measures on a bank’s market share and survival rate. The authors initially conduct analyses by grouping together crisis types, in order to find characteristic patterns for banks undergoing crises either caused by the banking sector or deriving from a stock market crash. Building upon the fundamental design, this thesis analyses a bank’s capital ratio and other bank characteristics during the pre-crisis phase and relates them to the market to book ratio during the crash, as well as to measures of risk during the build-up of the recent subprime crisis. Hence, the following evaluation considers only a single occurrence, enabling conclusions about specific regulatory or macroeconomic conditions only observed during this time. Peculiar to the episode of the 2007 downturn is first and foremost the real estate bubble fuelled by opaque securitized banking investments (Acharya et al., 2009; Jorda et al., 2016). After banks experienced large losses on extensive subprime mortgage positions, the shortfall spread through an immobilized short-term credit market into the global financial system (Gorton and Metrick, 2012a; 2012b). Moreover, active government intervention in terms of bailouts and monetary support were an important supplement for capital during the crash period (Berger et al., 2014), while regulatory frameworks created a false sense of security during the build-up (Perotti, Ratnovski and Vlahu, 2011).

(16)

In order to establish the crisis timeline (Figure 1), either the conventional understanding of the crisis based on movements in GDP, or estimations of the financial cycle are used. It is consequential to consider how pre-crisis capital will be estimated, since the hypotheses are based on this notion. The main analysis will be performed estimating the pre-crisis measures 1998 through 2003, the upswing from 2004 until the second quarter of 2007, and the crash from the third quarter of 2007 through 2009. This approach is consistent with gross domestic product behavior, since the measure exhibits the sharpest increase during 2004 presumably inducing a strong expectation of ongoing prosperity characteristic to the build-up of a crisis (Figure 2). On the other hand, GDP starts to diminish in mid 2007 heralding the outset of the recession. An alternative timeline constructed according to shifts in the financial cycle will be used to perform robustness tests, whereas implications and results are analyzed in section 6.5.

Notes: The figure shows the level of United States GDP in U.S. $ Billions from 1998 until 2009, measured by the red line. The data was gathered from the World Bank Database (The World Bank: United States: GDP, Source: World Development Indicators)

Figure 2- Gross Domestic Product, United States

4.1 Methodology

The following section will describe how the two different time periods of the subprime crisis, the upswing and the downswing, will be analyzed in specific detail based on the bank characteristics observed during the pre-crisis period. The crash period will be analyzed using the average market to book ratio of a bank, whereas the build-up period is estimated using a bank’s average credit

11,500 12,000 12,500 13,000 13,500 14,000 14,500 15,000 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 $ Bi ll ion s Years

Gross Domestic Product, United States

(17)

risk. Both regressions will then relate the dependent variable of interest to pre-crisis capital. All analyses are linear ordinary least squares regressions estimated using robust standard errors clustered by bank, to account for heteroskedasticity in the error terms.

4.1.1 Crash Regression

The OLS regression used to analyze the first hypothesis during the crash period is: %Δ 𝑀/𝐵_!!"#$! _{= 𝛽}

! + 𝛽!∗ 𝐿𝐸𝑉 𝑅𝐴𝑇𝐼𝑂!!"#$"%&%& + 𝛽!∗ 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠!!"#$"%&%&

Where the dependent variable %Δ 𝑀/𝐵_!!"#$! (or %ΔM/B CRASH) is the percentage change in the average market to book ratio of bank i during the crash, estimated from the third quarter of 2007 until the end of 2009. The market value or market capitalization is defined as shares outstanding in quarter t multiplied by the quarterly closing price at the same time. To find the market to book ratio, market value is divided by total assets, and the averages over the pre-crisis and crash periods are constructed. Specifically, total assets are a widely used proxy for size relating the extent of on-balance-sheet activities of an intermediary.4_{Deducting the average}

pre-crisis market to book ratio from the crash market to book ratio, dividing by the pre-pre-crisis market to book ratio and multiplying by 100 estimates the dependent variable. This regression should reveal a bank’s change in market value during the downturn, and how it relates to its pre-crisis equity capital. Market to book ratios are considered since they disclose the perceived value of an institution relative to the notional value of its assets, a measure particularly sensitive to the changing expectations of value occurring during the start of a market crash. As mentioned earlier, most variables in the sample were winsorized once at the 2.5% level, whereas constructed ratios are winsorized after calculation to correct for impossible observations and mitigate the effect of outliers. The independent variable as well as the controls are estimated as averages over the pre-crisis period, in order to reduce the influence of anomalies in the sample. 𝐿𝐸𝑉 𝑅𝐴𝑇𝐼𝑂!!"#$"%&%&

(or LEV RATIO) is defined as the ratio of common equity to total assets of bank i. In a later section, robustness tests using tier 1 and total regulatory capital ratios are conducted, and the estimations hold. To account for other bank characteristics and alleviate omitted variable bias, both main regressions include a range of controls.

