Credit Risk Securitization and Banking Stability

Evidence from the Micro-Level for Europe This draft: April 15, 2009 Tobias Michalak a Andr Uhde b *

Abstract: Using a unique sample of 743 cash and synthetic securitization transactions issued by 55 stock listed bank holdings in Western Europe plus Switzerland over the period from 1997 to 2007 this paper provides empirical evidence that credit risk securitization has a negative impact on the banks financial soundness as measured by the z-score technique while controlling for macroeconomic, bank-specific, regulatory and institutional factors. Moreover, as a result of further robustness checks we find a positive impact of credit risk securitization on the banks leverage and return volatility as well as a negative relationship between securitization and the banks profitability. Keywords: Credit risk securitization, Banking stability, European banking JEL classification: G21; G28

a b

Tobias Michalak, University of Bochum, Department of Economics, 44780 Bochum, Germany. Fax: ++49 234 32 05345, email: tobias.michalak@rub.de. Dr. Andr Uhde *(corresponding author), University of Bochum, Department of Economics, 44780 Bochum, Germany. Fax: ++49 234 32 02278, email: andre.uhde@rub.de. We thank Oliver Mueller and Carina Trimborn for thoughtful and helpful comments.

Introduction Worldwide and in particular in Europe the market for credit risk transfer has experienced a

remarkable growth in recent years. This refers not only to the volume of credit risk being transferred by financial institutions but also to the total number of securitization transactions. On a micro-level, the growing popularity of credit risk securitization can be put down to the fact that banks typically use the instrument of securitization to diversify concentrated credit risk exposures and to explore an alternative source of funding by realizing regulatory arbitrage and liquidity improvements when selling securitization transactions. On a macro-level, credit risk securitization is recommended since it is anticipated as a reduction of the overall concentration of credit risk in the entire financial system if risks are transferred to less fragile (non-financial) institutions (BIS, 2005; ECB, 2004; IMF, 2002). In response to the U.S. subprime mortgage crisis from mid-2007, however, a general reassessment of risks inherent in structured finance instruments is observed across the whole financial community. To date, the IMF values mark to market losses on structured finance instruments at approximately 1,405 billion USD (IMF, 2008). Moreover, it is assumed that about half of the amount of losses and write-downs on these instruments will explicitly affect the banking industry (IMF, 2008), probably constituting a serious threat of systemic fragility. The latter is supported by failures in valuating complex securitization instruments, a weak transparency in structured finance markets as well as weak forces of market discipline, which in sum have exposed the financial system to a serious funding and confidence crisis (BIS, 2008, 2008a; IMF, 2008, 2008a, 2007). Referring to these findings, the Basel Committee has recently finalized its proposals for enhancing the Basel II framework in the area of securitization (BCBS, 2009). The proposals aim at strengthening the framework and responding to lessons learned from the financial crisis. In particular, proposals mainly focus on (a) higher risk weights to securitization exposures and hence higher minimum capital standards (Pillar 1), (b) addressing the banks on- and off-balance sheet

securitizations within the frameworks of ICAAP and SREP as well as the banks MIS to enhance the banks and supervisors sensitivity to securitization (Pillar 2), and (c) strengthening disclosure requirements with regard to securitization activities and off-balance sheet vehicles to enhance transparency (Pillar3). Against this background this paper empirically investigates the impact of credit risk securitization on banking stability using a unique sample of 743 cash and synthetic securitization transactions issued by 55 stock listed bank holdings located in Western Europe plus Switzerland over the period from 1997 to 2007. Our analysis complements and extends previous empirical studies on this issue (Jiangli and Pritsker, 2008; Uzun and Webb, 2007; Dionne and Harchaoui, 2003) for several specific aspects. First, to the best of our knowledge this is the first study that empirically investigates the relationship between credit risk securitization and banking stability using a crosssectional time-series dataset for European banks. Second, while previous studies employ (regulatory) capital ratios (Uzun and Webb, 2007; Dionne and Harchaoui, 2003) or time deposit premiums (Jiangli and Pritsker, 2008) as respective proxies for the banks financial soundness, we complement empirical work by utilizing the z-score ratio as a time-variant distance to default measure. Third, by investigating the impact of credit risk securitization on single components of the z-score ratio (ROAA, capital ratio, volatility of ROAA), we try to shed more light on the nexus between credit risk securitization and banking stability. Finally, we extend previous empirical studies by performing a large variety of sensitivity analyses controlling for banking and capital market structure developments as well as the regulatory and institutional environment in Western Europe. The remainder of the paper is organized as follows. Section 2 presents related theoretical and empirical literature on the relationship between credit risk securitization and banking stability. Section 3 comprises our empirical analysis. While section 3.1 presents the data set, section 3.2 describes our empirical model. Empirical results are presented and discussed in section 3.3 and illustrated within the statistical appendix. Finally, section 4 concludes.

Related literature Economic theory provides countervailing predictions of the relationship between credit risk

securitization and banking stability (Shin, 2009; Krahnen and Wilde, 2008; Jiangli et al., 2007). This may be due to the fact that the relationship depends on both a direct and indirect impact. The direct impact of securitization on banking stability depends on how much credit risk is actually transferred to external investors. This relationship however is not distinct. While the securitizationstability view points out that the banks overall risk exposure is likely to be reduced if the tail risk of senior tranches being transferred to external investors exceeds the sum of default risks of the first-loss position which is typically retained by the bank (Jiangli et al., 2007), the securitizationfragility view replies that the major part of default risks typically remains within the banks firstloss position acting as a quality signal towards external investors (DeMarzo, 2005; Instefjord, 2005; Riddiough, 1997; Greenbaum and Thakor, 1987). In this context, it is additionally emphasized that former Basel I regulations have provided an incentive to keep the larger part of risks within the bank. Thus, as corporate and retail loans were not risk-adjusted but globally backed up with regulatory capital under Basel I regulations, keeping the major part of default risks within the firstloss piece typically provoked profits from regulatory arbitrage (Allen and Gale, 2006). The indirect impact of credit risk securitization on financial stability is determined by the banks strategy to utilize securitization as a source of additional funding to finance new assets with liquid capital that has become available from selling securitization transactions. Thus, the indirect effect of securitization is not obvious but rather depends on a wide range of investment policies and can more probably be defined by the way the banks overall asset portfolio risk is restructured (Krahnen and Wilde, 2008). In this context the securitization-stability view points out that reinvesting liquid capital into new assets may provoke a better diversification of the banks asset portfolio if remaining total assets are less correlated after securitization (Cebenoyan and Strahan, 2004; Demsetz, 2000). In contrast, the securitization-fragility view suggests that the actual effect on a banks financial soundness may depend on the risk-level of new assets being taken in, which again

is determined by the current level of competition in the respective asset market (Instefjord, 2005). Moreover, using liquid capital to extend the amount of total assets or to repurchase shares and pay higher dividends to shareholders may additionally lead to an increase in the banks leverage (Shin, 2009; Leland, 2007). Empirical evidence on the relationship between securitization and banking stability is ambiguous as well. To begin with, applying event study methodology Uhde and Michalak (2009), Hnsel and Krahnen (2007), Franke and Krahnen (2006) as well as Lockwood et al. (1996) provide empirical evidence that credit risk securitization has a positive impact on the increase of a banks systematic risk. This result holds even when controlling for the banks pre-event level of systematic risk, the type of securitization transaction, the regulatory framework as well as the underlying reference portfolio (Uhde and Michalak, 2009). Turning to panel data analysis, using balance sheet data from commercial banks in Canada for the period from 1988 to 1998, Dionne and Harchaoui (2003) find that credit risk transfer is inversely related to a banks regulatory capital supporting the capital arbitrage theory. Moreover, they provide empirical evidence that an increase in the volume of credit risk transfer has a negative impact on the banks asset quality and hence financial soundness. Similarly, Uzun and Webb (2007) examine the impact of credit risk securitization on banking stability using data from a sample of 112 financial institutions in the U.S. for the period from 2001 to 2005. They find that securitization is negatively related to a banks capital environment. Controlling for underlying assets they provide further empirical evidence that the decrease in financial soundness is predominately associated with securitizations of credit card receivables whereas securitizations of mortgage loans and home equity lines of credits have a positive impact on banking stability. Finally, Jiangli and Pritsker (2008) examine the effect of mortgage loan securitizations on bank stability, profitability and leverage using data from U.S. bank holding companies for the period from 2001 to 2007. In line with Uzun and Webb (2007) they find that mortgage securitizations tend

to reduce a banks financial fragility. In contrast, however, they also provide empirical evidence for a positive relationship between securitization and a banks leverage whereas profitability tends to increase due to securitization. 3 3.1 Empirical analysis Data and sources

