Excerpt
Contents
1 Introduction
2 Background
2.1 Bank Capital Structure
2.2 The Financial Crisis
3 Data, Variables and Descriptive Statistics
4 Methodology and Results
4.1 Baseline Model
4.2 Model Modification and Crisis Indicators
4.3 Insights on Liabilities
5 Discussion.
6 Conclusion
Appendix 1: Banks’ Countries of Origin
Appendix 2: Further Insights on Bank Liabilities
References
Honorable Declaration
List of Tables and Figures
Table 1 Descriptive statistics I
Table 2 Descriptive statistics II
Table 3 Correlations I
Table 4 Literature of leverage determinants
Table 5 Time and country fixed effects, 1991-2004 dataset
Table 6 Time and bank fixed effects, 1991-2004 dataset
Table 7 Time and country fixed effects, 2004-2015 dataset
Table 8 Time and bank fixed effects, 2004-2015 dataset
Table 9 Crisis and pre-crisis interactions, country fixed effects .
Table 10 Crisis and pre-crisis interactions, bank fixed effects
Table 11 Pre-crisis period, country fixed effects
Table 12 Pre-crisis period, bank fixed effects
Table 13 Crisis period, country fixed effects
Table 14 Crisis period, bank fixed effects
Table 15 Banks’ countries of origin
Table 16 Correlations II
able 17 Effects on deposits
Table 18 Effects on non-deposit liabilities
Table 19 Effects on deposits, pre-crisis period
Table 20 Effects on deposits, crisis period
Table 21 Effects on non-deposit liabilities, pre-crisis period
Table 22 Effects on non-deposit liabilities, crisis period
Figure 1 Equity and decomposed liabilities
Figure 2 Equity and decomposed liabilities
List of Abbreviations
Abbildung in dieser Leseprobe nicht enthalten
Abstract
As the last global financial crisis has shown, macroeconomic shocks have huge im- pacts on numerous economic sectors. The distortions caused by the crisis led to world- wide declines in trade, output and employment. This paper takes special emphasis on bank leverage ratios. Our longitudinal sample ranges from 2004 to 2015, including 5,069 observations of 483 banks in the United States and Europe. Our investigations are based on a modified Gropp and Heider (2010) style linear two-way error compo- nent model that uses leverage determinants of non-financial firms to describe bank capital structure. It is investigated how bank capital structure and its determinants re- spond to the financial crisis. The results show that the financial crisis had strong im- pacts.
1 Introduction
The last global financial crisis started in the banking sector and rapidly emerged to a global recession in numerous economic sectors. In the crisis year of 2008 the MSCI World index fell by ~42%, the S&P 500 index experienced a decline of ~39%, similar the Euro Stoxx 50 index fell by ~44% (Bloomberg Online, 2016).
Against this background, it would be interesting to examine how the financial crisis impacted bank capital structure. Since the irrelevance proposition of Modigliani and Miller (1958), which states that under perfect market conditions the value of a firm is unaffected by its financing, firms’ capital structure decisions and related determinants have been a widely discussed topic in corporate finance literature. However, it is often argued that it is unnecessary to investigate bank capital structure determinates as deviations from the irrelevance proposition are dominated by capital regulations (Mish- kin, 2000, 2007). Nonetheless, Gropp and Heider (2010) find that capital structure determinants of non-financial firms are able to explain variations in bank capital struc- tures and that regulation only plays a subordinate role.
Our paper investigates how bank capital structure and its determinants are affected by the last global financial crisis.
Our main sample contains 483 listed banks in the United States and Europe with av- erage total assets of more than one billion USD. The sample starts in 2004 and ends in 2015. Our linear two-way error component model is an extension of the Gropp and Heider (2010) style model. It is based on non-financial firms’ capital structure determi- nants and crisis and pre-crisis indicating variables. Our depended variable is either book or market leverage. Further, our model contains interaction terms between crisis and pre-crisis indicating variables and the other explanatory variables. For some spec- ifications shown in appendix 2 we use book or market deposits and non-deposit liabilities as depended variable.