4_{Total assets might not be entirely representative of a bank’s total activities since it does not account for its asset}

transformation capabilities, or off-balance-sheet items (Boot, Greenbaum and Thakor, 1993; Boot and Thakor, 1993). In order to address this issue, Berger and Bouwman (2013) additionally use a liquidity-creation based approach to estimate the relation of market share and capital holdings, and find similar results as compared to using measure of on-balance-sheet items.

(18)

First, all regressions include a proxy for profitability, since more lucrative entities are likely to have a greater market value during the downswing. However, high profitability might also represent easier access to credit and involve riskier investments. The main regression includes net interest margin (NIM), a measure of on-balance-sheet profitability that shows the difference between interest income and interest expenses, normalized by total interest earning assets. This measure has been retrieved from Wharton WRDS COMPUSTAT as is, whereas robustness tests will be performed with a constructed variable (ROA) and a proxy retrieved from Wharton WRDS CRSP (EPS).

As a placeholder for the exposure to retail deposits a bank’s total deposits divided by total assets is included in the regression (TOT DEPOSITS), since it is relevant to capture their unique influence. Retail deposits enjoy governmental protection and can induce asset-substitution hazard, increasing the risk taking in banks covered by the safety net (Merton, 1997). On the other hand, according to the Basel III Net Stable Funding Ratios (NSFR) retail deposits are a relatively safe type of funding, and might therefore decrease the overall risk character of an intermediary. As a robustness check, the amount of demand deposits divided by total assets (DEM DEPOSITS) will be included in a later section.

Real estate played a major role in the subprime crisis, with banks displaying an excessive increase in real estate and related holdings, which exposed them to significant risk related to the housing market (Jorda et al., 2016). Cheng, Raina and Xiong (2014) argue that the complex nature of banks real estate activities caused extensive losses during the downswing, after unexpectedly high failures in subprime mortgages. Evidence suggests that especially commercial real estate was an important determinant of bankruptcy in financial institutions (Cole and White, 2012). As descriptor of a bank’s exposure to the commercial real estate sector, a bank’s commitment to commercial real estate funds (secured by real estate) normalized by total assets (CRE LOANS) is included in the regression. Among the robustness checks, the analysis is repeated using total loans secured by real estate divided by total assets (RE LOANS) instead, and the results are solid. Further, cash holdings and funds due from other banks divided by total assets (CASH DFB) are included in the regression. Cash is the most liquid of assets and is expected to make banks less prone to losses, but can also provide excess liquidity if credit markets become rigid. Nonetheless, some scholars argue that excess cash can also lead to moral hazard and agency issues, especially when there is an abundance of promising investment opportunities. (Jensen, 1986; Malmendier

(19)

and Tate, 2005). To test the findings of the main regression, cash and short-term investments as a portion of total assets (CASH STI) will be used as a control in an additional analysis.

Total Basel I risk-weighted assets divided by total assets (CRED RISK) is included as a proxy for risk as well as complexity (Berger et al., 2012). The measure includes an institution’s assets and off-balance-sheet activities weighted by the perceived risk of the investment as classified by regulators. This variable is also used in the build-up regressions as a dependent variable, and will be discussed in more detail shortly. Robustness checks are performed using a measure of Basel I assets, derivatives and off-balance-sheet items in the Basel one hundred percent risk category (ADOBS).

Next, a measure for the effects of bank size is included, using the natural logarithm of total assets (LOG ATQ). Controlling for size is important because risk and contents of a bank’s asset portfolio are expected to deviate depending on the size of the entity (Berger et al., 2005). Specifically, large intermediaries seem lend more extensively to companies with adequate accounting records, and are less reliant on relationship banking. Moreover, larger banks are more likely to be bailed out, which could provide management and shareholders with additional incentives to accumulate risk. As can be seen from the descriptive statistics, larger banks seem to hold less capital, likely because larger consolidated banks that might have subsidiaries, sister- or parent-firms have the ability to raise additional equity from their related institutions. On the other hand, access to financial markets is relatively reliable for larger intermediaries, but can bear prohibitive costs for small banks. Particularly, since the dataset lacks the extent that would be necessary to regress banks grouped by size, it is essential to control for the magnitude of a bank’s activities.