Notes on variables and data sources are presented in Table 1. Table 2 reports descriptive statistics for the entire set of included variables. Correlation matrices are provided in Tables 9-12. Our empirical analysis focuses on consolidated balance sheet data from 55 stock listed bank holdings across the EU-11 1 plus Switzerland for the period from 1997 to 2007 following the beginning of credit risk securitizations in Europe in 1997. Banks consolidated balance sheet data are retrieved from BankScope database provided by Bureau van Dijk. Table 3 reports the geographical distribution of European banks in our sample. A. Banking stability We employ the banks distance to default as a proxy for financial soundness by including the z-score as our dependent variable. This ratio has become a popular measure of bank soundness in related empirical work on financial stability (Boyd and Runkle, 1993; De Nicol et al., 2004; Uhde and Heimeshoff, 2009) and is denoted as follows:

+k

(1)

We construct the z-score per bank holding and time by aggregating the banks consolidated balance sheet data and define as the return on average assets before taxes (ROAA), k as the equity capital in percent of total assets and as the standard deviation (volatility) of the ROAA. Thus, the z-score combines the banks profitability (), capital ratio (k) and return volatility () in one single
1

The EU-11 comprises Belgium, Denmark, France, Germany, Greece, Ireland, Italy, Netherlands, Portugal, Spain and the United Kingdom. 6

indicator. Obviously, the indicator will increase with the banks capital ratio and profitability, and decrease with increasing return volatility. Hence, the z-score measures the probability of a bank becoming insolvent when the value of assets falls under the value of debt. Hence, a higher (lower) z-score implies a lower (higher) probability of insolvency risk. Table 2 reports z-score ratios for European banks in our sample across countries and over time.

B. Credit risk securitization

Our initial and unique sample of 776 cash and synthetic credit risk securitizations issued by 60 stock listed bank holdings across the EU-15 plus Switzerland from 1997 to 2007 is obtained from offering circulars and presale reports provided by Moodys, Standard & Poors and FitchRatings. These reports provide detailed information on securitizations including the type and structure of the transaction as well as the underlying reference portfolio. We first of all omit those countries (Austria, Finland, Luxembourg and Sweden) exhibiting less than three securitizations over the entire sample period. Subsequently, we exclude data on transactions by banks whose shares are not traded on European Stock Exchanges. Summarized, this finally reduces the sample to 743 credit risk securitizations between 1997 and 2007 issued by 55 European bank holdings located in the EU-11 plus Switzerland. We include these securitization transactions as the log of the total extent of credit volumes being transferred each year (Table 2). As Table 4 reports, the cumulated volume of credit risk transfer amounts to 1,410,423 million. Moreover, our sample is mainly represented by the risk transfer of residential mortgage loans ( 751,227 million) and corporate loans ( 488,565 million). Figures 1 and 2 more precisely illustrate the distribution of credit risk securitizations over the sample period. Hence, in Europe a notable transfer of credit risks through securitization did not begin until 1997. Furthermore, with the exception of the year 2004 the number of credit risk transfer transactions has continuously increased over the sample period peaking in 2006 and decreasing afterwards probably as a result of the U.S. subprime crisis in mid-2007. Similarly, with the exception of the years 2001 and 2004 the volume

of credit risk transfer in Europe continuously increased over the sample period but did not decrease in 2007.

C. Further explanatory variables

When examining the relationship between securitization and stability it is imperative to control for macroeconomic, bank-specific, regulatory and institutional factors that are likely to affect bank soundness, securitization, or both, and hence help mitigate omitted variable biases. We lagged some of the variables to avoid simultaneity.

Macroeconomic control variables are retrieved from the World Development Indicator (WDI) database provided by the World Bank. We include the log of real GDP, the rate of real GDP growth, the annual change of inflation and real short term interest rates to cover macroeconomic
developments that are likely to affect the quality of bank assets and hence may influence credit risk securitization. The log of real GDP is included to control for a countrys overall level of economic environment. We assume banks operating under a higher level of economic environment to be more stable. The rate of growth of real GDP is a control variable since the banks investment opportunities may be correlated with business cycles (Laeven and Majoni, 2003). Hence, we expect a positive sign of the coefficient of real GDP growth if investment opportunities rise under economic booms. In addition, borrowers solvency should be higher under an increasing economic performance which raises the banks asset quality and may reduce credit risk securitization activities. The latter is indirectly confirmed by Stanton (1998) who provides empirical evidence that the number of securitization transactions increase in periods of cyclical downturns. However, as Estrella (2002) finds that securitization of mortgage loans tends to decline during economic recessions the evidence is not conclusive. We further include the one-period lagged changes in

inflation rates. The effect of changes in inflation rates depends on whether banks anticipated
inflation or not and whether inflation coincides with general economic fragility. Since interest rates tend to rise in the presence of inflation, inflation is probably associated with a higher realization of net interest margins and profitability. However, as the banks funding costs may also increase under
8

inflation (Hortlund, 2005) the effect on profitability and bank soundness depends on the net effect from increasing net interest margins and costs. Similarly, changes in real short term interest rates are likely to implicitly influence asset quality. Again, we expect an ambiguous effect. While a passing through of increasing short term interest rates to deposit rates will raise the banks funding costs, a handing down to lending rates should raise profitability but might let loan repayment be more difficult for borrowers which may result in higher loan default rates and decreasing asset quality. Obviously, the actual effect depends on the differences in the average maturity of assets and liabilities or the banks capability to reprice assets and liabilities. In this context, however, securitization may not only provoke a raise in asset quality by transferring credit risk to external investors but it may also provide an alternative funding source which in sum let the bank be more independent from interest rate changes (Goswami et al., 2009). Due to the fact that characteristics of securitizing banks in our sample vary across the EU-11 plus Switzerland we employ further bank-specific variables. We include the delta of log of total assets to control for changes in the banks size since Bannier and Hnsel (2008) as well as Martn-Oliver and Saurina (2007) provide empirical evidence that the financial institutions size is a strong determinant of the frequency of securitization activities. We further employ the banks net interest

margin to control for profitability, the one-period lagged non-performing loans to total assets as a
key measure for credit risk and loan-portfolio quality, the cost-income ratio to control for efficiency and the one-period lagged liquid assets to total assets as a proxy for the banks liquidity. We expect a positive sign of the coefficients of net interest margin and liquid assets to total assets and a negative sign of the coefficients of non-performing loans to total assets and cost-income ratio. Moreover, referring to securitization, Bannier and Hnsel (2008) provide empirical evidence that the number of securitization transactions tend to increase under the framework of lower asset quality, efficiency and liquidity. To draw accurate statistical inference concerning the relationship between securitization and banking stability we perform a large variety of sensitivity analyses. Thus, we control for cross-

country differences regarding the banking and capital market structure as well as the regulatory and institutional environment to provide information on possible linkages between these sensitivity measures and banking stability. We include two measures of banking market concentration. To begin with, 5-bank concentration is constructed as the fraction of assets of the total banking systems assets held by the five largest domestic and foreign banks per country (Uhde and Heimeshoff, 2009). The Herfindahl-Hirschman

Index (HHI) is computed as the sum of the squared market shares of a countrys domestic and
foreign banks. Calculating concentration ratios in this way addresses the fact that the banking industry is further globalizing and that banks compete not only within national boundaries but also cross-border. As both theoretical and empirical studies are not conclusive about the impact of banking market concentration on financial stability (Uhde and Heimeshoff, 2009; Beck et al., 2006; Schaeck et al., 2006; De Nicol et al., 2004), we expect an ambiguous effect of our concentration measures on financial stability. The competitiveness of a countrys banking market is proxied by the H-Statistic proposed by Panzar and Ross (1987). We estimate the H-Statistic on an aggregated, consolidated bank balancesheet basis cross-sectionally for each country in our sample for the period from 1997 to 2007. In contrast to related literature the H-Statistic includes interest bearing revenues only to make sure that the competition measure is more related to our sample of exclusively interest-bearing asset securitizations. 2 Moreover, since Schaeck et al. (2006) provide evidence that the effect of competition is reduced under a more sophisticated economic environment, the measure interacts
2

Hence, following Claessens and Laeven (2004) and Schaeck et al. (2006) the H-Statistic is based on revenue equations and measures the degree of market competitiveness by means of the banks elasticity of interest bearing revenues with respect to its input factor prices while controlling for a long-run market equilibrium. Thus, an increase in factor prices (a) will be mirrored by an equal-proportional increase in the interest bearing revenue under perfect competition (H = 1), (b) will be mirrored by an under-proportional increase in the interest bearing revenue under monopolistic competition (0 < H < 1) and (c) will not at all be reflected by an increase in the banks interest bearing revenue in the monopoly case (H 0). 10

with a countrys GDP to control for country-level heterogeneity. Similar to the nexus between concentration and stability, even the relationship between competition and stability is not conclusive (e.g., Beck 2008). Hence, we expect an ambiguous effect of this measure on banking stability. Next to cross-country differences concerning the banking market structure we additionally control for differences in a countrys stock market development. We again expect an ambiguous effect of this measure. On the one hand, a well-developed stock market may provoke disintermediation tendencies and spur competition for retaining bank customers which results in lower financial soundness. On the other hand, however, we assume that the level of stock market development is an appropriate proxy for the number and quality of potential external investors engaging in securitization transactions on capital markets. If this is true, higher developed stock markets are anticipated to support credit risk securitization activities and, as a possible result, the securitizing banks financial soundness. Turning to cross-country differences in the regulatory environment, we employ three timevariant measures of banking regulation and supervision proposed by Barth et al. (2004). 3 The

capital regulatory index is built by means of principal component analysis and describes a summary
measure of initial capital stringency and overall capital requirements. To the extent that greater capital stringency encourages prudent behavior of bank managers and equity capital is an appropriate measure of a banks solvency, we expect better capitalized banks to be more stable.