To begin with, our results in general support the findings of Gropp and Heider (2010) that non-financial firms’ capital structure determinants are able to explain bank leverage variations. Our first crisis related finding is a leverage increase during the precrisis period which is shown by significant positive effects of the pre-crisis dummy in our model. Further, our results support the findings of Gropp and Heider (2010) that huge fractions of variations in book leverage are due to unobserved bank fixed effects. However, we find that the explanatory power of these effects is comparably small for market leverage. Against the background of the financial crisis, we find that the fraction of variation in book and market leverage that is explained by bank fixed effects is at its greatest during the crisis period. In our opinion this is due to the fact that the financial crisis affects individual banks differently. Therefore, each bank adjusts its capital structure strategy according to its individual situation. Consequently, the explanatory power of bank fixed effects increases. In contrast, the fraction of variance in market and book deposits and non-deposit liabilities that is explained by bank fixed effects is at its greatest during the pre-crisis period. Our results further show, that negative leverage effects coming from asset risk are weaker during the pre-crisis period. This could be an indication that bank debt holders were less risk averse during the years prior to the crisis. Further, a shift in the decomposition of bank liabilities is found. Namely, fractions of bank non-deposit liabilities decline while fractions of bank deposits increase during the crisis period. This shift can be explained by banks substituting risky with less risky assets during the crisis period.
The remainder of this paper is structured as follows. Section 2 provides a literature review on capital structure theories, bank capital structure and the financial crisis. Section 3 describes our dataset, variables and correlations between our variables. Section 4 presents our regressions’ results. Section 5 discusses our main findings. Section 6 concludes.
2 Background
A good starting point for the purpose of this thesis is found in the capital structure literature of non-financial firms and its most prominent theory, the irrelevance proposition of Modigliani and Miller (1958). It has been the first capital structure theory that has been widely known and accepted. The model assumes perfect market conditions without taxes or other disruptions. In the model firms divide cash flows among investors dependent on the portions of equity and debt in their financing structure. The investors can further create any level of leverage in excess or below of what level of leverage is offered by the firm. Therefore, a firm’s enterprise value is independent from its capital structure and a levered firm has the same value as an unlevered firm (Modigliani and Miller, 1958, 1963). However, in reality the theorem fails due to many deviating factors like transaction costs, taxes, agency conflicts, asymmetric information, bankruptcy cost or the non-separability between operations and financing. Therefore, one could argue that these results are more or less irrelevant. However, even though perfect market conditions do not exist in reality, the theorem helps to understand capital structure decisions. For instance, it is an arguable point that every time the capital structure of a firm is relevant it must be due to at least one deviation from the theorems assumptions. Therefore it makes sense to investigate on factors leading to deviations (Luigi and Sorin, 2009; Miller, 1988).
Consequently, capital structure theories which diverge from perfect capital markets assumptions in Modigliani and Miller (1958) emerged over time. Nonetheless, a universal theory of capital structure choice is nonexistent (Myers, 2001). The three most important theories are firstly, the trade-off theory which assumes that firms trade off costs and benefits of equity and debt financing under imperfect market conditions to find an appropriate capital structure (Fama and French, 2002). Secondly, the peckingorder theory which assumes that firms minimize information asymmetry between managers and shareholders by following a financing hierarchy (Leary et al., 2010; ShyamSunder and Myers, 1999; Myers, 1984; Myers and Majluf, 1984). Thirdly, the markettiming theory, which is the most recent of the three theories. It goes back to Baker and Wurgler (2002). The theory assumes that the present capital structure is determined by efforts to time the equity market in the past (Luigi and Sorin, 2009).
The following sections purpose is to provide further insights into the literature of capital structure theories starting with the irrelevance proposition, followed by the trade-off and pecking-order and finally the market-timing theory.
The Trade-off Theory
The trade-off theory describes a family of related theories that have in common that a firm’s management evaluates costs and benefits of different capital structure decisions (Luigi and Sorin, 2009; Frank and Goyal, 2007). The initial trade-off theory comes from Modigliani and Miller (1963) when they added tax costs to their model. This produced debt benefits in form of an income tax shield. This implied a sole debt financing as there have been no offsetting costs of debt like bankruptcy costs. Bankruptcy costs were later added by Kraus and Litzenberger (1973). They find a trade-off between tax benefits and higher expected bankruptcy costs coming from higher debt levels (Frank and Goyal, 2007). Myers (1984) argues that firms set leverage ratio goals by balancing these costs and benefits. Frank and Goyal (2007) further separate the theory into a static and a dynamic part. The static trade-off theory is a one period capital structure choice model going back to Bradley et al (1984). The model captures tax benefit and bankruptcy costs, agency costs of debt arguments and the possible loss of tax shields on non-debt in non-default states (Bradley et al., 1984). The model shows, that there is an inverse relationship between the optimal leverage ratio to the expected costs of financial distress and to the level of non-debt tax shields. Further, Bradley et al (1984) show that if there are significant costs of financial distress than the leverage ratio is inversely related to the variations of earnings. Frank and Goyal (2007) complain that the model’s main elements cannot be observed directly. Further, they argue that supplementary assumptions are needed to be able to applicate the theory on real world data. For this reason several modifications of the model are found in literature. Such modifications could include product market competition, corporate governance, asset substitution, underinvestment, risk shifting or inclusions of the free cash flow hypothesis (Frydenberg, 2011). However, all static models have in common that a comparison of actual with target leverage determines capital structure decisions for each period.