An intermediary’s exposure to short-term credit is measured as its debt in current liabilities divided by total assets (CURR DEBT). The ratio is included to further adjust for the influence of a bank’s funding and strategic decisions since an intermediary relying on more stable funding might be trying to avoid contagion. Gorton and Metrick (2012a, 2012b) point out that the highly interrelated short-term credit market caused high losses for banks, but quickly became a large liability also for healthy institutions during the recent subprime crisis.

Drehmann (2012) elaborates the nature of the financial cycle, including the reinforcing nature of debt levels and housing prices. As real estate is often used as collateral, and funded to large extent by credit, it will increase the debt in the financial system, as well as feed off it, which can lead to a

(20)

boom-like expansion. At the very least, an increase in real estate prices associated with an extension of lending output as measured by the financial cycle is thought to enhance bank performance. The 2007 banking crisis was especially associated with the failure of mortgages, hence the quarterly percentage change in the state-specific Housing Price Index (HPI) is included in the regression. It is constructed by subtracting the one quarter lagged state-level HPI in quarter

q from the value in quarter q itself, dividing it by the lagged HPI and multiplying by one hundred.

Since the dataset is missing information on bank activities according to state, the location of the headquarter is used to determine the state a bank is likely to be most active in. The data used to estimate the change in HPI was retrieved from the Federal Housing Finance Agency, using an expanded index. Expanded Indices contain enterprise and FHA data and refine it using Real Property County Recorder Data (ww.fhfa.gov/DataTools).

4.1.2 Build-up Regression:

While academics have argued that equity ratios can in fact make a bank safer, more profitable, and apparently enhance liquidity services (Mehran and Thakor, 2011; Admati et al., 2013; Berger and Bouwmann, 2009), there might be a crucial difference in the way capital affects banks in different stages of a crisis. According to theory, capital could be utilized to produce a swift increase of risk and market exposure on a bank’s balance sheet since it conveys a structural reliability, a capacity for losses, which might not exist (Martynova, Ratnovski and Vlahu, 2014). This is occurring through the acquisition of risky market-based investments using credit that is cheaply available due to perceived safety of a bank. Intermediaries are unlikely to internalize the losses of tail risk heavy activities while capital ratios are used to make the bank seem safe by making it less prone to failure in shocks of medium magnitude (Perotti, Ratnovski and Vlahu, 2011). To analyze whether the effects of capital on the relative valuation act through this channel in the expected way, credit risk is regressed on a similar main estimation, where pre-crisis market value is included as a control variable.

The main regression used to estimate risk accumulation during the build-up of the crisis will be: %Δ𝐶𝑅𝐷 𝑅𝐼𝑆𝐾_!!"#$%!!" = 𝛽!+ 𝛽!∗ 𝐿𝐸𝑉 𝑅𝐴𝑇𝐼𝑂!!"#$"%&%&+ 𝛽!∗ 𝐶𝑜𝑛𝑡𝑟𝑜𝑙𝑠!!"#$"%&%&

Where %Δ𝐶𝑅𝐷 𝑅𝐼𝑆𝐾_!!"#$%!!" (or % 𝚫 CRD RISK) describes the percentage change in average credit risk of bank i. The risk factor is measured as average during the build-up period as of January 2004 until the end of the second quarter of 2007. Credit risk is analyzed using the Basel I

(21)

weighted assets, defined as assets and off-balance-sheet items of a bank multiplied by risk-weights, divided by total assets (Basle Committee on Banking Supervision, 1988). The change is then constructed as the pre-crisis average ratio deducted from the average value during the build-up, normalized by the average ratio during the pre-crisis phase and multiplied by one hundred. This proxy for risk captures the on- and off-balance-sheet exposure of banks, theoretically including exposure to derivatives or CDOs. Hence it should not only capture traditional banking risk, but also illuminate the complex nature of securitized banking so characteristic to this period. Moreover, regulatory Basel I risk-weighted assets have been used in various empirical studies to assess the risk factor of a bank (Berger et al., 2012; Berger and Bouwman, 2013; Dionne and Harchaoui, 2008; Logan, 2001). The set of controls for the build-up regression is slightly different from the one originally used, excluding pre-crisis credit risk but including pre-crisis market to book ratio. The estimation is adapted by eliminating credit risk as a control to account for correlation in the measures, especially since there is less of a lag between pre-crisis and build- up phases. Similar robustness checks will be performed on the build-up regression after the empirical results have been reported.

