Securitized banking and the Euro financial crisis: Evidence from the Italian banks risk-taking

G Model JEB-5675; No. of Pages 16

ARTICLE IN PRESS Journal of Economics and Business xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

Journal of Economics and Business

Securitized banking and the Euro financial crisis: Evidence from the Italian banks risk-taking Francesca Battaglia a,1, Angela Gallo b,∗, Maria Mazzuca c,2 a Dipartimento di Studi Aziendali e Quantitativi, Università Parthenope, Via Generale Parisi 13, 80132 Napoli, Italy b Dipartimento di Studi Aziendali, Università degli Studi di Salerno, Via Ponte don Melillo, 84084 Fisciano (SA), Italy c Dipartimento di Scienze Aziendali, Università della Calabria, Via p. Bucci cubo 3C, 87036 Rende (CS), Italy

a r t i c l e

i n f o

Article history: Received 6 May 2013 Received in revised form 15 February 2014 Accepted 18 February 2014 JEL classification: G12 G21 G24 Keywords: Securitization Bank equity risk Systematic risk

a b s t r a c t This research explores the effects of securitization on banks equity risk exposure. A widespread opinion before the crisis of 2007–2008 was that securitization enhances financial stability. We provide empirical evidence of the impact of securitization on the market’s perception of the originating banks’ risk exposure before and after the crisis, in terms of systematic and idiosyncratic risk. Using a sample of Italian listed banks over the period 2000–2009, we find evidence of increasing systematic and idiosyncratic risk for originating banks, in particular in the post-crisis period. We also find that securitization increases the probability of the originator banks to contribute to a market crisis. © 2014 Elsevier Inc. All rights reserved.

1. Introduction Over the past decade and prior to the sub-prime financial crisis, the amount of securitization activity has impressively expanded, both in terms of the development and amount of innovative and sophisticated instruments to transfer risk and in the number of countries using such techniques.

∗ Corresponding author. Tel.: +39 089963122. E-mail addresses: [email protected] (F. Battaglia), [email protected] (A. Gallo), [email protected] (M. Mazzuca). 1 Tel.: +39 0815474852. 2 Tel.: +39 0984492276. http://dx.doi.org/10.1016/j.jeconbus.2014.02.003 0148-6195/© 2014 Elsevier Inc. All rights reserved.

Please cite this article in press as: Battaglia, F., et al. Securitized banking and the Euro financial crisis: Evidence from the Italian banks risk-taking. Journal of Economics and Business (2014), http://dx.doi.org/10.1016/j.jeconbus.2014.02.003

G Model JEB-5675; No. of Pages 16 2

ARTICLE IN PRESS F. Battaglia et al. / Journal of Economics and Business xxx (2014) xxx–xxx

This growth has been exponential also outside the US, recording strong growth rates in Asia and Europe (see European Central Bank, 2008a), even if the credit crisis that broke out in 2007 is having a strong negative impact on the securitization market with a large decline in securitization activity (Joint Forum, 2008). The macroeconomic factors behind this expansion can be recognized in the financial market globalization, technological and financial innovations, and the general trend toward a more market-based financial system. The main purposes for the use of securitization are to obtain additional funding and transfer risk to third-party investors, generating fee income, managing profits, or minimizing regulatory capital requirements. The literature has widely investigated the impact on banks’ risk profiles of securitization activity depending on the above motivations (among others, Allen & Carletti, 2006; Ambrose, LacourLittle, & Sanders, 2005). In this research we mainly focus on securitization as one of the main techniques to manage credit risk, since banks have dramatically increased their risk transfer activities prior to the financial crisis. Although, in principle, a properly done transfer of risk should reduce the banks’ risks, the empirical evidences are mixed. On one hand, banks can shift risks outside their balance sheet as well as to achieve portfolio and funding diversification by means of transfer risk activity (European Central Bank, 2008b). On the other hand, securitization could also lead banks to take on additional risks by acquiring credit risk on the market or adopting more risky funding strategies. We examine the relationship between securitization activity and originator banks’ systematic and idiosyncratic risk over the period 2000–2009 by focusing on a sample of Italian listed banks Although the topic have been widely investigated before, this research contributes to the empirical literature on asset securitization and bank risks in several respects. First, the period of the analysis allows us to compare potentially different impacts on systematic and idiosyncratic before and after the crisis broke out. Secondly, since the prior literature interprets the beta as a measure of systematic risk but also as a proxy of systemic risk, we use a different measure of systemic risk to verify this interpretation. In particular, in our test we use the marginal expected shortfall (MES) as a measure of the systemic risk, defined by Acharya et al., 2010 as a bank’s losses in the tail of the aggregate banking sector’s loss distribution. The difference between MES and beta arises from the fact that systemic risk is based on tail dependence rather than average covariance, so that it better fits the definition of systemic risk in terms of expected losses of each financial institution in a future systemic event in which the overall financial system is experiencing losses. Thirdly, despite the importance of the Italian securitization market, there is a research void on it compared to other European countries. Our results provide evidence of positive effects of securitization on both systematic and idiosyncratic risk; in addition, after 2007 these increases appear to be relatively higher. We also find that securitization increases the probability that the analyzed banks to contribute to a market crisis, but we find no difference in the comparison of the pre-crisis and post-crisis period. This suggests that the risky exposures of these banks are still as high as before the crisis with severe implications for financial stability. The remainder of the paper is organized as follows. In Section 2, we discuss the relevant literature. In Section 3, we describe the estimation framework, sample and data, and variables. In Section 4, we present and discuss the empirical analysis and its results. In Section 5, we debate the results of the robustness tests. In Section 6, we conclude. 2. Literature review The literature considers the phenomenon of securitization from different points of view. A first stream of studies deals with the effects of securitization on the banks’ lending activity and on the monetary policy (Altunbas, Gambacorta, & Marques-Ibanez, 2010; Estrella, 2002; Loutskina & Strahan, 2009). Another strand focuses on the role that securitization has on banks’ risk-taking behavior. Jiangli and Pritsker (2008) use US data for bank holding companies and find that banks active in the securitization market tend to have lower insolvency risk and higher profitability. Rajan (2005) stresses that more market-based pricing exacerbates the incentive structures driving banks and institutional investors, which could (under extreme circumstances) lead to excessive risk taking behavior. Casu Please cite this article in press as: Battaglia, F., et al. Securitized banking and the Euro financial crisis: Evidence from the Italian banks risk-taking. Journal of Economics and Business (2014), http://dx.doi.org/10.1016/j.jeconbus.2014.02.003