Activity restrictions is a key determinant for the scope of a banks business by aggregating measures
of whether and how far a bank is allowed to engage in securities, insurance and real estate markets. To the extent that activity restrictions keep banks from operating in too risky lines of business, banking systems with greater restrictions are assumed to be more stable (Beck et al., 2006; Barth et. al., 2004). In contrast, if a high level of activity restrictions prevents banks from diversifying asset risks outside traditional business, banking systems with greater restrictions may become more

We combine data from three World Bank Surveys on Bank Regulation and Supervision conducted in 1997, 2001 and 2005 to construct time-variant data series. 11

fragile. We further include the private monitoring index as a measure of the degree to which regulations empower, facilitate and encourage the private sector to monitor banks. As a higher degree of market discipline is anticipated to support state banking supervision, we expect a positive sign of the coefficient of the private monitoring index. Finally, we employ Basel I as a dummy variable to control for structural breaks in our time series of securitization transactions. 4 Structural breaks over the sample period may arise due to modified banking regulations concerning credit risk securitization when changing from Basel I to Basel II regulations. Though both regulatory frameworks obligate banks to raise equity capital as a buffer against potential default losses of securitized assets, former Basel I regulations provided an incentive to realize regulatory arbitrage if the major part of default risks was retained within the banks first-loss position. 5 Taking this into account, we expect a negative sign of the coefficient of Basel I. Apart from regulatory aspects we finally control for the institutional environment. To begin with, we include the privatization index provided by Abiad et al. (2008) to control for cross-country differences in the ownership structure of banks. We expect banking markets exhibiting a lower fraction of government-owned banks to be more efficient and stable since government-owned banks are assumed to suffer from moral hazard problems and X-inefficiency (Berger et al., 2004). This is due to the fact that these banks may anticipate to be bailed out in case of a financial distress encouraging bank managers to be less committed to prudent risk-taking behavior and efficiency aspects. Furthermore, the soundness of a bank is likely to depend on the influence exerted by shareholders since bank managers being closely monitored by shareholders are expected to avoid excessive risk-taking behavior. With regard to securitization, these bank managers are assumed to
4

We assume that new regulations on securitizations by Basel II have at the latest been anticipated by banks when the New Basel Accord has been published in June 2004.

Addressing this concern, Basel II now follows a substance over form principle whilst determining the required regulatory capital for all retained tranches of a securitization. As a consequence, Basel II provides stronger incentives to transfer subordinated tranches and in particular the first loss position of a securitization to external investors. 12

transfer higher amounts of credit risk to external investors and use liquid capital to take in less risky assets or to release own liabilities ex post. We account for these effects of shareholder influence by including the shareholder rights index provided by La Porta et al. (1998) and expect a positive relationship between bank stability and shareholder influence. Finally, we employ the rule of law as a proxy for the strength and quality of a countrys institutional environment. The index is obtained from the Worldwide Governance Indicators provided by the World Bank. We expect that greater strength and quality of institutions are key factors for a well-developed and stable financial system. 3.2 Empirical model

To test the hypothesis that credit risk securitization affects banking stability, we utilize a bankspecific random-effects model and set time dummies to control for time-specific effects. 6 Since some of the banks in our sample continuously securitize credit risk over the entire sample period while others do not, we additionally address to heterogeneous securitization frequencies by clustering standard errors on a bank-level following the generalized method based on Huber (1967) and White (1980). We estimate the financial soundness of bank i at time t as follows:

yit = it + 1cit + k xit ,k + it

(2)

where yit represents the z-score ratio as our measure of banking stability and cit is the measure of credit risk securitization. The vector xit , k includes control variables described above. it is an error term and and the ' s denote the parameters to be estimated.

As Table 2 reports, the number of observations in our panel varies. Thus, in addition to random effects, we apply the consistent estimator for the variance components by Baltagi and Chang (1994) as a robustness check to avoid possible biases resulting from our unbalanced panel. However, as results did not differ significantly from the ordinary random effects estimations, we do not comment them in this paper. 13

The Hausman test clearly rejects a fixed-effects specification in favor of the panel estimation with bank-specific random effects. Hence, assuming that it can be composed into a bank-specific timeinvariant component and a component it capturing the remaining disturbance that is assumed to be uncorrelated over time so that the equation it = + it holds, the equation can be estimated with the random effects model. With regard to our analysis, the random effects model is a consequent strategy, as most variations should be observed over time. Moreover, random effects allow for the inclusion of time-invariant variables among regressors which, in particular holds for the major part of our sensitivity measures. Moreover, considering banking regulation, all Western European countries in our sample follow the European Capital Requirement Directive (transformation of Basel II) and the European Banking Directive respectively. In this context, regulatory policies and national supervisory institutions have remained almost unchanged over the sample period. The absence of time variation in regulatory and supervisory control variables as well as the existence of a considerable time lag between regulatory changes and their effect on bank performance are commonly accepted in related literature and pointed out by Barth et al. (2004). Hence, from this point of view, financial markets in Western Europe form a homogenous entity. As a consequence, variation within the cross-section between regulatory and institutional explanatory variables is low and applying the random effects techniques is appropriate. 3.3 Empirical results

We present main empirical results in Table 5. Regression (1) reports our baseline regression results assessing the impact of credit risk securitization on banking stability as measured by the z-score-technique. While regression specifications (2)-(3) omit macroeconomic and bank-specific variables, regressions (4)-(5) are additional robustness checks to control for endogeneity and possible reverse causality between bank soundness and credit risk securitization using instrumental variable regressions. Results from the first-stages of instrumental variable regressions are presented in Table 6. Table 7 reports further empirical results from regressing credit risk securitization on
14

single components of the z-score, whereas Table 8 presents empirical results from a large variety of sensitivity analyses. 3.3.1 Main findings from z-score regressions

As Table 5 reports, securitization enters regression specification (1) significantly negative at the five-percent level indicating that an increase in credit risk securitization has a negative impact on Western European banks financial soundness. In line with the securitization-fragility view from related theoretical literature, our empirical results indicate that Western European banks may predominantly employ securitization as a source of capital relief by retaining the major part of their credit risk exposures within the first-loss position (direct effect of securitization on stability). Moreover, our empirical findings suggest that banks may further utilize securitization as an additional funding source to take on new but risky assets using liquid capital from selling securitization transactions (indirect effect of securitization on stability). These hypotheses are in line with empirical evidence provided by previous event studies (Uhde and Michalak, 2009; Hnsel and Krahnen, 2007; Franke and Krahnen, 2006; Lockwood et al., 1996) applying nearly the same sample of banks as well as from panel data analysis for the U.S. and Canadian banking market (Uzun and Webb, 2007; Dionne and Harchaoui, 2003). In contrast, our findings do not support theoretical arguments and earlier empirical findings provided by Jiangli and Pritsker (2008) promoting the securitization-stability view. Among the control variables, the log of GDP enters the regression significantly positive at the five-percent level indicating that securitizing banks operating under a more sophisticated economic environment are less prone to financial fragility. As expected, one-period lagged non-performing

loans to total assets and cost-income ratio enter the regression significantly negative at the five- and
one-percent level respectively, suggesting that higher asset quality and operational efficiency have a positive impact on financial soundness. Introducing net interest margin, this variable enters the regression significantly positive at the one-percent level indicating that securitizing banks exhibiting a higher level of profitability are more stable.
15