4.2 Dataset and Descriptive Statistics

In order to create a coherent and significant set of observations various databases are utilized as sources. The basis is provided by bank fundamentals of institutions incorporated in the United States, which were retrieved from Wharton WRDS “COMPUSTAT Bank Fundamentals Quarterly”. The rudimentary dataset incorporates mostly commercial banks, excluding entities that have less than $ 1 billion in assets or were founded in Puerto Rico. This study will focus on commercial banks for two reasons: First, investment banks hold more complex assets and carry increased market risk, which could create bias in the estimation due to differences in strategy and behavior, as well as obstacles in obtaining a significant dataset. Second, commercial banks and other depository institutions enjoy greater safety net protection but have to endure stricter regulation. Consequently, unique incentives are present to accumulate market risk among such institutions and they are required to provide extensive documentation by law. Small banks ($1 billion of assets and below) are excluded from this sample since Berger et al. (2005) report differences in the asset portfolio of a bank depending on its size5_{. Observations of banks}

5_{Furthermore, Berger and Bouwman (2013) find endogeneity issues in small banks ($1 billion assets and below), but}

not in medium or large banks, when analyzing the effect of pre-crisis capital ratios. In their additional analysis of individual crises, the authors conclude that their variable of choice, the effective tax rate of a bank, is not a significant instrument for the recent subprime crisis. The results for effective tax rate as an instrument during the recent crisis have been reaffirmed and are discussed in Appendix 3. Therefore, while the reported estimations should

(22)

established in Puerto Rico are excluded due to legislation that hinders local intermediaries from using U.S. bankruptcy procedures, as well as a weaker economy in general. These special circumstances mitigate benefits derived from governmental support during distress or might potentially skew the estimation otherwise. Furthermore, entities that have less than four years of continuous observations as well as banks starting operations after 2003 are dropped to exclude banks that would not provide enough data for consistent estimation.

The resulting dataset was then merged with figures from Wharton WRDS “CRSP6_/

COMPUSTAT Merged Fundamentals Quarterly” to add stock prices and shares outstanding, as well as the PERMCO permanent company identifier used by CRSP. The Federal Reserve Bank of New York created and maintains a PERMCO-RSSD record, which enables the user to merge CRSP data with WRDS Wharton “Bank Regulatory” data. This database contains information on regulated depository institutions such as bank holding companies (unconsolidated parent-only data was used), commercial banks as well as savings and loans institutions (Federal Reserve Bank of New York, 2014). Linking to parts of the call reports and data archive of the Federal Reserve Bank of Chicago, the RSSD-IDs were used to gather more intricate balance sheet information such as quarterly risk weighted assets or exposure to real estate loans. The final dataset contains 227 state, federal and national commercial banks (Standard Industry Classification code = 6020) and 40 savings institutions (SIC = 6030), of which 23 are federally chartered (SIC = 6035) and 17 are not (SIC = 6036). Observations not present in both the Wharton WRDS Bank Regulatory dataset as well as the initial one are dropped. Lastly, the quarterly expanded house price index by state was added from the Federal Housing Finance Agency, which refines the measure with more granular data such as county specific prices and aggregates it. The datasets were combined using extensive coded commands in the STATA MP statistics package, whereas most variables have been winsorized at the 2.5% level to ensure that all ratios are valid and to reduce the effect of outliers. The STATA code used for creating the datasets, constructing the variables as well as conducting regressions and robustness checks is not included in the thesis, but available upon request.

(cont.) be valid for larger entities, including small banks in the analysis would require a more sophisticated dataset. Further, medium and large banks are grouped together in this thesis because they can be seen to respond in a similar fashion when analyzing pre-crisis capital during banking crises.