G Model JEB-5675; No. of Pages 16

ARTICLE IN PRESS F. Battaglia et al. / Journal of Economics and Business xxx (2014) xxx–xxx

3

et al., 2011, focusing on a sample of US bank holding companies (BHC) during the period 2001–2007, show that bank credit-risk taking behavior (measured as the ratio of risk-weighted assets to total assets) is negatively associated with securitization (explained by the “recourse hypothesis”). When they examine securitizations by the type of underlying assets, they find that the negative relationship is primarily driven by securitizations of mortgages and home equity lines of credit, while all other types of securitizations have no significant impact on bank credit-risk taking behavior. Casu et al. (2013) study the relationship between securitization and bank performance, taking into account both risk and profitability. Analyzing the period 2001–2008 and focusing on a sample of US BHC, they employ a propensity score matching approach to compare the ex post (securitization) performance of securitizer banks to the performance of a counterfactual group of equivalent non-securitizer banks (what would have happened to the securitizing banks if they had not securitized). As indicators of banks performance, the authors use the cost of funding, some risk measures (such as non-performing loans ratio or loan loss provision ratio), some operating performance measures (such as ROA or net interest margin). The analysis does not find evidence that securitization has a beneficial impact upon bank performance, as securitizers would have had comparable cost of funding, credit risk and profitability had they not securitized. Franke and Krahnen (2005) analyze the impact of CDO (Collateralized debt obligations) transactions on the default risk exposure of the originating bank. Such risk effects are measured as the impact on the bank’ default losses and on its stock beta (which proxies the systematic risk in the stock market). Using 73 securitization announcements of 27 European banks between 1999 and 2002, they provide empirical evidence that securitization has positive effects on the systematic stock market risk of the bank measured by its beta. They suggest that the risk-reduction effect of securitization is undermined because banks reinvest liquid capital into riskier projects. Moreover, they propose that risk-reduction by means of securitization is basically determined by the technique of tranching the securitization’s issues. Hence, a post-event increasing beta should result from the fact that the first-loss piece exhibits a higher probability of failure than less risky senior tranches being transferred to external investors. Hansel and Krahnen (2007) confirm the previous findings, showing that the credit risk transfer activity enhances the systematic risk, proxied by the equity beta of the issuing bank and that overall credit securitization increases the bank’s risk appetite. Udhe and Michalak (2010) study securitization and systematic risk in the European banking sector, by analyzing a sample of (cash and synthetic) European listed banks over the period 1997–2007. They find that credit risk securitization has a positive impact on the increase of European banks’ systematic risk, proxied by banks’ beta. In addition, they find that the increase in systematic risk is more relevant for larger banks that repeatedly engage in securitization. Wu et al. (2011) analyze asset securitization and banks’ risk exposure for a sample of US bank holding companies over the period 2002–2007, to investigate whether the market perceives asset securitization as increasing or reducing a bank’s risk. They measure the market perceived risk in terms of banks’ equity risk and distinguish between systematic and idiosyncratic risk. Their findings show that the market seems to view asset securitization as reducing banks’ systematic risk exposure, but there is no evidence of increasing idiosyncratic risk. Moreover, they find that larger banks tend to have higher systematic risk and lower idiosyncratic risk because of diversification. Finally, they identify significant structural break in 2007, when securitizing banks experienced jumps in both systematic and idiosyncratic risks. A number of recent papers investigate the effects of securitization on the financial system stability. following the 2007–2008 credit turmoil. For Shin (2008), securitization has proven to be deleterious from a financial stability standpoint because it allows banks to overextend their balance sheet (for a given level of capital) and lower their credit standards. Based on Allen and Gale (2004), Allen and Carletti (2006) show that credit risk transfer could produce a reduction of welfare through the creation of contagion in others. However, several papers emphasize the benefits deriving from securitization activity because of the opportunity to smooth out the risk among many investors. In other words, credit risk can be more easily and potentially widely transferred across the financial system, allowing banks to hold less risk simply due to diversification and more tradability (Berger et al., 2010). The transfer of credit risk can produce a more efficient use of bank’s capital and a reduction in the cost of raising capital for loan intermediation (Duffie, 2007). Nijiskens and Wagner, (2011) argue that securitization allows banks to shed idiosyncratic exposures, such as the specific risk associated with their area of Please cite this article in press as: Battaglia, F., et al. Securitized banking and the Euro financial crisis: Evidence from the Italian banks risk-taking. Journal of Economics and Business (2014), http://dx.doi.org/10.1016/j.jeconbus.2014.02.003