Graham and Leary (2011) argue that for models that develop an optimal dynamic strategy this does not need to always be the case. Literature shows that models accounting for adjustment costs are able to explain deviations and inconsistencies of leverage characteristics that occur in the static trade-off model. Dynamic models based on adjustment costs imply that a firm’s actual capital structure does not permanently match the firm’s target leverage ratio (Dudley, 2007). The dynamic model of Fischer et al. (1989) assumes that firms trade off tax benefits and financial distress costs like in the static models. But, in contrast to the static view the model leads to a leverage range instead of a fixed target. This is due to the model’s assumption that firms are exposed to annual asset value variations and costs of recapitalization (Graham and Leary, 2011; Fischer et al., 1989). This further implies that firms do not adjust the capital ratio as long as leverage stays within that range. Firms therefore only adjust their capital structure if the leverage ratio crosses an upper or lower recapitalization threshold. Others, like Bhamra et al. (2010) include adjustment costs when examining the impact of time-varying macroeconomic conditions on firms’ capital structures. In other dynamic models the firms’ investment and cash flow decisions are endogenous. For example, Hennessey and Whited (2005) develop a model with firms having endogenous investment, leverage and distribution choice while facing several cost factors like taxes, financial distress costs and equity flotation costs. Their study is able to explain several inconsistencies of empirical findings based on the static trade-off theory.
The Pecking-order Theory
An alternative view to the trade-off theory is the pecking-order theory. The first suggestion of the theory comes from Donaldson (1961). However, Myers (1984) and Myers and Majluf (1984) modified the theory which led to the most common model. They assume that firms have a financing hierarchy which is implied by information asymmetry. Firms’ leverage ratios are determined by costs of capital and access to financing. These costs increase with rising information asymmetry. Therefore, firms prefers internal over external financing. If external financing is required the firm prefers debt over equity as equity implies the highest information asymmetry (Myers, 1984). Tong and Green (2005) state that the pecking-order theory implies that there is an optimal capital structure at any point in time and that net cash flows are an important capital structure determinant as net cash flows determine the amount of funds available. However, they argue that there is no sole optimal capital structure that firms target in the long run as in the (static) trade-off theory. Helwege and Liang (1996) test the peckingorder theory by examining the financing decisions of firms after their IPO. Their results show that the pecking-order theory is not followed by firms that enter the capital market when deciding which type of securities they are going to offer. It is further indicated by their results, that the probability of using external funding is not related to shortfalls of internal funding. Further, indications are found for a preference of internal over external financing by firms that have cash surpluses (Graham and Leary, 2011). Frank and Goyal (2003) argue that small size and high growth rates are related to higher information asymmetries. Consequently, such firms should have higher incentives to follow the theory. However, Fama and French (2005) find equity financing to be common especially for high growth firms and firms of small size. But, Lemmon and Zender (2010) state that on the one hand small firms are indeed likely to face higher levels of asymmetric information but that they are also likely to have higher relative growth opportunities. Further they show, that these firms have very restrictive constraints on debt capacity. Based on these findings they argue that one cannot necessarily conclude that the above mentioned studies do stand in contrast to the theory. Nevertheless, also others like Leary and Roberts (2010) find that the theory does not necessary predict issuance decisions as expected. Their findings are even supported for subsamples in which the theory is strongly expected to hold. Halov and Heider (2011) suggest the standard model to be a special case of an adverse selection problems. They argue that when there is adverse selection about enterprise value, the firms would prefer debt over equity financing. But, if there is asymmetric information about risk, the adverse selection arguments would apply for debt and therefore the fi rm would prefer equity over debt financing.