(23)

Table 1 Panel A exhibits summary statistics for the main regression variables during the crash period, while Panel 2 shows the same for the build-up regression. The final sample contains 267 different U.S. banks yielding 161 bank observations in the crash-period regressions and 198 in the build up regressions. Even though the crash regression shows fewer observations, the values of all independent variables, measured as averages during the pre-crisis period, are largely similar. Overall, intermediaries seem to adhere to capital regulation with a reported average leverage ratio (LEV RATIO) of roughly 8.5 percent as well as a relatively small standard deviation. As expected, depository institutions in the sample can also be seen to hold large amounts of debt in form of deposits, 74.14 and 73.29 percent of total assets (TOT DEPOSITS) for crash and build-up periods, respectively. When assessing the dependent variables based on percentage changes in the first row of Panel A and B, one can find large amounts of variation as depicted by the relatively high standard deviation. The finding relates the great differences an institutions idiosyncratic characteristics and behavior can have on market to book ratio during the crash or acquisition of credit risk during the build-up.

Table 2 displays different cross-sectional measures of a bank’s leverage ratio and portion of risky assets, to analyze determinants of heterogeneity in the main variables of interest. To further investigate, banks are divided along the median of four different categories before the means and standard deviations of the pre-crisis averages are constructed. On average, larger banks in the sample (institutions above the median in SIZE) are found to operate with less equity while holding more credit risk, compared to smaller entities. The observation implies that governmental monetary intervention awarded to systemically important financial institutions (SIFIs) and too big to fail entities might be used as a substitute for capital buffers. On the other hand, larger entities could have a higher probability of being part of a bank holding company, or having other affiliates, which can facilitate raising capital should they come under distress. The riskier banks in the sample (estimated to be above the median in CRD RISK and RE LOANS) can indeed be seen to hold, on average, more capital to compensate for the potential shortfall. However, observed financial intermediaries that hold proportionally more liquidity risk (above the median in CURR DEBT) on their balance sheet, tend to have less equity despite the relevant exposure. Since regulators do not induce banks holding potentially risky short-term liabilities to increase their capital, banks might abuse the privilege. Merely comparing these crude averages can already relate the origin of significant heterogeneity in the variables of interest.

(24)

Table 1

Descriptive Statistics

Notes: Summary statistics shown for all main regression variables, where all monetary values used for variable construction are nominal and in U.S.$ millions. Crash- variables are estimated from the third quarter of 2007 until December 2009, Build-up-variables use data from January 2004 until the second quarter of 2007, while Pre-Crisis averages were constructed starting January 1998 through December 2002. Panel A shows descriptives for the crash period regression, from a selected set of bank fundamental data including 161 U.S. commercial banks. Market value is constructed by multiplying shares outstanding with the quarterly closing price. The market to book ratio of a bank during the downturn is calculated as market value over total assets, multiplied by one hundred. % Δ M/B CRASH denotes the percentage change in the market to book ratio, measured as the average ratio during the crash period minus the average value over the pre-crisis period, divided by the average pre-crisis value and multiplied by one hundred. All independent variables are measured as averages during the pre-crisis period. LEV RATIO is the leverage ratio, estimated as total common equity divided by total assets. NIM denotes the net interest margin of a bank, retrieved from Wharton WRDS COMPUSTAT. TOT DEPOSITS are a bank’s total deposits, normalized by total assets. CRE LOANS denotes the commercial real estate loans as a proportion of total assets. CASH DFB is cash and funds due from banks divided by total assets. CRD RISK captures the Basel I risk weighted assets and off-balance sheet items. SIZE measures the natural logarithm of total assets. CURR DEBT is the debt in current liabilities divided by total assets. HPI is the quarterly percentage change in the housing price index by state, retrieved from FHFA. M/B denotes the market to book ratio. Panel B shows summary statistics for the crash period regression, from a selected set of bank fundamental data including 198 U.S. banks. . % Δ CRD RISK denotes the percentage change in the risk weighted assets to total assets ratio, measured the average ratio during the upswing minus the average value over the pre-crisis period, divided by the average pre-crisis value and multiplied by one hundred.