G Model

ARTICLE IN PRESS

JEB-5675; No. of Pages 16 4

F. Battaglia et al. / Journal of Economics and Business xxx (2014) xxx–xxx

lending and expose them to a bigger funding risk, which can be considered mostly systemic in nature, as current events have shown. The idiosyncratic share in a bank’s risk may also be lowered because banks may hedge any undiversified exposures they may have by buying protection using CDS, while simultaneously buying other credit risk by selling protection in the CDS markets. Banks may thus end up being more correlated with each other (Acharya & Yorulmazer, 2007; Elsinger, Lehar, & Summer, 2006). To conclude the literature discussion, we add some final remarks about the sample studied in previous works. To date, Agostino and Mazzuca (2011) and Battaglia and Mazzuca (2013) are the only authors who have focused on the Italian securitization market by testing the determinants of banks’ securitization and the effects of securitizations on banks’ liquidity risk respectively. Few other studies refer to specific European countries (Cardone-Riportella, Samaniego-Medina, & Trujillo-Ponce, 2010; Martínez-Solano, Yagüe-Guirao, & López-Martínez, 2009, both considering the Spanish market). It seems worth considering other geographical contexts with differently developed capital markets, different banking sector structures. 3. Estimation framework, data and variables In this section, first we discuss our empirical framework, then we examine the sample and the data, and finally we focus on the dependent variables (changes in banks systematic and idiosyncratic, and systematic risks), the identification/explanation of the key independent variables and the control variables. 3.1. Estimation framework Our empirical framework analyze the relationship between asset securitization and bank’s risk profile. First, following the existing literature, we measure the market perceived risk of banks by means of the equity beta. Second, we distinguish between systematic and idiosyncratic risk, by using two complementary approaches (i.e. a simple Capital Asset Pricing Model and the methodology used by Campbell et al., 2001). Successively, following Acharya et al. (2010), we focus on systemic risk employing the marginal expected shortfall (MES). MES has been originally proposed by Tasche (2000), and later used by Yamai and Yoshiba (2002). MES is defined as the expected equity loss per dollar invested in a particular bank if the overall market decline by a certain amount. One example of this approach is provided in Engle and Brownlees (2010). The banks with the highest MES are the banks that contribute the most to the market decline during a crisis. In comparison with other measures of firm-level risk, MES have shown a higher predictive power in detecting a bank’s contribution to a crisis (Acharya et al., 2010).3 To address our research, we regress changes in banks risks on a securitization dummy, a previous securitization dummy and on a set of control variables. Since our aim is to model the probability of a securitizing bank to become riskier during the analyzed period on a number of factors, we choose to apply a probit model (Altunbas et al., 2010). In particular, following Berger et al., 2010a; Berger et al., 2010b, we adopt an ordered model, because the employing of an ordinal dependent variable allows us to analyze the behavior of the bank i in terms of probability, distinguishing between changes in bank behavior and relatively constant behavior.4 Moreover, given that on average the return volatility of

3 The literature on systemic risk is relatively recent and can be broadly divided into two groups of studies, those taking a structural approach using contingent claims analysis of the financial institution’s assets (Gray, Merton, & Bodie, 2009; Lehar, 2005), and those taking a reduced-form approach focusing on the statistical tail behavior of institutions’ asset returns (Adrian and Brunnermeier, 2008; De Jonghe, 2010; Hartmann, Straetmans, & de Vries, 2005; Huang, Zhou, &, Zhu, 2010; Segoviano and Goodhart, 2009). Therefore, MES is not the only systemic risk measures currently proposed. Among others, there is the CoVaR, tail betas and measures based on credit default swaps. 4 Differently from Berger, Bouwman et al. (2010) and Berger, Molyneux et al. (2010), we adopt an ordered probit model instead of a logit one. This is because our dependent variable can be considered as a synthesis of a latent continuous variable. As a robustness check we also run a logit model and we find consistent results.

Please cite this article in press as: Battaglia, F., et al. Securitized banking and the Euro financial crisis: Evidence from the Italian banks risk-taking. Journal of Economics and Business (2014), http://dx.doi.org/10.1016/j.jeconbus.2014.02.003

G Model JEB-5675; No. of Pages 16

ARTICLE IN PRESS F. Battaglia et al. / Journal of Economics and Business xxx (2014) xxx–xxx

5

our sample is approximately 32% over the investigated period, we focus on changes (cut-off) of 30%.5 In particular, our dependent variables take on the following values:

a. 1 if the bank experienced a drop in risk-taking (relative to the previous year) of more than 30% (DECR); b. 2 if the banks’ risk- taking moved within a range of ±30% (CONST); c. 3 if the banks’ risk-taking increased by more than 30% (INCR).

We estimate the following model specification:  Pr(yi,t = j) = ˚(ˇXi,t + ctrli,t−1 ),

j = 1, 2, 3

(1)

where the ordered outcomes are modeled to arise sequentially as the latent variable, y*, crosses progressively higher thresholds.6 In detail, yit = 1, 2 or 3 depending on whether the change in risk-taking of bank i at time t respectively decreases, remains within a definite range or increases; Pr is the probability, ˚ is the standard cumulative normal probability distribution, Xi,t is the vector of securitization (dummy) variables and ctrli,t−1 is the vector of the control variables (all described in subsection 3.3); as usual, ˇ and  parameters are estimated by maximum likelihood. In an ordered probit model the sign of the regression parameters, ˇ, can be immediately interpreted as determining whether the latent variable, y*, increases with the regressor. If ˇj is positive, then an increase in xij necessarily decreases the probability of being in the lowest category (yi = 1) and an increase the probability of being in the highest category (yi = 3). So, for example, a positive coefficient for the securitization dummy or the previous securitization dummy indicates that an increase of these variables corresponds to an increase of the probability for the bank of belonging to the upper risktaking category, i.e. greater than 30%. On the opposite, a negative coefficient indicates that an increase of the securitization variables determines a decrease in the probability for bank i to belong to the higher risk-taking class. Following Wu et al. (2011); Berger et al. 2010a; and Berger et al. 2010b, to address a potential endogeneity concern in our analysis and recognizing that this may not be sufficient, we use lagged control variables (i.e. bank-specific attributes from the previous period but dummy variables from the current period). As variables from the previous period can be viewed as predetermined, this approach help us to lax the endogenous variables problem. Furthermore, it allows us to take into account that there is a time lag for bank-specific variables to affect a bank’s equity return.