3.3.2

Robustness checks

By means of regressions (2)-(5) we investigate the robustness of our main results. To begin with, as Table 9 indicates, both the log of GDP as well as the rate of growth of real GDP are highly correlated with several control variables which is especially true for our bank-specific measures. Due to this, we omit both macroeconomic measures in regression specification (2). However, as specification (2) reiterates the negative relationship between credit risk securitization and banking stability while significances from other coefficients remain robust, it is not sensitive to include GDP variables in our regressions. Moreover, the negative relationship between securitization and banking stability is likely to suffer from endogeneity with regard to our baseline regression specification (1). Hence, we first of all address this statistical problem by eliminating the bank-specific control variables in regression (3) to control for bank-specific endogeneity. As shown, even though bank-specific variables are excluded, our main finding of a negative relationship between securitization and bank soundness is reconfirmed. Hence, we rule out that results are driven by bank-specific endogeneity. Among the control variables, the log of GDP, the rate of growth of real GDP and the delta of log of total assets enter the regression significantly positive indicating that banks issuing credit risk securitizations under a more sophisticated economic environment being attended with economic boom phases are less prone to financial fragility. We further address possible endogeneity problems by applying 2SLS instrumental variable techniques in regression specification (4). We employ the log of a banks net loans as an instrumental variable since the accounting value of a banks total loans is considered to be the main source for credit risk securitizations in our sample. 7 Concerning the validity of our instrumental variable Table 10 reports that the banks net loans are highly correlated with the securitization variable whereas its correlation with the z-score ratio is suffieciently low. In addition, the validity is

With the exception of Collateral Bond Obligations (0.5 per cent of the total volume of transactions in our sample) all underlyings are classified as net loans. 16

confirmed by results from the first stage of the 2SLS regression. (Table 6). As shown by regression specification (4), results from instrumental variable regressions entirely reiterate our main findings from the standard random effects model. Hence, we rule out that our baseline estimation results may be driven by endogeneity. Finally, the causality running from securitization to banking stability is not clear since it is not obvious if the frequency of securitization activities itself depends on the banks financial soundness. Hence, reverse causality may arise as a particular case of endogeneity, for example, if a bank exhibiting a high risk exposure or a shortage of liquidity tends to increase credit risk securitization activities to gamble for resurrection. Thus, to address likely reverse causality, we again apply instrumental variable techniques using a 2SLS panel estimator in regression (5) and employ two-

period lagged securitization as an instrumental variable (Tables 10 and 6 indicate the validity of our
instrument). As reported by regression specification (5), the instrumental variable regression reconfirms our baseline results from the standard random effects model suggesting that the negative impact of credit risk securitization on bank stability is not biased by reverse causality. 3.3.3 Main findings from z-score components regressions

By means of regressions (1)-(3) in Table 7 we try validate our hypotheses from our baseline regression suggesting that securitization is predominantly utilized as a source of capital relief and additional funding and that both strategies may provoke a decrease in the banks financial soundness. We scrutinize our hypotheses by regressing the securitization variable on single components of the z-score ratio. We use the same set of control variables being employed in the baseline regression but have to omit single bank-specific variables that are highly correlated with each of the new dependent variables to avoid biased estimation results (Table 11). To begin with, we include the banks ROAA as the dependent variable in specification (1).

Securitization enters the regression significantly negative at the five-percent level indicating that
increasing securitization activities may reduce bank profitability. This result supports our hypotheses from the baseline regression as it is assumed that retaining the major part of credit risk
17

exposure within the banks first-loss position and taking on new risky assets may reduce bank profitability and soundness due to a higher probability of future loan losses. Moreover, Altunbas et al. (2007) and Goderis et al. (2006) provide empirical evidence that securitization has a positive impact on the banks target loan levels. Hence, applying traditional industrial organization theory to banking, an explanation for the negative relationship between securitization and profitability may be the increase in competition in loan markets triggered by credit risk securitization which should result in a decrease in the banks interest and profit margins. Among the control variables log of

GDP enters the regression significantly negative confirming the trend of declining bank margins,
which especially holds for well-developed Western European banking markets for the last decade. As expected, the coefficient of rate of growth of real GDP exhibits a significantly positive sign suggesting that the banks investment opportunities and borrowers solvency may rise under economic booms. Finally, real short term interest rate enters the regression significantly positive indicating that banks may predominantly hand down increasing interest rates to borrowers rather than to depositors. By means of regression (2) we assess the relationship between securitization and the banks

capital ratio as the second component of the z-scores numerator. Securitization enters the
regression significantly negative at the five-percent level indicating a positive relationship between credit risk securitization and the banks leverage. Again, this finding supports our hypotheses from the baseline regression as it is assumed that the banks leverage may increase if the bank realizes regulatory capital arbitrage by means of credit risk securitization. This causality is also proposed by related theoretical literature (Leland, 2007; Cebenoyan and Strahan, 2004) and has been confirmed by previous empirical studies (Jiangli and Pritsker, 2008). Among the control variables inflation enters the regression significantly negative supporting theoretical arguments and empirical findings that the banks funding costs may increase under inflation (Schaeck et al., 2006; Hortlund, 2005). We finally include the volatility of the ROAA as the z-scores denominator in regression specification (3). If it is true that credit risk is the main source of a banks overall risk exposure, the

18

return volatility is a measure of a banks loan portfolio quality. Securitization enters the regression significantly positive but weak at the ten-percent level indicating that securitization tends to increase the volatility of bank asset returns which results in a decrease in the banks loan portfolio quality. Hence, in line with results from regressions on the ROAA empirical findings support our hypotheses from the baseline regression as it is assumed that retaining the more risky first-loss position and taking on new risky assets may increase the banks return volatility due to a higher probability of future loan losses. As expected, among the control variables liquid assets to total

assets enters the regression significantly negative.

To sum up, taking the single results from regressions on z-score components into account, main rationales for the negative impact of credit risk securitization on Western European banks financial soundness may be derived from the positive relationship between credit risk securitization and the banks leverage and return volatility as well as from a negative relationship between securitization and bank profitability. In this way, each single result from the z-score components regressions validates our hypotheses suggesting that securitization is predominantly utilized as a source of capital relief and additional funding and that both the direct and indirect effect of securitization may provoke a decrease in the banks financial soundness. 3.3.4 Sensitivity analyses

We perform a large variety of sensitivity analyses. As a general result, our main finding of a negative relationship between securitization and banking stability holds even when controlling for cross-country differences concerning the banking market structure, the capital market development as well as the regulatory and institutional environment. Due to high correlation between these control variables (Table 12), we include them in turn in separate regressions (Table 8). With regard to cross-country differences in the banking market structure, we first of all include the 5-bank concentration ratio in regression specification (1) and the HHI in regression specification (2). As Table 8 reports, both concentration measures enter the regressions significantly positive indicating that securitizing banks in more concentrated banking markets are less prone to
19

financial fragility. Corresponding with theoretical arguments and empirical findings provided by related literature on this issue (e.g., Uhde and Heimeshoff, 2009) we suggest that larger banks may engage in credit rationing more heavily since fewer high quality credit investments will increase the return of the singular investment and hence foster financial soundness. Moreover, larger banks may exhibit comparative advantages in providing credit monitoring services and may be able to diversify loan portfolio risks more efficiently due to higher economies of scale and scope. Introducing H-Statistic, this variable enters regression (3) significantly negative at the one-percent level indicating that securitizing banks operating under increasing market competition are more prone to financial fragility. This result is in line with theoretical models and empirical findings predicting that in a more competitive environment with higher pressures on profits, banks have higher incentives to take more excessive risks, resulting in higher fragility. In addition, banks are anticipated to earn fewer informational rents from their relationship with borrowers in competitive markets, which may reduce their incentives to properly screen borrowers, again increasing the risk of fragility (e.g. Beck, 2008). We finally control for cross-country differences concerning the capital market development. Stock market capitalization enters regression specification (4) significantly positive. This result does not correspond with theoretical arguments and empirical evidence stressing that a well-developed capital market may support financial disintermediation and spur competition for retaining bank customers which results in lower financial soundness (Schaeck et al., 2006, Bikker, 2004). In contrast, evidence suggests that the level of stock market development may be an appropriate proxy for the number and quality of potential external investors engaging in securitization transactions on capital markets. Hence, higher developed stock markets are anticipated to support credit risk securitization. However, as our main findings reveal, the effect of a higher developed stock market on banking stability depends on the amount of credit risks that are actually transferred by banks. Turning to the regulatory environment, we initially include the capital regulatory index. This variable enters regression specification (5) significantly positive at the five-percent level suggesting