Panel A

Variables Mean Std. Dev. 25th p. Median 75th p. Nr. of Obs. % 𝚫 M/B CRASH -33.09 32.66 -56.97 -41.01 -16.00 161 LEV RATIO 8.524 1.624 7.551 8.309 9.473 161 NIM 4.140 0.614 3.758 4.131 4.559 161 TOT DEPOSITS 74.14 8.385 68.49 75.44 80.59 161 CRE LOANS 3.605 2.450 1.844 2.965 4.914 161 CASH DFB 4.044 1.816 3.006 3.567 4.644 161 CRD RISK 72.63 10.15 66.31 72.88 78.39 161 LOGATQ 8.336 1.360 7.300 7.795 8.930 161 CURR DEBT 7.271 5.234 3.049 6.253 10.51 161 HPI 0.0555 2.347 -0.317 0.0694 0.629 161 Panel B

Variables Mean Std. Dev. 25th_p. _Median ₇₅th_p. _{Nr. of Obs.}

% 𝚫 CRD RISK 3.836 7.996 -0.490 4.279 8.774 198 LEV RATIO 8.461 1.700 7.468 8.241 9.427 198 NIM 4.106 0.635 3.734 4.110 4.550 198 TOT DEPOSITS 73.29 8.586 66.85 74.19 79.58 198 CRE LOANS 3.560 2.483 1.819 2.894 4.914 198 CASH DFB 3.965 1.886 2.914 3.497 4.575 198 LOGATQ 8.415 1.359 7.357 7.862 9.148 198 M/B 18.41 7.528 14.07 16.68 21.29 198 CURR DEBT 7.312 5.126 3.070 6.481 10.46 198 HPI 0.0949 2.143 -0.327 0.0696 0.622 198

(25)

Table 2

Selected Additional Summary Statistics

Notes: Mean and Standard Deviation shown for 267 U.S. commercial banks, where all monetary values used for variable construction are nominal and in U.S.$ millions. All variables are measured as averages during the pre-crisis period, constructed with data starting January 1998 through December 2002. Summary statistics are shown for four different categories, with observations split along the median. SIZE measures the natural logarithm of total assets. CRD RISK captures credit risk, defined as Basel I risk weighted assets, as a proportion of total assets. RE LOANS is defined as real estate loans divided by total assets. CURR DEBT shows debt in current liabilities, normalized by total assets. LEV RATIO is the leverage ratio, estimated as common equity divided by total assets.

Mean Std. Dev. Mean Std. Dev.

Above median: SIZE Above median: RE LOANS

LEV RATIO 8.4674 1.8916 LEV RATIO 8.7717 2.0142

CRD RISK 74.5267 11.3052 Below median: RE LOANS

Below median: SIZE LEV RATIO 8.2959 1.7747

LEV_RATIO 8.6547 1.9564

Above median: CURR DEBT

CRD RISK 70.0994 8.4915

Above median: CRD RISK

LEV RATIO 8.1965 1.7692 CRD RISK 72.5927 10.7087

LEV RATIO 8.6383 1.9057 Below median: CURR DEBT

Below median: CRD RISK LEV RATIO 8.9277 2.0066

LEV RATIO 8.4617 1.9482 CRD RISK 72.1118 9.7980

5. EMPRICAL RESULTS

The results of the main regressions will be discussed in this part, elaborating the influence of pre-crisis capital in the three estimations outlined above as observed in the sample. First the consequences of increased capital will be estimated using the change in average market to book ratio during the crash, and afterwards on the change in a bank’s credit risk during the build-up of the recent crisis.

5.1 Crash Period Results

Table 3, column 1, contains the findings of the crash regression, where the average of the independent variable is estimated during the downswing from the second quarter of 2007 until the end of 2009. T- statistics are reported in parenthesis based on robust standard errors clustered by bank, to correct for heteroskadicity in the error terms.

(26)

Table 3

Main regression results (both Crash and Build-up)

Notes: All monetary values used for variable construction are nominal and in U.S.$ millions. Crash- variables are estimated from the third quarter of 2007 until December 2009, Build-up-variables use data from January 2004 until the second quarter of 2007, while Pre-Crisis values were constructed starting January 1998 through December 2003. Regression results shown for main crash period analysis in Column 1, from a selected set of bank fundamental data including 161 U.S. banks. The market value of a bank is estimated by multiplying shares outstanding with the quarterly closing price, and the market to book ratio is obtained by dividing the market value by total assets. The independent variable % Δ M/B CRASH denotes the percentage change in the market to book ratio during the crash, measured as the difference between the average ratio during the crash period and the average value over the pre-crisis period, divided by the average pre-crisis value and multiplied by one hundred. Column 2 reports regression results for main build-up period analysis, from a selected set of bank fundamental data including 198 U.S. banks. Credit risk (CRD RISK) is measured by Basel I risk- weighted assets, divided by total assets. Basel I risk-weighted assets are assets and off-balance-sheet items multiplied by the risk weight assigned by the Basle Committee on Banking Supervision, 1988. % Δ CRD RISK denotes the percentage change in the credit risk to assets ratio, measured as the average ratio during the upswing minus the average value over the pre-crisis period, divided by the average pre-crisis value and multiplied by one hundred. All following variables are independent measured as averages during the pre-crisis period. LEV RATIO is the leverage ratio, estimated as total common equity divided by total assets. NIM denotes the net interest margin of a bank, retrieved from Wharton WRDS COMPUSTAT. TOT DEPOSITS are a bank’s total deposits, normalized by total assets. CRE LOANS denotes a measure for commercial real estate loans as a proportion of total assets. CASH DFB is cash and funds due from banks divided by total assets. CRD RISK captures the Basel I risk-weighted assets as a proportion of total assets. SIZE measures the natural logarithm of total assets. CURR DEBT is the debt in current liabilities divided by total assets. HPI is the quarterly percentage change in the expanded state-level Housing Price Index, retrieved from the FHFA. M/B measures the market to book ratio of a bank. The regression uses robust standard errors, clustered by bank.