3.2. Sample and data We select all the Italian banks having placed at least one cash securitization during the period we considered (2000–2009) based on the securitization. it database, a site providing information on all the cash securitizations carried out in Italy from 1999 onwards. From this database we draw a list of deals solely originated from banks. Banks’ financial statements (balance sheets and income statements) and the measure of capital for regulatory purposes are drawn from the Bankscope-Bureau van Djik database. Our sample includes all intermediaries present in the supervisory register of the Bank of Italy (according to the article No. 106 of TUB, the Italian Banking Law) and classified as commercial banks or savings banks. In particular, we only consider securitizing listed banks because our dependent variables are market-based risk measures. This implies a small sample of banks and it may be a concern for our analysis. However, our sample size is in line with those of other studies referring to the most developed European securitization markets in terms of volumes (see Cerrato et al., 2012 for the UK market; see Martin-Oliver and Saurina, 2007 for the Spanish market). Due to the lack of values in different years for the different variables, the final sample is composed of 83 securitizations undertaken by 21 listed banks over the period 2000–2009. Please cite this article in press as: Battaglia, F., et al. Securitized banking and the Euro financial crisis: Evidence from the Italian banks risk-taking. Journal of Economics and Business (2014), http://dx.doi.org/10.1016/j.jeconbus.2014.02.003

G Model

ARTICLE IN PRESS

JEB-5675; No. of Pages 16 6

F. Battaglia et al. / Journal of Economics and Business xxx (2014) xxx–xxx

Table 1 The Italian securitization market over the period 2000–2009: deal numbers and issuance volumes. Year

Deal numbers

Issuance volume (millions of euros)

Issuance volumes per deal

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

25 59 41 40 39 40 39 29 30 26

12.086 33.967 30.606 30.141 35.028 40.804 35.707 33.919 n.a n.a

483.44 575.71 746.49 753.53 898.15 1020.1 915.56 1169.62 n.a n.a

Source: Securitization.it.

Table 1 reports descriptive statistics for the Italian securitization market (listed and not-listed banks) since its introduction in 1999. The deal numbers per year increased rapidly in the early 2000s, reaching a peak in 2001 with 59 transactions placed. From 2001 to 2006, the number of average transactions per year was 43. Since 2001, the issuance volumes have consistently been significant and have never fallen below the threshold of 30 billion euros. Only in 2007, the deal numbers dropped, as expected, because of the financial crisis. 3.3. Variables 3.3.1. Dependent variables: systematic risk and idiosyncratic risk To capture the effects on the banks’ systematic risk we use the asset beta. In particular, we estimate the beta by adopting a simple Capital Asset Pricing Model (CAPM). In our application, the CAPM model is based on the following equation: Ri,t = ˇi,t ∗ RM,t + εit

(2)

where Ri,t are the daily stock market logarithmic abnormal excess returns from each bank i, RM,t are the daily stock market abnormal returns from the broad stock market index M. The term εi,t is the bank specific residual. For each bank i we calculate our systematic component ˇi,t by running separate regressions on daily data for every year from 1999 to 2009. In this way, our risk proxies can be matched with the other variables having yearly frequency. We also consider the idiosyncratic risk as dependent variable. By running different estimations for this risk and systematic risk, we are able to take into account the possibility that the securitization activity affects the two component of the total risk in opposite directions. To take into account the effect of securitization of banks’ specific risk, following Altunbas et al. (2010), we use two complementary approaches. In the first one, according to the standard CAPM, the idiosyncratic component (IDSC1) is constructed as the average of the squared of the unexplained component of each regression for bank i over each year: IDSC1 =

m  (εi,t )2 t=1

m

(3)

where m is the number of trading days in each year. The second approach follows Campbell et al. (2001), who build on Merton (1980), and decompose stock market volatility into total market, banking sector and individual bank level volatility. In particular, by assuming that these different components of stock market returns are orthogonal to one

5

We specify that in a robustness test, we use alternative cut-offs. We decide to adopt an ordered probit model because our dependent variables can be considered as a synthesis of a latent continuous variable. 6

Please cite this article in press as: Battaglia, F., et al. Securitized banking and the Euro financial crisis: Evidence from the Italian banks risk-taking. Journal of Economics and Business (2014), http://dx.doi.org/10.1016/j.jeconbus.2014.02.003

G Model

ARTICLE IN PRESS

JEB-5675; No. of Pages 16

F. Battaglia et al. / Journal of Economics and Business xxx (2014) xxx–xxx

7

another, each risk component can be calculated by means of simple variance decomposition. Therefore, individual bank idiosyncratic risk for year y can be computed as: IDSC2 =

m  (Ri,y,t − RB,y,t )2 t=1

m

(4)

where for each time t, Ri,t and RB,t are the individual bank i and the banking sector logarithmic returns respectively. The idiosyncratic measure of bank risk (IDSC2) is calculated for each year y, where m is the number of daily observations for each year available for bank i. To capture the effects on systemic risk we use the MES (Acharya et al., 2010). MES derives from the concept of Expected Shortfall (ES). Denote by r the day t return of a given index (I), the Expected Shortfall is defined as: ESt (C) = Et (rtl |rtl < C)

(5)

where C is a known threshold. Expected Shortfall is thus a partial moment capturing the expected value of the lower tail of the index distribution (the threshold is generally defined to be negative, or equal to the Value-at-Risk (VaR) at a given confidence level). Notably, as shown by Acerbi and Tasche (2002), ES is a coherent risk measure. Given equation (5), Acharya et al. (2010) and Engle and Brownlees (2010) derive Marginal Expected Shortfall of company i as the derivative of the market Expected Shortfall with respect to company i weight in the market index and ultimately define MES as: MESi,t (C) = E(ri,t |rtl < C)

(6)

where ri,t is the day t return of the bank i. 3.3.2. Key independent variables: securitization dummy and previous securitization dummy To capture the securitization activity placed by the Italian banks we consider two different securitization dummy variables. First, we include a securitization dummy (Sec), coded 1 if the specific bank securitizes in a given year and 0 otherwise (Agostino and Mazzuca, 2011). Next, following the previous literature (Franke & Krahnen, 2005; Udhe & Michalak, 2010), we explicitly consider the repeated/previous securitizations by each sample bank including the previous securitization regressor (Prev sec) that accounts for the previous securitizations undertaken by a bank in the previous years. This variable allows us to take into account the expertise of each bank in the securitization process and it is coded 1 if the bank i placed at least one securitization in the period before the considered year and 0 otherwise. 3.3.3. Control variables We include a set of control variables in the probit Model (1). All of the control variables (except for the dummy variables) are measured in changes. For ease of exposition, we discuss these variables below in levels. To control for specific bank characteristics, we include variables that might explain differences in the bank’s risk taking. In particular, referring to the most recent studies on the topic (Agostino & Mazzuca, 2011; Casu et al., 2011; Jiangli & Pritsker, 2008; Nijiskens & Wagner, 2011), we consider the following control variables: the loans ratio (total loans/total assets, Loans), the equity ratio (equity/total assets %, Equity ratio), the total assets (natural log of total assets, size), the liquidity ratio (liquid assets/customer and short term funding, Liquidity), the impaired loan ratio (impaired loans/gross loans, Imp loans), the Tier1 ratio (Tier1), the return on equity (net income/total equity, Roe). The sign of the expected relation between the previous variables and the dependent ones, varies from one regressor to another, and it is related to the dependent variable (i.e. beta, IDSC1, IDSC2 and MES) that in turn we consider. Please cite this article in press as: Battaglia, F., et al. Securitized banking and the Euro financial crisis: Evidence from the Italian banks risk-taking. Journal of Economics and Business (2014), http://dx.doi.org/10.1016/j.jeconbus.2014.02.003