20

that better capitalized securitizing banks are more stable and that greater capital stringency may encourage prudent behavior of bank managers. We further employ activity restrictions to control for governmental restrictions on banking business. As regression (6) reports, the coefficient of activity restrictions exhibits a significantly negative sign suggesting that securitizing banks operating under higher restrictions are more prone to financial fragility. Our findings correspond with theoretical predictions and empirical findings stressing that restricting the banks business activities is likely to reduce the banking systems efficiency and stability (Barth et al., 2004) since it detains financial institutions from reaping benefits of diversification opportunities. As expected, the private

monitoring index enters regression specification (7) significantly positive at the weak ten-percent
level indicating that forces of higher market discipline may support state bank supervision and hence, help promoting financial stability. Finally, Basel I enters regression (8) significantly negative. Hence, evidence suggests that securitizing credit risk under former Basel I regulations affects financial stability negatively since Basel I regulations provided an incentive to realize regulatory arbitrage from securitization if the major part of credit default risks were retained within the first loss position. We finally control for cross-country differences concerning the institutional environment. Introducing the privatization index, this variable enters regression (9) significantly positive at the one-percent level indicating that banking markets with a lower amount of government-owned banks are less prone to financial fragility. This result was expected since government-owned banks are anticipated to strongly suffer from moral hazard problems and X-inefficiency. Employing the

shareholder rights index, this measure enters regression specification (10) significantly positive
confirming theoretical assumptions and empirical findings that managers of securitizing banks being closely monitored by shareholders tend to avoid excessive risk-taking behavior and may transfer the major part of credit risks to external investors (Park and Peristiani, 2007). Finally, the coefficient of rule of law exhibits a significantly positive sign in regression specification (11)

21

suggesting that securitizing banks operating under a stronger institutional environment of higher quality are less prone to financial fragility. 4 Conclusion Using a unique sample of 743 cash and synthetic securitization transactions issued by 55 stock listed bank holdings located in the EU-11 plus Switzerland over the period from 1997 to 2007, this paper provides empirical evidence that credit risk securitization has a negative impact on Western European banks financial soundness as measured by the z-score technique while controlling for macroeconomic, bank-specific, regulatory and institutional factors. Empirical results from panel estimations hold when applying instrumental variable techniques to address probable endogeneity and performing a variety of further sensitivity analyses. Our empirical findings support theoretical assumptions provided by the securitization-fragility view and confirm empirical evidence from previous studies by Uzun and Webb (2007) and Dionne and Harchaoui (2003). Investigating single z-score components we additionally find a positive impact of credit risk securitization on the banks leverage and return volatility as well as a negative relationship between securitization and the banks profitability. Hence, in line with baseline regressions on the z-score ratio we suggest that securitization is predominantly utilized as a source of capital relief and additional funding and that both the direct and indirect effect of securitization may provoke a decrease in Western European banks financial soundness. Moreover, evidence from sensitivity analyses reveals that securitizing banks operating in more concentrated banking markets and higher developed capital markets are less prone to financial fragility whereas a higher level of banking market competitiveness negatively affects the banks financial soundness. While capital regulations support financial stability, high restrictions on banking activities do not. Finally, stronger forces of market discipline and a higher shareholder influence tend to support state banking regulation and supervision resulting in greater banking stability.

22

Against the background of our empirical results we point out that recent proposals by the Basel Committee to enhance the Basel II framework in the field of securitization are a step in the right direction. First of all, we expressly emphasize the necessity for further focusing on improved minimum capital standards concerning a banks securitization exposures in Pillar 1 since our empirical findings reveal that Western European banks tend to utilize securitization as a source of capital relief. Furthermore, strengthening disclosure requirements in the area of securitization is assumed to be a useful instrument, however, as banks do not yet reveal whether the first-loss position is actually transferred out of the balance sheet, disclosure requirements should be further reinforced in this regard. Finally, addressing securitization within the frameworks of ICAAP, SREP and MIS in order to enhance the banks and supervisors sensitivity to securitization was expected to be a remedy for lessons learnt from the recent financial crisis. However, one has to reconsider if modifying the Basel Accord in this way could really help to prevent future financial turmoil as long as state supervisors lack relevant knowledge to fully understand and monitor high complex structured finance instruments and, as long as external investors knowingly invest in risky instruments to make fast money at the expense of a safeguard.

23

References
Abiad, A.G., Tressel, T., Detragiache, E., 2008. A new database of financial reforms. Working Paper No. 08/266, IMF. Allen, F., Gale, D., 2006. Systemic risk and regulations. In: Carey, M., Stulz, R.M. (Eds.), The Risks of Financial Institutions. University of Chicago Press: Chicago, pp. 341-376. Altunbas, Y., Gambacorta, L., Marqus, D., 2007. Securitisation and the bank lending channel. Working Paper No. 838, ECB. Bank for International Settlements (BIS), 2005. Credit Risk Transfer. Basel. Bank for International Settlements (BIS), 2008. Credit Risk Transfer: Developments from 2005 to 2007. Basel. Bank for International Settlements (BIS), 2008a. Fair Value Measurement and Modeling: An Assessment of Challenges and Lessons Learned from the Market Stress. Basel. Bannier, C.E., Hnsel, D.N., 2008. Determinants of European banks engagement in loan securitization. Working Paper No. 10/2008, Deutsche Bundesbank. Baltagi, B., Chang, Y., 1994. Incomplete panels: A comparative study of alternative estimators for the unbalanced oneway error component regression model. Journal of Econometrics 62, 67-89. Barth, J.R., Caprio, G., Levine, R., 2004. Banking regulation and supervision: What works best? Journal of Financial Intermediation 13, 205-248. Basel Committee on Banking Supervision (BCBS), 2009. Proposed enhancements to the Basel II framework. Consultative Document January 2009, Bank for International Settlements. Beck, T., 2008. Competition and stability: Friends or foes? Working Paper No. 4656, World Bank. Beck, T., Demirg-Kunt, A., Levine R., 2000. A new database on financial development and structure. World Bank Economic Review 14, 597-605. Beck, T., Demirg-Kunt, A., Levine, R., 2006. Bank concentration, competition, and crises: First results. Journal of Banking and Finance 30, 1581-1603. Berger, A.N., Demirg-Kunt, A., Levine, R., Haubrich, G.J., 2004. Bank concentration and competition: An evolution in the making. Journal of Money, Credit and Banking 36, 433-450. Bikker, J. A., 2004. Competition and Efficiency in a Unified European Banking Market. Cheltenham: Edward Elgar. Boyd, J.H., Runkle, D.E., 1993. Size and performance of banking firms, Journal of Monetary Economics 31, 47-67. Cebenoyan, S., Strahan, P.E., 2004. Risk management, capital structure and lending at banks. Journal of Banking and Finance 28, 19-43. Claessens, S., Laeven, L., 2004. What drives bank competition? Some international evidence. Journal of Money, Credit and Banking 36, 563-583.

24

DeMarzo, P.M., 2005. The pooling and tranching of securities: A model of informed intermediation. Review of Financial Studies 18, 351-61. Demsetz, R.S., 2000. Bank loan sales: A new look at the motivations for secondary market activity. Journal of Financial Research 23, 197-222. De Nicol, G., Bartholomew, P., Zaman, J., Zephirin, M., 2004. Bank consolidation, internalization, and conglomerization. Working Paper No. 03/158, IMF. Dionne, G., Harchaoui, T.M., 2003. Banks capital, securitization and credit risk: An empirical evidence for Canada. Working Paper No. 03-01, HEC Montral. Estrella, A., 2002. Securitization and the efficacy of monetary police. Economic Policy Review, FRBNY. European Central Bank (ECB), 2004. Credit Risk Transfer by EU Banks: Activities, Risks, and Risk Management. Frankfurt/Main. Franke, G., Krahnen, J.P., 2006. Default risk sharing between banks and markets: The contribution of collateralized debt obligations. In: Carey, M., Stulz, R.M. (Eds.), The Risks of Financial Institutions. University of Chicago Press: Chicago, pp. 603-634. Goddard, J., Molyneux, P., Wilson, J.O.S., Tavakoli, M., 2007. European banking: An overview. Journal of Banking and Finance 31, 1911-1935. Goderis, B., Marsh, I.W., Castello, J.V., Wagner, W., 2006. Bank behavior with access to credit risk transfer markets. Working Paper, University of Oxford. Goswami, M., Jobst, A., Long, X., 2009. An investigation of some macro-financial linkages of securitization. Working Paper No. 09/26, IMF. Greenbaum, S.I., Thakor, A.V., 1987. Bank funding modes: Securitization versus deposits. Journal of Banking and Finance 11, 379-401. Hnsel, D.N., Krahnen, J.P., 2007. Does credit securitization reduce bank risk? Evidence from the European CDO market. Working Paper, University of Frankfurt/Main. Hortlund, P., 2005. Do inflation and high taxes increase bank leverage?. Working Paper Series in Economics and Finance No. 612, Stockholm School of Economics. Huber, P.J., 1967. The Behavior of Maximum Likelihood Estimates under Nonstandard Conditions. Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability. Berkeley, CA: University of California Press. International Monetary Fund (IMF), 2002, Global Financial Stability Report: A Quarterly Report on Market Developments and Issues. Washington DC. International Monetary Fund (IMF), 2007. Global Financial Stability Report: Financial Market Turbulence: Causes, Consequences, and Policies. Washington DC. 25