The Market-timing Theory
The market-timing theory challenges the previous two theories. The theory argues that conditional on the financing needs, firms use debt financing when they perceive the relative equity costs as high and use equity financing when they perceive the relative equity costs as low. Therefore, firms repurchase own shares when they perceive an undervaluation and issue stocks when they perceive an overvaluation of their shares (Luigi and Sorin, 2009). Baker and Wurgler (2002) show high persistency of the influence of market-timing and claim it as first order determinant of firms’ capital structure choice. Following Baker and Wurgler (2002), there is a lot of evidence for markettiming in the real economy. They argue that investigations of actual financing decisions show that on average market-timing is successful. Moreover, they argue that managers admit to time the market in several surveys. They further find, that firms tend to issue equity when investors seem to be too enthusiastic about prospected earnings. Huang and Ritter (2005) explain that firms’ decision makers are able to assess the relative costs of debt and equity by knowing themselves and their industry better and due to the possibility that they follow specific psychological patterns. The survey of Graham and Harvey (2001) shows that equity market prices are regarded as more important than several other factors when firms decide to either issue common shares or convertible debt. Of the questioned CFOs two-thirds say that when issuing equity the amount of overor undervaluation of their shares is an important consideration. Support for Baker and Wurgler (20002) comes from Chang et al. (2006) who find that market-timing has a huge explanatory power for capital structure decisions of small companies.
However, the empirical evidence for the market-timing theory is still mixed, especially when considering the persistency of effects coming from market-timing. For instance Alti (2006) states that most market-timing tests are based on the positive relationship of a firm’s valuation and its decision to issue equity. However, Alti (2006) argues that there is a variety of capital structure determinants that are likely to lead to that relationship. As an example he states that typically high market-to-book values are usual for firms that have growth opportunities. Alti (2006) further argues that such firms may prefer equity financing over debt financing to maintain flexibility. By examining firms’ capital structures around their IPO he found evidence for the market-timing theory in the short run. But also, that market-timing effects completely disappear during the post-IPO years. This is supported by Leary and Robert (2005) who argue that if the adjustment of the capital structure is costly there should be a weaker long term markettiming effect. Moreover, Kayhan and Titman (2007) find that leverage is widely determined by the average of the market-to-book ratios itself and not by its market-timing component.
The literature above is related to non-financial firms. However, banks are different to other firms as they function as financial intermediaries. Banks bear the cost of capital regulation on the one hand but are able to issue federally insured debt on the other hand (Harding et al., 2009).
2.1 Bank Capital Structure
Literature on capital structure decisions is in general based on non-financial firms. Diamond and Rajan (2000) ask whether bank capital structure matters at all. They refer to Gorton and Pennacchi (1990) arguing that banks do not have to face issuance costs based on asymmetric information. Diamond and Rajan (2000) find that bank capital decisions are important as they have an effect on a bank’s safety, its capability to refinance, its capability to receive borrowers’ repayment and its willingness to liquidate the repayments.
Gropp and Heider (2010) further state that the standard textbook’s opinion would be that it is unnecessary to investigate on the capital structure of banks as deviations from Modigliani and Miller are dominated by capital regulations (Mishkin, 2000; Mishkin, 2007). Indeed, large parts of finance literature assume that especially capital requirements determine banks’ capital ratios (e.g. Jackson et al. 1999). However, the results of Gropp and Heider (2010) stand in contrast to these findings as they indicate that regulation is only a subordinate bank capital structure determinant.
There are studies examining the effects on bank capital coming from risk (Shrieves and Dahl, 1992), asset risk in the environment of regulatory innovations (Flannery and Rangan, 2008), liquidity (Myers and Rajan, 1995), market discipline (Allen et al. 2011), tax effects (Schepens, 2016) or the operating environment (De Jonghe and Ötzekin, 2015; Brewer et al, 2008). Others, like Harding et al. (2007) find evidence that in the presence of deposit insurance, a fixed capital standard and taxes, there is an interior optimal bank capital ratio.
The starting point of our investigation is the paper of Gropp and Heider (2010) which examines the determinants of bank capital structure. They use a linear regression model consisting of non-financial firm capital structure determinants. They use a sample of the 200 biggest banks in the United States and Europe for the timeframe of 1991 to 2004.