G Model

ARTICLE IN PRESS

JEB-5675; No. of Pages 16 8

F. Battaglia et al. / Journal of Economics and Business xxx (2014) xxx–xxx

Table 2 Summary statistics for the dependent variables, the key-independent variables, and the control variables. Variable

Obs.

Mean of level

Mean of change

Dependent variables

Beta MES IDSC1 IDSC2

166 166 165 167

0.575886568 −0.041979 0.0303068 0.0441

0.2058852 −0.6065153 0.2187314 0.2762326

Key independent variables

Sec Prev sec

210 210

0.3428571 0.7238095

Control variables

Loans Equity ratio Size Liquidity Imp loans Tier1 Roe

201 185 203 177 183 180 202

0.3422239 0.0741959 6.931395 13.122176 1.752285 5.9345976 0.3700048

n.a. n.a. 0.0925363 0.0345064 0.0068586 0.1145027 0.5554443 0.0488155 −0.3143841

Notes: Sec is a dummy coded 1 if a bank places at least one securitization in a year, zero otherwise; Prev sec is a dummy coded 1 if a bank places at least one securitization in the period before the analyzed year, zero otherwise; Size is measured as ln of total assets; the other variables are described in Section 3.3. Continuous variables are winsorised at 5%. While regressions are run in changes, we report means of levels and changes of the variables.

4. Empirical analysis and results Table 2 presents some relevant statistical information for our variables. While the regressions are run in changes, we report means of levels and changes for all the variables. We winsorise (at 5%) our variables with large positive and negative outliers, to prevent extreme values from biasing the results of our study, without loss of observations (Barth et al., 2006). In all cases, the observations are clustered at the bank level. We estimate two specifications of the baseline Model 1. In detail, first we run the following regression for each dependent variable: Model 1: Pr(yit = ji ) = ˚(ˇ1 sec1it + ˇ2 prev sec2it + ˇ3 loans3i(t−1) + ˇ4 equity ratio4,i(t−1) +ˇ5 size5,i(t−1) + ˇ6 liquidity6,i(t−1) + ˇ7 imp loans7,i(t−1) + ˇ8 tier18,i(t−1) + ˇ9 roe9,i(t−1) )

(7)

where yit is our dependent variable (i.e. beta, idsr1, idsr2, mes), sec and prev sec are the key independent dummy variables and ˚ is the cumulative distribution function of the standard normal distribution. Finally, to control for potential cycle effects, common to all banks but varying over the analyzed period, we include time effects that we omit in Eq. (7) for ease of exposition. To test whether the structural change occurs in the relationship between the securitization and the banks’ risk exposure over the period 2007–2009, we estimate a second specification of Model 1, where we include an interaction variable between the securitization and the dummy CRISIS, that takes the value of 1 during the years 2007–2009 and zero in the previous ones (2000–2006).7 In particular, for each dependent variable, we rerun the following regression8 : Model 2: Pr(yit = ji ) = ˚(+ˇ1 sec1it ∗ crisis + ˇ2 prev sec2it + ˇ3 loans3i(t−1) + ˇ4 equity ratio4,i(t−1) +ˇ5 size5,i(t−1) + ˇ6 liquidity6,i(t−1) + ˇ7 imp loans7,i(t−1) + ˇ8 tier18,i(t−1) + ˇ9 roe9,i(t−1) )

(8)

7 We selected the year 2007 as the start time of the crisis for several reasons. First, the crisis has started in June 2007, with losses from subprime mortgages exposing large scale vulnerabilities, and in August 2007 the turmoil has spread to interbank markets, signaling the advent of a broader financial market crisis (Bank for International Settlements, 2009). Second, the structured finance suffered the crisis effects starting from the second half of 2007 because the majority of investors pulled back due to mark to market losses. These crisis effects were particularly severe in Italy, where the second half of 2007 was one of the slowest period ever experienced (Moody’s Investors service, 2008). Therefore, since we use annual data, to capture the effects also for the year 2007 we specified our crisis sub-period from 2007 to 2011. 8 Also in Eq. (6) we control for time effects that we omit for ease of exposition.

Please cite this article in press as: Battaglia, F., et al. Securitized banking and the Euro financial crisis: Evidence from the Italian banks risk-taking. Journal of Economics and Business (2014), http://dx.doi.org/10.1016/j.jeconbus.2014.02.003

G Model JEB-5675; No. of Pages 16

ARTICLE IN PRESS F. Battaglia et al. / Journal of Economics and Business xxx (2014) xxx–xxx