(1) (2)

VARIABLES Crash-Period

Dep Var: % 𝚫 M/B CRASH

Build-up

Dep Var: % 𝚫 CRD RISK

LEV RATIO -3.098* 0.852** (-1.930) (2.259) NIM -3.275 -0.697 (-0.701) (-0.567) TOT DEPOSITS 0.150 0.0559 (0.294) (0.509) CRE LOANS -2.734** 0.185 (-2.106) (0.763) CASH DFB 2.843* -0.209 (1.876) (-0.517) CRD RISK -0.237 (-0.760) LOGATQ -6.657*** -1.455*** (-2.835) (-2.768) CURR DEBT 0.540 0.276* (0.742) (1.749) HPI -1.235 -0.0327 (-1.507) (-0.185) M/B -0.0629 (-0.635) Constant 62.95 6.950 (1.118) (0.710) Observations 161 198 R-squared 0.147 0.107

Robust t-statistics in parentheses *** p<0.01, ** p<0.05, * p<0.1

(27)

STATA reports an R-squared of 14.67%, and the F-test shows significant explanatory power at the one percent level for the set of variables included. Most importantly, the results on the exogenous independent variable show that the average pre-crisis leverage ratio (LEV RATIO) seems to be associated with a sharper decline in the average market to book ratio (% Δ M/B CRASH), over the full crisis cycle. More specifically, an increase of one standard deviation in the leverage ratio is expected to create a 4.972 percent decrease in the average market to book ratio. These results are consistent with the first hypothesis, relatable to the concerns about capital regulation supporting the wrong incentives discussed in section 3. Nonetheless, they seem to stand in contrast to the finding of many other academics (e.g. Lee and Hsieh, 2013) indicating a beneficial effect of capital on profitability and risk taking, which should also translate into a positive link between the employed variables capital and market to book ratio. However, the distinctive difference in this estimation is the timing of the measurement, where prosperous times create abundant opportunities followed by the 2007 crash. Hence there is a prudent economic interpretation of the reported results affiliated with the theories used to derive the hypothesis. The results show some support for the claim that pre-crisis equity capital will create a larger drop in market to book ratio during the downswing. In the context of the theory, since pre-crisis capital creates additional possibilities and incentives to accumulate risk, the exposure gained during this prosperous build-up period will become a significant liability during the downswing, decreasing a bank’s market to book ratio (Martynova, Ratnovski and Vlahu, 2014). The effect the leverage ratio has on the change in post-peak market to book value is significant at the 10 percent level, with a p-value of 5.5 percent. When assessing the control variables in Table 4, the amount of CRE LOANS significantly impacts the market to book value in a negative way. Since commercial real estate exposure was a major driver of bankruptcy during the subprime crisis, the coefficient was expected to be negative (Cole and White, 2012). Next, there is some evidence for extremely liquid assets having a positive effect on the dependent variable, as measured by cash and funds due from other institutions (CASH DFB). Therefore, the results seem to show that larger cash holdings do not necessarily have an adverse effect on the risk attitude of bank managers by creating a moral hazard that leads to over-investing, as opposed to other non-financial corporations (Jensen, 1986; Malmendier and Tate, 2005). Finally, in accordance with the law of diminishing returns, the measure for SIZE shows a significant negative coefficient, since the larger a bank is to begin with, the more difficulties it will experience in increasing its market to book ratio.