Gropp and Heider (2010) investigate whether regulation really withstands as first order capital structure determinant. They argue that if regulation is a first order determinant this would imply that there is no cross-sectional variance in the capital structure of banks that fall under the same regulatory regime. They find this not to be the case as they find a large variation in leverage ratios of the banks in their sample. One explanation for these findings could be found in the buffer view. The theory is based on the findings that banks hold capital levels much higher than required by regulation (e.g. Brewer et al., 2008; Allen et al., 2010). The buffer view assumes that the issuance of equity implies costs due to asymmetric information. It could be that a bank has to bear such costs unexpectedly when the equity ratio falls below the regulatory capital requirements. Therefore, banks are expected to hold equity buffers in excess to avoid these costs, see Myers and Majluf (1984). The results of Gropp and Heider (2010) are not in line with the predictions coming from the buffer view. However, Gropp and Heider (2010) find that leverage determinants of non-financial firms are able to explain variations in bank capital structure. A comparison with Frank and Goyal (2003) and Rajan and Zingales (1995), who investigate non-financial firms’ capital structures, shows that the effects found by Gropp and Heider (2010) are identical in sign and significance. Further, they find that unobserved time-invariant bank fixed effects have a strong influence on bank capital structure as they explain huge fractions of the variations in bank leverage. These findings are in line with Lemmon et al. (2008) for nonfinancial firms. Based on their results they declare bank fixed effects as first order bank capital structure determinant.
2.2 The Financial Crisis
As mentioned above, testing of capital structure theories leads to ambiguous results. One explanation for such differences comes from Zhang and Mirza (2015). They argue that capital structure theories are not reliable under different economic conditions. Therefore, it is interesting to investigate how an economic shock like a global financial crisis could influence firm and bank capital structure and their determinants. Against this background the last financial crisis raises an interesting research area.
The last global financial crisis started in the financial sector and rapidly emerged to a global recession in numerous economic sectors along with declines in output, trade and employment, a great reduction of security issuance of firms and lending by financial institutions (Fosberg, 2012; Ahn et al., 2011; Milesi-Ferretti and Tille, 2011; Hobza et al., 2009). Since the crisis and its distortions, many researchers like Admati (2016) claim that bank capital requirements are insufficient.
It is often argued, that the lending boom in the years prior to the crisis led firms to increase their leverage ratios to exceptional levels. After turning into recession, accompanied by a strong decrease in leverage and contraction of banks’ credit commitments the economic collapse evoke (Thakor, 2015; Fahri and Tirole, 2011; Fostel and Geanakoplos, 2008).
Chen et al. (2014) find too high leverage and too low collateralization, over-reliance in short term debt and inability to raise capital as main drivers for the Lehman Brothers default. Mendoza and Terrones (2008) argue that increased fragility in the banking sector is caused by excessively increased leverage ratios and that this situation has been associated with economic and financial crises in the past. Supporting, Geanakoplos (2009) sees the financial crisis as the bottom of what he calls the leverage cycle, in which leverage gradually increases too high with as sudden fall to a too low level. Support comes from Demirguc-Kunt et al. (2015) who find that during and after the crisis leverage declined in advanced and developing countries but also in countries that did not experience a crisis. The strongest decrease in leverage they find for privately held, small companies in countries with less efficient legal systems, shallower banking systems, weaker information-sharing and stronger entry barriers for banks. For listed companies they find least significant evidence regarding decreases in leverage ratios. They argue that these firms are in general bigger and therefore have easier access to capital. Similarly, Fosberg (2012) finds a strong significant increase in debt in the pre-crisis period. He finds, that this increases were reversed rapidly after the worst part of the crisis has been passed as by the end of 2010 when the sample firms showed a capital structure close to the pre-crisis period. Graham et al. (2015) find an explanation for declining leverage ratios during crises periods in rare investment opportunities. They argue that due to rare investment opportunities there is no necessary demand for external debt financing what consequently leads to lower leverage levels. The study of Iqbal and Kume (2014) finds an overall leverage ratio increase during the pre-crisis period as well as in the crisis period followed by a decrease in leverage ratios in the post-crisis period for UK, French and German firms. Additionally, they find that firms that had a lower leverage level as their industry peers in the pre-crisis period experience a steady increase in their leverage ratio in the after crisis period. The opposite they find for firms that had higher leverage levels compared to their peers in the pre-crisis period.