9

We specify that the statistical analysis of the daily stock prices’ time series reveals the presence of a structural break in year 2007. We investigate the presence of that break through the implementation of one of the methods ad hoc used. Generally, these techniques test the presence of changes in the regressions’ coefficients through the use of the F-statistics. We decide to employ the Chow test (Chow, 1960), which tests the null hypothesis of breaks’ lack. The value of F-statistics allows us to reject the null hypothesis of breaks’ lacks for both the models with a significant level lower than 1%. As in the probit models the sign of the coefficients provides only information on the direction of the impact of each determinant (independent variables) on the probability for the bank i to became riskier, to obtain the direct impact of each explanatory variable on the probability that the event of interest occurs (which depends on the value taken by all the other regressors included in the equation), one has to compute the marginal effect of each explanatory variable by assuming a given (representative) value for all other variables. In the present work, for the computation of each partial effect, we consider all other variables at their sample mean. We do not need to standardize regressors, because they are all expressed in the same measurement units.9 Columns 3 of Tables 3–6 show the marginal effect referring to the third category, which corresponds to the riskier class, where changes are >30%. Table 3 presents estimation results of our models. Panel A and B reports the ordered-probit estimations for Models 1 and 2, respectively. The robust standard errors are reported in columns 3 and 6. Columns 4 and 7 report the marginal effects. The pseudo R-square is 0.1099 and 0.0975, respectively. In the estimates of Model 1 we find a positive and significant coefficient for the dummy sec and a negative and positive coefficient for the dummy prev sec. This suggests that securitizing banks have a higher probability to became riskier in terms of correlation with the market, but this relationship is negatively influenced by the fact that the same banks undertaken securitization activity in the previous years and is almost unaffected by our control variables. Among these, the roe is the only variable with a positive and significant coefficient. The highest marginal effect is attributable to the size variable (positive but insignificant) allowing us to control for potential size effect, that is to say, the possibility that larger banks tend to have higher systematic risk exposure and lower idiosyncratic risk because of diversification. Then, this variable allows us to distinguish between the systematic component of beta due to the securitization from the component related to the size. In the Model 2, we find that the coefficient of the interaction variable Int sec crisis is positive and highly significant. This is consistent with the finding of Wu et al. (2011), even if they only confirmed the existence of a structural change in 2007, whereas we find that securitization leads to a higher systematic risk in the years after the crisis broke out in 2007 until 2009, compared to the pre-crisis period. In Tables 4 and 5, we report the results of our estimations when the idiosyncratic risk is assumed as a dependent variable, for the IDSR1 and IDSR2 measures. In the first case the dependent variable is estimated from the same model as beta and we find again that securitizing banks have a higher probability of became riskier in terms of specific risk, but in this case the prev sec variable is highly significant and positive. It suggests that the idiosyncratic risk is higher for banks that have been more involved in the securitization market in the previous years. This is consistent with the idea that securitization implies higher systematic risk in the year of the operation, but also that it will strongly affect the idiosyncratic risk in the subsequent years, likely depending on the extent of the risk retention for the originating bank. If combined with the evidence in Table 3 where we find a negative relationship between beta and previous securitization, we could say that the total risk of a securitizing bank is especially driven by previous operations in the component of the idiosyncratic risk. Our control variables are now significant in the case of loans, size, Imp loans and Roe. Again the size variable has the highest marginal effect, controlling for bias due to the larger banks. When considering the estimation of Model 2, the results confirm the increase in the impact of securitization on the idiosyncratic risk after 2007. The Table 5 reports results by adopting a different measure of specific risk, which is not model-based but market-based. This finding confirms the evidence for IDSR2, so that we can draw conclusion on the robustness of our results to different specification of specific risk. Table 6 reports results for the estimation of our models when the MES is assumed as dependent variable. In Model 1 we find that the coefficient for the dummy sec is positive and significant,

9

In fact, with the exception of the dummy variables, ratios are expressed as a percentage.

Please cite this article in press as: Battaglia, F., et al. Securitized banking and the Euro financial crisis: Evidence from the Italian banks risk-taking. Journal of Economics and Business (2014), http://dx.doi.org/10.1016/j.jeconbus.2014.02.003

Panel A: Model (1) Coefficients

Numero di oss. Log pseudolikelihood Numero di banche (clusters) Prob > chi2 Pseudo R2

−0.4297302 0.0289239 0.2038907 3.023402 0.126492 0.086282 0.003573 0.1218579 111 −102.17629 21 0 0.1099

Robust standard errors

***

0.197116

**

0.2367144

***

0.1141654 0.1905357 8.471673 0.1023221 0.084504 0.0847794 0.0468882

Marg. effects

Coefficients

Sig.

Robust standard errors

Marg. effects

0.1764859 −0.161395 0.0103955 0.0732803 1.08664 0 .0454624 0.0310106 0.0012842 0.0437969

0.7639893 −0.3037583 0.012981 0.1839039 0.028431 0.1084491 0.0845184 0.0596434 0.0743944

*** *

***

0.3041138 0.3616014

0.2930349 −0.1120885

0.1283913 0.1882729 9.859008 0.1094299 0.0739571 0.1089864 0.042454

0.0046189 0.654371 0.101164 0.0385886 0.0300736 0.0212224 0.0264712

111 −103.80855 21 0 0.0957

Dependent variable – Panels A and B: changes in beta ( beta) of 30%. Notes: (a) we estimate ordered probit models for changes in beta of 30%. In Panel A we run the Model 1 (explanatory variables: sec, prev sec, loans, equity ratio, size, liquidity, imp loans, tier1 and roe), while in Panel B we run the Model 2 (int sec crisis, prev sec, loans, equity ratio, size, liquidity, imp loans, tier1 and roe). The dependent variable takes on the value 1 if there was a drop in the beta of at least 30% relative to the previous year, it takes on the value 2 if beta remained within the interval ±30%, and it takes on the value 3 if there was an increase in beta of more than 30%; (b) all the control variables are lagged 1 year; (c) all the explanatory variables (except the dummy variables) are measured in changes; (d) columns 2 of Panel A and Panel B report robust standard errors; (e) columns 3 of Panel A and Panel B present the marginal effects referring to the outcome 3, that is the upper beta category (changes > 30%); (f) time effects are included in all estimations. * Significant at 10%. ** ***

Significant at 5%. Significant at 1%.

ARTICLE IN PRESS

Control variables Loans Equity ratio Size Liquidity Imp loans Tier1 Roe

0.4796293

Sig.