International Monetary Fund (IMF), 2008. Global Financial Stability Report: Financial Stress and Deleveraging MacroFinancial Implications and Policy. Washington DC. International Monetary Fund (IMF), 2008a. Global Financial Stability Report: Containing Systemic Risks and Restoring Financial Soundness. Washington DC. Instefjord, N., 2005. Risk and hedging: Do credit derivates increase bank risk? Journal of Banking and Finance 29, 33345. Jiangli, W., Pritsker, M., Raupach, P., 2007. Banking and securitization. Working Paper, FDIC, Federal Reserve Board and Deutsche Bundesbank. Jiangli, W., Pritsker, M., 2008. The impacts of securitization on U.S. bank holding companies. Working Paper, FDIC and Federal Reserve Board. Krahnen, J.P., Wilde, C., 2008. Risk transfer with CDOs and systemic risk in banking. Working Paper, University of Frankfurt/Main. La Porta, R., Lopez-de-Silanes, F., Shleifer, A., 1998. Law and finance. Journal of Political Economy 106, 1113-1155. Laeven, L., Majoni, G., 2003. Loan loss provisioning and economic slowdowns: Too much, too late? Journal of Financial Intermediation 12, 178-197. Leland, H.E., 2007. Financial synergies and the optimal scope of the firm: Implications for mergers, spinoffs, and structured finance. Journal of Finance 62, 765-807. Lockwood, L.J., Rutherford, R.C., Herrera, M.J., 1996. Wealth effects of asset securitization. Journal of Banking and Finance 20, 151-64. Martn-Oliver, A., Saurina, J., 2007. Why do banks securitize assets? Working Paper, Banco de Espagna. Panzar, J. C., Rosse J. N., 1987. Testing for monopoly equilibrium. Journal of Industrial Economics 35, 443-456. Park, S., Peristiani, S., 2007. Are bank shareholders enemies of regulators or a potential source of market discipline? Journal of Banking and Finance 31, 2493-1515. Riddiough, T.J., 1997. Optimal design of asset backed securities. Journal of Financial Intermediation 6, 121-52. Schaeck, K., ihk, M., Wolfe, S., 2006. Competition, concentration and bank soundness: New evidence from the micro-level. Working Paper No. 06/143, IMF. Shin, H.S., 2009. Securitisation and financial stability. The Economic Journal 119, 309-332. Stanton, S., 1998. The underinvestment problem and patterns in bank lending. Journal of Financial Intermediation 7, 466-486. Uhde, A., Michalak, T.C., 2009. Securitization and systematic risk in European banking: Empirical evidence. Working Paper, University of Bochum.

26

Uhde, A., Heimeshoff, U., 2009. Consolidation in banking and financial stability in Europe: Empirical evidence. Journal of Banking and Finance 33, 1299-1311. Uzun, H., Webb, E., 2007. Securitization and risk: empirical evidence on US banks. Journal of Risk Finance 8, 11-23. White, H., 1980. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica 48, 817-830.

27

Statistical appendix2
Table 1 Notes on variables and data sources
Variable Z-score Description Ratio of the sum of equity capital to total assets and the return on average assets before taxes (ROAA) to standard deviation of ROAA (sdROAA). Accounting value of a banks return on average assets before taxes (ROAA). Accounting value of a banks ratio of equity capital to total assets. Standard deviation of a banks ROAA. Log of the total extent of cash and synthetic credit risk securitizations issued by 55 banks across the EU-11 plus Switzerland from 1997 to 2007. Lag (2) of the total extent of cash and synthetic credit risk securitizations issued by 55 banks across the EU-11 plus Switzerland from 1997 to 2007. Log of real GDP. Rate of real GDP growth at constant 2000 prices (annual percentage change). Lag (1) of annual change of GDP deflator. Annual change of real short term interest rate, adjusted for inflation (GDP deflator). Proxy for the banks size. Delta of log of the accounting value of a banks total assets. Proxy for the banks profitability. Log of accounting value of a bank's net interest revenue as a share of its interest-bearing (total earning) assets. Data Sources BankScope, authors calc.

ROAA Capital ratio sdROAA Securitization

BankScope CN 4006 BankScope CN 2095 BankScope CN 4006, authors calc. Moodys, Standard & Poors and FitchRatings

Securitization (t-2)

Moodys, Standard & Poors and FitchRatings

GDP GDP growth

World Development Indicators (WDI) World Development Indicators (WDI)

Inflation (t-1) Interest rate Total assets Net interest margin

World Development Indicators (WDI) World Development Indicators (WDI) BankScope CN 2025 BankScope CN 2035

Non-performing loans (t-1) Proxy for the banks asset quality. Lag (1) of the accounting value of a banks non-performing loans as a share of its total assets. Cost-income ratio Proxy for the banks efficiency. Accounting value of the ratio of a banks overhead costs to its total revenue. Proxy for the banks liquidity. Lag (1) of the accounting value of a banks liquid assets to its total assets. Variable constructed as 1 net loans to total assets. Log of accounting value of a banks net loans. EU-11 plus Switzerland concentration ratios: Fraction of assets of a countrys total banking system's assets held by the largest 5 domestic and foreign banks. Herfindahl-Hirschman Index computed as the sum of the squared market shares of a countrys domestic and foreign banks.

BankScope CN 2170

BankScope CN 4029

Liquid assets (t-1)

BankScope CN 4032

Net loans 5-bank concentration

BankScope CN 5190 ECB statistics, national central banks

HHI

ECB statistics, national central banks

28

Table 1 (continued) Notes on variables and data sources

Variable H-Statistic Description H-Statistic estimated on an aggregated consolidated bank balance-sheet basis cross-sectionally for each country for the period from 1997 to 2007. H-Statistic comprises banks interest bearing revenues. Variable is interacted with a countrys GDP. Proxy for the development of the capital market. Proportion of the banking sector assets to stock market capitalization. Index that measures the overall capital stringency. Index is built by first principal component analysis of initial capital stringency and overall capital stringency. Higher index values indicate greater capital stringency. Index aggregates measures that indicate whether bank activities in the securities, insurance and real estate markets, ownership and control of non-financial firms are unrestricted, permitted, restricted, or prohibited. Index ranges between (0) and (10), with higher values indicating greater activity restrictions arising from legal requirements. Index aggregates measures that indicate the degree to which regulations empower, facilitate, and encourage the private sector to monitor banks. Index ranges between (0) and (9), with higher values indicating higher market discipline. Dummy variable that takes on the value of 1 if credit risk securitizations were issued under Basel I regulations (19972003), and zero otherwise. Aggregate index of banking system privatization. Index ranges from 0 to 4, with higher indices indicating higher privatization. Aggregated index for the emphasis on shareholder rights, with higher values indicating more shareholder rights. Indicator that measures individuals degree of confidence in rules of society and the likelihood of crime and violence. Scores range between 2.5 and 2.5, with higher scores corresponding with better outcomes. Data Sources BankScope, authors calc.

Stock market capitalization

Beck et al. (2000)

Capital regulatory index

Barth et al. (2004), Authors calc.

Activity restrictions

Barth et al. (2004)

Private monitoring index

Barth et al. (2004)

Basel I

Authors calc.