The study of Zhang and Mirza (2015) shows that there has been a change of capital structure determinants, both on the firm and the macroeconomic level after the financial crisis for Chinese listed non-financial firms.
Harrison & Widjaja (2014) find that the financial crisis significantly impacted firms’ capital structure decisions in the United States. Campello et al. (2010) find that the reduction in demand and cash flows during a crisis leads to perceived increases in costs of funding. Therefore, they argue that there has been no additional funding demand during the crisis. Kahle and Stulz (2013) claim that the last financial crisis lead to an increase in cash holdings and a fall in debt and equity issuance due to restrained investments. They further explain that firms in general used debt financing via bond markets especially during the first part of the crisis, however, they argue that debt financing decreased in 2008.
The empirical work of Papanikolaou (2013) shows that banks increased leverage on and off their balance sheets prior to the crisis. It further shows, that banks decreased leverage when they were disposing risky assets during the crisis.
Our linear two-way error component model is an extension of the Gropp and Heider (2010) style model. It is based on non-financial firms’ capital structure determinants and crisis and pre-crisis indicators. Further, it contains interaction terms between the indicator variables and the other explanatory variables. Several robustness modifications are made. We use panel data. Our main sample contains 483 banks in the United States and the European Union for the timeframe beginning in 2004 and ending in 2015. Banks in the sample have average total assets of more than 1 billion USD. The years 2007 and 2008 are considered to be the pre-crisis years. The years 2008 and 2009 are considered to be crisis years.
The previous sections leads to the hypothesis that banks like non-financial firms follow the leverage cycle described in Geanakoplos, J. (2010).
Therefore, we assume our results to show that the pre-crisis period has a significant positive effect on banks’ book and market leverage. In contrast, there should be a negative response of bank leverage to the crisis period.
It is further investigated whether the effects on book and market leverage of non-bank capital structure determinates change during the crisis periods. Moreover, the decomposition of bank liabilities is investigated against the background of the crisis.
3 Data, Variables and Descriptive Statistics
We use consolidated balance sheet and income statement data from the Bankscope database of the Bureau van Dijk and data on stock volatility and market values from the Thompson Financial’s Datastream database. We focus on commercial banks, therefore, central banks, Islamic banks and other finance companies included in the bank scope dataset are dropped. Further, banks having their reporting date between April and September are dropped. To get unified yearly observations, data published between January and March is considered to belong to the previous year. The sample used for our main analyses starts in the year 2004 and ends in year 2015. A second sample is used to compare the outcomes of our baseline to Gropp and Heider (2010). Therefore, this second sample starts in year 1991 and ends in year 2004. The sample ranging from 1991-2004 contains 260 listed banks in the United States and 14 European countries, the sample ranging from 2004-2015 contains 483 listed banks in the United States and 21 European countries (see appendix 1). The samples only include banks with average total assets of at least one billion USD and banks that are still existent. The reason to drop out banks that do not exist anymore is to diminish variations coming from financial distress for regressions without bank fixed effects. The 2004-2015 sample consists of 5,069 bank year observations, the 1991-2004 sample consists of 2,603 bank year observations (see appendix 1). Currencies and units have been unified to USD and amounts in millions. The years 2008 and 2009 are defined as crisis years and the years 2006 and 2007 as pre-crisis years (see: Demirguc-Kunt et al., 2015; Pattani et al., 2011 and Iqbal and Kume, 2014).
As the following graph shows, there has been a significant change in the relative composition of book capital. Figure 1 shows an increase in the fraction of non-deposit liabilities (long and short term debt) until the end of 2008 followed by a sharp decline in 2009. The following years, non-deposit liabilities further decline gradually. In contrast, the evolution of deposits goes in the opposite direction. Further, the fraction of book equity becomes slightly larger during and after the crisis years.
Figure 1 Equity and decomposed liabilities
The sample consists of European and U.S. banks having average total assets higher than one billion USD for the timeframe 2004-2015. Data coming from Bankscope and Thompson Financial’s Datastream. This figure shows the fractional composition of total book assets by book equity, book deposits and book debt.
Abbildung in dieser Leseprobe nicht enthalten
The variables used in the regression models are shown in the following descriptive statistics tables. All data has been winsorized at the 1% level.