F. Battaglia et al. / Journal of Economics and Business xxx (2014) xxx–xxx

Key independent variables Sec Int sec crisis Prev sec

Panel B: Model (2)

G Model

Dependent variable beta: Pr(riskieri = 3)

JEB-5675; No. of Pages 16

10

Please cite this article in press as: Battaglia, F., et al. Securitized banking and the Euro financial crisis: Evidence from the Italian banks risk-taking. Journal of Economics and Business (2014), http://dx.doi.org/10.1016/j.jeconbus.2014.02.003

Table 3 Ordered probit models focusing on beta changes of 30% over the period 2000–2009.

Panel A: Model (1) Coefficients

Control variables Loans Equity ratio Size Liquidity Imp loans Tier1 Roe Number of obs. Log pseudolikelihood Number of banks (clusters) Prob > chi2 Pseudo R2

Marg. effects

Sig.

Robust standard errors

0.3751938

*

0.2281575

0.1764859

0.769781

***

0.23318

0.2121372

0.7020236 0.5718481

0.3613679 −0.0311057 11.85772 −0.0258383 0.0811677 0.1167495 −0.0327131

***

0.1258889 0.2171385 8.52791 0.667231 0.376213 0.2525908 0.076503

0.1189761 −0.0102412 10.48877 −0.0085069 0.0267235 0.0384384 −0.0107704

0.3658739 −0.0435407 9.36615 −0.0261301 0.0953824 0.0741012 −0.0336874

111 −109.98074 21 0 0.0885

***

**

***

Coefficients

Sig.

*** *

***

Robust standard errors

Marg. effects

0.4076063 0.222874

0.2591893 0.1657805

0.1306204 0.2206821 10.96817 0.0598759 0.0372016 0.26018 0.0076728

0.1204325 −0.014332 7.666278 −0.0086011 0.0313964 0.0243914 −0.0110887

111 −109.54897 21 0 0.0921

Dependent variable – Panels A and B: changes in IDSR1 ( IDSR1) of 30%. Notes: (a) we estimate ordered probit models for changes in IDSR1 of 30%. In Panel A we run the Model 1 (explanatory variables: sec, prev sec, loans, equity ratio, size, liquidity, imp loans, tier1 and roe), while in Panel B we run the Model 2 (int sec crisis, prev sec, loans, equity ratio, size, liquidity, imp loans, tier1 and roe). The dependent variable takes on the value 1 if there was a drop in the IDSR1 of at least 30% relative to the previous year, it takes on the value 2 if IDSR1 remained within the interval ±30%, and it takes on the value 3 if there was an increase in IDSR1 of more than 30%; (b) all the control variables are lagged 1 year; (c) all the explanatory variables (except the dummy variables) are measured in changes; (d) columns 2 of Panel A and Panel B report robust standard errors; (e) columns 3 of Panel A and Panel B present the marginal effects referring to the outcome 3, that is the upper IDSR1 category (changes > 30%); (f) time effects are included in all estimations. * ** ***

Significant at 10%. Significant at 5%. Significant at 1%.

ARTICLE IN PRESS

Key independent variables Sec Int sec crisis Prev sec

Panel B: Model (2)

G Model

JEB-5675; No. of Pages 16

Dependent variable IDSR1: Pr (riskieri = 3)

F. Battaglia et al. / Journal of Economics and Business xxx (2014) xxx–xxx 11

Please cite this article in press as: Battaglia, F., et al. Securitized banking and the Euro financial crisis: Evidence from the Italian banks risk-taking. Journal of Economics and Business (2014), http://dx.doi.org/10.1016/j.jeconbus.2014.02.003

Table 4 Ordered probit models focusing on IDSR1 changes of 30% over the period 2000–2009.

Panel A: Model (1) Coefficients

Number of obs. Log pseudolikelihood Number of banks (clusters) Prob > chi2 Pseudo R2

Robust standard errors

0.3146164

*

0.2295427

0.0908672

0.7752416

**

0.3094656

0.171853

0.2693418 −0.0473965 14.37956 −0.0146974 0.0342943 −0.0108073 −0.0381715

**

0.1096724 0.2327859 8.696345 0.0666743 0.0653333 0.2367853 0.0091696

0 .0747146 −0.0131476 6.762818 −0.004077 0.0095131 −0.0029979 −0.0105887

112 −108.93783 21 0 0.0904

***

**

***

Coefficients

Sig.

0.734868 0.5939006

***

0.2842606 −0.0522415 12.90343 −0.0111598 0.0484022 −0.0508129 −0.039065

***

*

***

***

Robust standard errors

Marg. effects

0.2462679 0.3177273

0.2449506 0.1377075

0.1049536 0.2394585 8.415762 0.0592451 0.0710697 0.2334839 0.0087829

0.0781743 −0.0143669 5.298654 −0.0030691 0.013311 −0.013974 −0.0107432

112 −107.86754 21 0 0.0993

Dependent variable – Panels A and B: changes in IDSR2 ( IDSR2) of 30%. Notes: (a) we estimate ordered probit models for changes in IDSR2 of 30%. In Panel A we run the model 1 (explanatory variables: sec, prev sec, loans, equity ratio, size, liquidity, imp loans, tier1 and roe), while in Panel B we run the model 2 (int sec crisis, prev sec, loans, equity ratio, size, liquidity, imp loans, tier1 and roe). The dependent variable takes on the value 1 if there was a drop in the IDSR2 of at least 30% relative to the previous year, it takes on the value 2 if IDSR2 remained within the interval ±30%, and it takes on the value 3 if there was an increase in IDSR2 of more than 30%; (b) all the control variables are lagged 1 year; (c) all the explanatory variables (except the dummy variables) are measured in changes; (d) columns 2 of Panel A and Panel B report robust standard errors; (e) columns 3 of Panel A and Panel B present the marginal effects referring to the outcome 3, that is the upper IDSR2 category (changes > 30%); (f) time effects are included in all estimations. * Significant at 10%. ** ***

Significant at 5%. Significant at 1%.

ARTICLE IN PRESS

Control variables Loans Equity ratio Size Liquidity Imp loans Tier1 Roe

Marg. effects

Sig.