Privatization index

Abiad et al. (2008)

Shareholder rights index Rule of law

La Porta et al. (1998) Worldwide Governance Indicators (WGI)

29

Table 2 Descriptive statistics Variable Z-score ROAA Capital ratio sdROAA Securitization Securitization (t-2) Net loans GDP GDP growth Inflation (t-1) Interest rate Total assets Net interest margin Non-performing loans (t-1) Cost-income ratio Liquid assets (t-1) 5-bank concentration HHI H-Statistic Stock market capitalization Capital regulatory index Activity restrictions Private monitoring index Basel I Privatization index Shareholder rights index Rule of law N 568 568 568 568 303 224 578 605 605 550 568 567 579 497 579 531 597 597 594 605 605 605 605 605 605 605 605 Mean 0.6934 0.0067 0.0531 0.0422 7.6825 7.4856 11.0773 27.3182 0.026 0.001 0.0568 0.1304 -4.0571 0.0349 0.6182 0.4491 0.4118 567.89 0.7885 0.8928 0.9172 7.1091 6.0182 0.6364 2.3736 2.6 1.3651 SD 1.2620 0.0046 0.0234 0.0585 1.2834 1.2420 1.2526 0.9442 0.0173 0.0286 0.5961 0.2079 0.6008 0.0367 0.1401 0.1629 0.1868 494.57 0.1485 0.7033 0.5207 2.0878 1.2293 0.4814 0.8194 1.6373 0.4335 Min 0 -0.0174 0.0079 0.0001 3.2189 3.2189 6.8628 25.0997 -0.0074 -0.2005 -2.3579 -0.66 -6.5023 0 0.2035 0.11 0.17 114 0.5057 0.1191 -0.6393 3 4 0 0 0 0.36 Max 14.1327 0.0222 0.1606 0.5349 10.5308 10.3315 13.9462 28.4388 0.1168 0.2928 5.2874 2.66 -2.7031 0.2958 1.1683 0.88 0.86 2108 0.9779 4.2785 1.4352 10 8 1 3 5 2.08

30

Table 3 Geographical distribution of banks in the sample Country Belgium Denmark France Banks Dexia KBC Groupe Danske Bank Sydbank BNP Paribas Crdit Agricole Natixis Socit Gnrale Bayr. Hypo- u. Vereinsbank Commerzbank Deutsche Bank Deutsche Postbank Hypo Real Estate Holding IKB Dt. Industriebank EFG Eurobank Ergasias Anglo Irish Banks Bank of Ireland DePfa Bank Banca Antonveneta Banca Carige Banca Lombarda Banca Monte Dei Paschi Banca Naz. Lavoro Banca Popolare Milano Banca Popolare Italiana Capitalia Intesa Sanpaolo Mediobanca San Paolo IMI Unicredito Italiano Country Netherlands Banks ABN Amro Holding Fortis Bank SNS Reaal Banco BPI Banco Espr. Santo BCP BBVA Banco De Sabadell Banco De Valencia Banco Pastor Banco Popular Espanol Banco Santander Bankinter Abbey National Alliance & Leicester Barclays Bank Bradford & Bingley HBOS HSBC Holdings Lloyds TSB Group Northern Rock Royal Bank of Scotland Standard Chartered Credit Suisse UBS

Portugal

Spain

Germany

United Kingdom

Greece Ireland

Italy

Switzerland

31

Table 4 Descriptive statistics of securitization transactions in the sample (in million ) N Total Total Transactions Single Underlyings Collateralized Debt Obligations from small and medium enterprises from large enterprises Residential Mortgage Backed Securities Commercial Mortgage Backed Securities Credit Cards Receivables Consumer Loans Others 743 1,410,423 1,898 1,076 2,192 25 22,000 Total Volume Mean Median Standard Deviation Minimum Maximum

233 107 126 312 84 20 44 50

488,565 181,712 306,853 751,227 74,999 16,340 33,066 46,226

2,097 1,698 2,435 2,408 893 817 752 925

1,335 1,250 1,500 1,490 661 855 490 704

2,396 1,640 2,850 2,415 907 404 654 736

60 60 196 87 199 56 25 28

16,863 7,728 16,863 22,000 7,092 1,658 3,000 3,100

32

Figure 1 Development of the volume of credit risk transfer through securitizations during the sample period (in million )
Volume of Credit Risk Transfer
350,000 300,000 250,000 200,000 150,000 100,000 50,000 0 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

Years

Figure 2 Development of the number of securitization transactions during the sample period

33

Table 5 Z-score and securitization (1) Z-score Securitization GDP GDP growth Inflation (t-1) Interest rate Total assets Net interest margin Non-performing loans (t-1) Cost-income ratio Liquid assets (t-1) Time Dummies No. of Obs. No. of Groups Wald 2 Adj. R2 0.1638 (0.016)** 0.1833 (0.042)** 2.9212 (0.698) 0.6310 (0.634) 0.0388 (0.625) 0.3810 (0.390) 0.7751 (0.004)*** 6.3266 (0.022)** 2.2221 (0.000)*** 0.6755 (0.269) yes 260 52 146.77*** 0.24 (2) Z-score 0.1447 (0.023)** (3) Z-score 0.2249 (0.007)*** 0.1393 (0.098)* 21.7043 (0.005)*** 1.0313 (0.277) 0.0363 (0.517) 0.9649 (0.031)** (4) Z-score 0.2227 (0.059)* 0.2107 (0.067)* 2.7683 (0.755) 0.8320 (0.818) 0.0488 (0.721) 0.3953 (0.464) 0.7674 (0.001)*** 6.7152 (0.009)*** 2.0141 (0.023)** 0.6246 (0.368) yes 259 52 72.30*** 0.23 (5) Z-score 0.3532 (0.034)** 0.2562 (0.111) 12.5745 (0.349) 0.6357 (0.871) 0.0285 (0.856) 0.8361 (0.338) 0.5672 (0.123) 8.1254 (0.035)** 2.3225 (0.042)** 0.1353 (0.884) yes 155 44 84.20*** 0.40

0.4095 (0.739) 0.0320 (0.694) 0.4439 (0.307) 0.6846 (0.003)*** 5.2742 (0.011)** 2.4431 (0.000)*** 0.5175 (0.371) yes 260 52 119.99*** 0.23

yes 291 55 28.80** 0.11

The panel model estimated is Z-score (i=bank, j=time) = + 1 Securitizationi,t + 2 GDPi,t + 3 GDP growthi,t + 4 Inflationi,t-1 + 5 Interest ratei,t + 6 Total assetsi,t + 7 Net interest margini,t + 8 Non-performing loansi,t-1 + 9 Cost-income ratioi,t + 10 Liquid assetsi,t-1 +i,t. GDP and GDP growth are omitted in specification (2) and bank-specific variables are omitted in specification (3). Securitization is instrumented using net loans in specification (4) and the two-period lagged securitization variable in specification (5). Regressions (4) and (5) are estimated by means of a 2SLS instrumental variable regression. Constant term included but not reported. Heteroscedasticity consistent P-values are in parenthesis. ***, **, *: statistically significant at the 1, 5 and 10% level.

34

Table 6 First stage regressions (instruments) (1) Net loans Securitization (t-2) GDP GDP growth Inflation (t-1) Interest rate Total assets Net interest margin Non-performing loans (t-1) Cost-income ratio Liquid assets (t-1) No. of Obs. Wald 2 0.0023 (0.978) 6.7188 (0.306) 0.0374 (0.989) 0.0660 (0.512) 0.0492 (0.902) 0.6341 (0.000)*** 1.6609 (0.379) 0.2138 (0.744) 2.1702 (0.000)*** 259 238.00*** 0.7083 (0.000)*** 0.5892 (0.000)*** 0.065 (0.571) 0.6379 (0.948) 1.1212 (0.695) 0.0055 (0.962) 1.7222 (0.006)*** 0.5458 (0.031)** 3.0856 (0.261) 0.1264 (0.879) 1.0013 (0.141) 155 125.00*** (2)

Securitization is instrumented by net loans in specification (1). It is instrumented by the two-period lagged securitization variable in specification (2). P-values are in parenthesis. ***, **, *: statistically significant at the 1, 5 and 10% level.

35

Table 7 Z-score components and securitization (1) ROAA Securitization GDP GDP growth Inflation (t-1) Interest rate Total assets Non-performing loans (t-1) Liquid assets (t-1) Time Dummies No. of Obs. No. of Groups Wald 2 Adj. R2 0.0018 (0.429) yes 291 55 87.27*** 0.30 yes 260 52 88.82*** 0.20 0.0004 (0.036)** 0.0012 (0.007)*** 0.0597 (0.073)* 0.0019 (0.606) 0.0004 (0.005)*** 1.59e-07 (1.000) (2) Capital Ratio 0.0017 (0.029)** 0.0020 (0.452) 0.0262 (0.796) 0.0284 (0.016)** 0.0001 (0.778) 0.0006 (0.871) 0.0016 (0.976) (3) sdROAA 0.051 (0.064)* 0.022 (0.597) 0.3150 (0.424) 0.0751 (0.479) 0.0029 (0.174) 0.0313 (0.289) 0.1346 (0.403) 0.0121 (0.647) yes 260 52 31.45** 0.10

The panel model estimated is Z-score (i=bank, j=time) = + 1 Securitizationi,t + 2 GDPi,t + 3 GDP growthi,t + 4 Inflationi,t-1 + 5 Interest ratei,t + 6 Total assetsi,t + 7 Net interest margini,t + 8 Non-performing loansi,t-1 + 9 Cost-income ratioi,t + 10 Liquid assetsi,t-1 +i,t. Z-score is substituted by its single components ROAA, capital ratio and standard deviation of ROAA in specifications (1)-(3). Net interest margin, non-performing loans (t-1) and cost-income ratio are omitted in specification (1). Net interest margin, cost-income ratio and liquid assets (t-1) are omitted in specification (2). Net interest margin and cost-income ratio are omitted in specification (3). Constant term included but not reported. Heteroscedasticity consistent P-values are in parenthesis. ***, **, *: statistically significant at the 1, 5 and 10% level.