Average total assets amount 60 billion USD in the 1991-2004 sample and 110 billion USD in the more recent sample. The sample for the latter timeframe shows on average lower market-to-book ratios, asset risks, profits, collateral and dividend pay-out ratios. However, they are not comparable due to different datasets. The average book leverage (market leverage) amounts 91% (86%) percentage for the 1991-2004 timeframe and 89% (88%) for the 2004-2015 timeframe (see table I. and table II.).
Table 1 Descriptive statistics I.
The sample consists of European and U.S. banks having average total assets higher than one billion USD for the timeframe 1991-2004. Data coming from Bankscope and Thompson Financial’s Datastream.
Abbildung in dieser Leseprobe nicht enthalten
Table 2 Descriptive statistics II.
The sample consists of European and U.S. banks having average total assets higher than one billion USD for the timeframe 2004-2015. Data coming from Bankscope and Thompson Financial’s Datastream.
Abbildung in dieser Leseprobe nicht enthalten
Table 3 Correlations I.
The sample consists of European and U.S. banks having average total assets higher than one billion USD for the timeframe 2004-2015. Data coming from Bankscope and Thompson Financial’s Datastream.
Abbildung in dieser Leseprobe nicht enthalten
Table 3 Continued
The calculation of variables is based on the paper of Gropp and Heider (2010) and Frank and Goyal (2004, 2009).
Book leverage = 1 (book value of equity / book value of total assets) Market leverage = 1 (market cap / (market value of equity + book value of liabilities))
Size = book value of total assets
Profits = (pre-tax profits + interest expenses) / book value of total assets Market-to-book ratio = market value of assets / book value of assets
Collateral = (fixed assets + reverse repos and cash collateral + trading securities + derivatives + available for sale securities + held to maturity securities + at equity investments + money market instruments + CDs and other deposits + other securities) / book value of total assets
Dividend dummy = 1 for each year in which a bank pays a dividend
Asset risk = annualized standard deviation of daily stock price returns * (market cap / (market cap + market value of liabilities))
Pre-crisis dummy = 1 for the years 2006 and 2007
Crisis dummy = 1 for the years 2008 and 2009
Book deposits = total deposits / book value of assets
Market deposits = total deposits / (market cap + market value of liabilities)
Book non-deposit liabilities = book leverage – book deposits
Market non-deposit liabilities = market leverage – market deposits
4 Methodology and Results
Following Frank and Goyal (2009) the finance literature on capital structure theories often refers to Harris and Raviv (1991) or Titman and Wessel (1988) whose results stand in contrast to each other. Harris and Raviv (1991) state that the available studies on capital structure in general agree that leverage increases with growth opportunities, fixed assets, firm size and non-debt tax shield. Further, they state that it is generally agreed on that with increased volatility, profitability, expenditure in advertising and research and development, probability of bankruptcy and the uniqueness of a product, leverage is expected to decrease. The results of Titman and Wessels (1988) disagree, as they do not show effects of collateral, non-debt tax shields, volatility and growth opportunities on debt ratios. Therefore, Frank and Goyal (2009, p.2) state that depending on the underlying theory of capital structure (see section 2) there are “diametrically opposing well-known summaries of “what we all know” from the previous literature”. Frank and Goyal (2009) therefore examine a large set of factors in literature that have a consistent interaction with leverage. In industry leverage, market-to-book ratio, asset tangibility, profits and size they find the most promising leverage determinants. In a previous paper they further find dividends to be a reliable leverage de terminant. Since Gropp and Heider (2010) further confirm these variables and asset risk to also be solid bank capital structure determinants we use these variables as basis for our model. Table 4 summarizes the predictions of the signs of the variables effects on leverage following the pecking-order theory, the trade-off theory, the market-timing theory and the buffer view. Further, signs of significant variables’ effects on leverage found in Gropp and Heider (2010), Frank and Goyal (2009) and Rajan and Zingales (1995) are shown.
Table 4 Literature of leverage determinants
+/denotes the sign of significant variables’ effects on leverage as the depended variable. P OT = pecking-order theory; TOT= trade-off theory; MTT = market-timing theory; BV = Gropp and Heider (2010) predictions for the buffer view; G&H (2010) = Gropp and Heider (2010); F&G (2009) = Frank and Goyal (2009); R&Z (1995) = Rajan and Zingales (1995).
Abbildung in dieser Leseprobe nicht enthalten
[...]