Table 8 Sensitivity analyses: Market structures, regulatory and institutional environment Z-score Securitization 5-bank concentration HHI H-Statistic Stock market capitalization Capital regulatory index Time Dummies No. of Obs. No. of Groups Wald 2 Adj. R2 yes 258 52 148.42*** 0.25 yes 258 52 142.85*** 0.24 yes 259 51 151.72*** 0.26 yes 260 52 136.32*** 0.25 (1) 0.1981 (0.007)*** 1.7753 (0.046)** (2) 0.1902 (0.009)*** (3) 0.2080 (0.004)*** (4) 0.1900 (0.008)*** (5) 0.1612 (0.019)**

0.0004 (0.050)* 2.02e-12 (0.006)*** 0.2740 (0.020)** 0.5021 (0.044)** yes 260 52 180.11*** 0.25

The panel model estimated is Z-score (i=bank, j=time) = + 1 Securitizationi,t + 2 GDPi,t + 3 GDP growthi,t + 4 Inflationi,t-1 + 5 Interest ratei,t + 6 Total assetsi,t + 7 Net interest margini,t + 8 Non-performing loansi,t-1 + 9 Cost-income ratioi,t + 10 Liquid assetsi,t-1 +i,t. Constant term included but not reported. Heteroscedasticity consistent P-values are in parenthesis. ***, **, *: statistically significant at the 1, 5 and 10% level.

36

Table 8 (continued) Sensitivity analyses: Market Structures, regulatory and institutional environment Z-score Securitization Activity restrictions Private monitoring index Basel I Privatization index Shareholder rights index Rule of law Time Dummies No. of Obs. No. of Groups Wald 2 Adj. R2 yes 260 52 152.27*** 0.25 yes 260 52 123.56*** 0.24 yes 260 52 146.77*** 0.24 yes 260 52 139.38*** 0.25 yes 260 52 149.11*** 0.25 (6) 0.1843 (0.010)** 0.1069 (0.057)* (7) 0.1866 (0.010)** (8) 0.1638 (0.016)** (9) 0.1844 (0.007)*** (10) 0.1812 (0.010)** (11) 0.1792 (0.012)**

0.1303 (0.098)* 0.5402 (0.044)** 0.2149 (0.003)*** 0.1259 (0.027)** 0.4274 (0.073)* yes 260 52 142.67*** 0.24

The panel model estimated is Z-score (i=bank, j=time) = + 1 Securitizationi,t + 2 GDPi,t + 3 GDP growthi,t + 4 Inflationi,t-1 + 5 Interest ratei,t + 6 Total assetsi,t + 7 Net interest margini,t + 8 Non-performing loansi,t-1 + 9 Cost-income ratioi,t + 10 Liquid assetsi,t-1 +i,t. Constant term included but not reported. Heteroscedasticity consistent P-values are in parenthesis. ***, **, *: statistically significant at the 1, 5 and 10% level.

37

Table 9 Correlation matrix (z-score regressions)

Non-performing loans (t-1)

Net interest margin

Securitization GDP GDP growth Inflation (t-1) Interest rate Total assets Net interest margin Non-performing loans (t-1) Cost-income ratio Liquid assets (t-1)

1.00 0.27*** -0.15*** -0.01 0.11* 0.06 -0.29*** -0.07 0.05 0.14** 1.00 -0.41*** 0.03 0.02 -0.03 -0.11*** 0.34*** 0.01 0.01 1.00 -0.03 0.09** 0.16*** 0.13*** -0.29*** -0.25*** -0.12*** 1.00 -0.01 -0.01 -0.07 0.04 -0.02 0.05 1.00 0.08* -0.07 -0.09* -0.01 -0.05 1.00 0.04 -0.08* -0.11*** -0.03 1.00 0.10** -0.06 -0.54*** 1.00 0.27*** 0.09* 1.00 0.37*** 1.00

Table 10 Correlation matrix (instruments) Z-score Z-score Securitization Net loans Securitization (t-2) 1.00 -0.21*** -0.16*** -0.26*** 1.00 0.62*** 0.61*** 1.00 0.55*** 1.00 38 Securitization Net loans Securitization (t-2)

Liquid Assets (t-1)

Cost-income ratio

Securitization

Inflation (t-1)

GDP growth

Interest Rate

GDP

Total assets

Table 11 Correlation matrix (z-score components regressions)

Non-performing loans (t-1)

Net interest margin

Securitization ROAA Capital Ratio sdROAA GDP GDP growth Inflation (t-1) Interest rate Total assets Net interest margin Non-performing loans (t-1) Cost-income ratio Liquid assets (t-1)

1.00 -0.21*** -0.40*** 0.10* 0.27*** -0.15*** -0.01 0.11* 0.06 -0.29*** -0.07 0.05 0.14** 1.00 0.47*** -0.38*** -0.28*** 0.43*** -0.01 0.07 0.19*** 0.45*** -0.31*** -0.48*** -0.24*** 1.00 -0.16*** -0.04 0.03 -0.02 -0.08* 0.04 0.61*** 0.03 -0.26*** -0.28*** 1.00 0.05 -0.11*** -0.09** 0.03 -0.13*** -0.17*** 0.15*** 0.26*** 0.09** 1.00 -0.41*** 0.03 0.02 -0.03 -0.11*** 0.34*** 0.01 0.01 1.00 -0.03 0.09** 0.16*** 0.13*** -0.29*** -0.25*** -0.12*** 1.00 -0.01 -0.01 -0.07 0.04 -0.02 0.05 1.00 0.08* -0.07 -0.09* -0.01 -0.05 1.00 0.04 -0.08* -0.11*** -0.03 1.00 0.10** -0.06 -0.54*** 1.00 0.27*** 0.09* 1.00 0.37*** 1.00

39

Liquid Assets (t-1)

Cost-income ratio

Securitization

Inflation (t-1)

Capital Ratio

GDP growth

Interest Rate

sdROAA

ROAA

GDP

Total assets

Table 12 Correlation matrix (country level variables)

Stock market capitalization Private monitoring index Shareholder rights index Capital regulatory index

5-bank concentration

Activity restrictions

Privatization index

Securitization

Inflation (t-1)

GDP growth

Interest rate

Securitization GDP GDP growth Inflation (t-1) Interest rate 5-bank concentration HHI H-Statistic Stock market capitalization Capital regulatory index Activity restrictions Private monitoring index Basel I Privatization index Shareholder rights index Rule of law

1.00 0.27*** -0.15*** -0.01 0.11* 0.02 0.06 0.21*** 0.28*** 0.11* -0.32*** 0.08 -0.23*** 0.06 0.04 0.21*** 1.00 -0.41*** 0.03 0.02 -0.31*** -0.24*** 0.26*** 0.20*** 0.04 -0.26*** 0.01 -0.06 0.02 0.01 -0.08** 1.00 -0.03 0.09** 0.09** 0.02 -0.41*** 0.09** 0.12*** -0.09** 0.30*** 0.08** 0.30*** 0.45*** 0.24*** 1.00 -0.01 0.02 0.01 0.03 -0.14*** 0.06 0.04 0.01 0.11** 0.12*** 0.04 -0.05 1.00 0.07 0.08* -0.02 0.12*** 0.08* -0.11** 0.13*** -0.26*** 0.09** 0.09** 0.09** 1.00 0.97*** -0.79*** 0.09** 0.16*** -0.01 -0.04 -0.14*** 0.17*** -0.08** 0.21*** 1.00 -0.73*** 0.05 0.12*** -0.03 -0.10** -0.11*** 0.11*** -0.18*** 0.25*** 1.00 0.03 -0.22*** -0.06 -0.14*** -0.07* -0.21*** -0.13*** -0.28*** 1.00 0.37*** -0.53*** 0.34*** -0.31*** 0.34*** 0.46*** 0.31*** 1.00 -0.67*** 0.51*** 0.01 0.43*** 0.56*** 0.64*** 1.00 -0.31*** 0.01 -0.15*** -0.46*** -0.83*** 1.00 0.01 0.43*** 0.85*** 0.33*** 1.00 -0.08* 0.01 0.12*** 1.00 0.51*** 0.17*** 1.00 0.41*** 1.00

40

Rule of law

H-Statistic

Basel I

GDP

HHI