F. Battaglia et al. / Journal of Economics and Business xxx (2014) xxx–xxx

Key independent variables Sec Int sec crisis Prev sec

Panel B: Model (2)

G Model

Dependent variable IDSR2: Pr (riskieri = 3)

JEB-5675; No. of Pages 16

12

Please cite this article in press as: Battaglia, F., et al. Securitized banking and the Euro financial crisis: Evidence from the Italian banks risk-taking. Journal of Economics and Business (2014), http://dx.doi.org/10.1016/j.jeconbus.2014.02.003

Table 5 Ordered probit models focusing on IDSR2 changes of 30% over the period 2000–2009.

Panel A: Model (1) Coefficients

Control variables Loans Equity ratio Size Liquidity Imp loans Tier1 Roe Number of obs, Log pseudolikelihood Number of banks (clusters) Prob > chi2 Pseudo R2

0.4837307

Sig.

Robust standard errors

**

0.2145443

0.1643511

0.3049763

0.0501183

0.1598064 0.0195394

0.1184184 0.3555952 7.783002 0.3049763 0.0541187 0.148239 0.0162013

0.1194214 −0.405234 8.191812 −0.13281 0.0170457 0.0339461 0.0040302

0.3634363 −1.14161 15.83109 −0.4089718 0.0514672 0.0579279 0.0120371

0.1582578 0.3652774 −1.2395 18.11523 −0.4062286 0.0521382 0.103832 0.0123274 111 −106.36065 21 0 0.1092

*** ** *** * *

Marg. effects

Coefficients

Sig.

*** ** *** *** *

Robust standard errors

Marg. effects

0.4161426 0.3596745

0.0548216 0.0064455

0.0855087 0.5736283 6.790291 0.2239516 0.0499545 0.1830783 0.0180055

0.1203234 −0.3779547 8.551939 −0.1353989 0.0170393 0.0191783 0.0039851

111 −108.22225 21 0 0.0936

Dependent variable – Panels A and B: changes in MES ( MES) of 40%. Notes: (a) we estimate ordered probit models for changes in MES of 40%. In Panel A we run the model 1 (explanatory variables: sec, prev sec, loans, equity ratio, size, liquidity, imp loans, tier1 and roe), while in Panel B we run the model 2 (int sec crisis, prev sec, loans, equity ratio, size, liquidity, imp loans, tier1 and roe). The dependent variable takes on the value 1 if there was a drop in MES of at least 40% relative to the previous year, it takes on the value 2 if MES remained within the interval ±40%, and it takes on the value 3 if there was an increase in MES of more than 40%; (b) all the control variables are lagged 1 year; (c) all the explanatory variables (except the dummy variables) are measured in changes; (d) columns 2 of Panel A and Panel B report robust standard errors; (e) columns 3 of Panel A and Panel B present the marginal effects referring to the outcome 3, that is the upper MES category (changes > 40%); (f) time effects are included in all estimations. * Significant at 10%. ** Significant at 5%. *** Significant at 1%.

ARTICLE IN PRESS

Key independent variables Sec Int sec crisis Prev sec

Panel B: Model (2)

G Model

JEB-5675; No. of Pages 16

Dependent variable MES: Pr (riskieri = 3)

F. Battaglia et al. / Journal of Economics and Business xxx (2014) xxx–xxx 13

Please cite this article in press as: Battaglia, F., et al. Securitized banking and the Euro financial crisis: Evidence from the Italian banks risk-taking. Journal of Economics and Business (2014), http://dx.doi.org/10.1016/j.jeconbus.2014.02.003

Table 6 Ordered probit models focusing on MES changes of 40% over the period 2000–2009.

G Model

ARTICLE IN PRESS

JEB-5675; No. of Pages 16 14

F. Battaglia et al. / Journal of Economics and Business xxx (2014) xxx–xxx

meaning that the securitization increases the systemic risk, while the prev sec is positive but insignificant. Notably, we find a negative and significant coefficient for the size and equity ratio, which can be considered as a rough proxy for leverage ratio. This is consistent with the evidence of Engle and Brownlees (2010) that MES is an increasing function of market size and financial leverage. The results from Model 2 show that the relationship between securitization and MES is unaffected by the financial crisis. Since the financial crisis illustrated how securitization activity in the years before 2007 had negative implications for the financial stability, this result confirms that the consequence of the securitization in terms of higher systemic risk are still there. 5. Robustness tests To further verify our results, we implement a set of robustness checks concerning the model specification and the estimation method. First, we use alternative cut-offs for the dependent variables to check whether our results are sensitive to our choice of 30% cut-off. In particular, we rerun all calculations employing 20 and 40% changes of the dependent variables. The results we obtain generally confirm the findings of the main regressions with reference to the key independent variables. So, we can conclude that our baseline regression estimates are robust to these alternative cut-offs. All estimations and Tables are available upon request. As a second sensitivity check referring to the model specification, we change some control variables. In detail, recognizing that banks of different size classes have different balance sheet compositions (Berger et al., 2005), we rerun all calculations including a size dummy variable, coded 1 if the size of bank i in the year t is below the yearly median bank size and zero otherwise,10 with no change in the main results.11 Then, assuming that the different capitalization of a securitizing bank can affect in a different way its risk-taking, we split our sample into poorly and better capitalized banks and create a new dummy variable (equity ratio dummy) by using as cut-off the yearly median bank’s equity ratio.12 Then we reestimate our regressions, but our findings remain qualitatively unchanged.13 With reference to the robustness test concerning the estimation method, we rerun all our regressions by using an ordered logit model, in order to verify our results are insensitive to the choice of the modeling technique. The general formula for an ordered logit model with three categories expresses the probability of observation i of variable Y falling into category j in year t as: P(Yi,t>j ) =

exp(˛j + ˇXi,t−1 ) 1 + exp(˛j + ˇXi,t−1 )

j = 1, 2

,

(9)

and P(Yi,t ) = 1 − P(Yi,t > 1)

(10)

where Xi,t−1 is the vector of the independent variables for observation i in year t−1, the ˛’s are the intercepts and the ˇ’s are the slope coefficients.14 All estimations and Tables are available upon request. 10

We split our sample into small and large banks using the yearly median bank size as the cut-off. All estimations are available upon request. The dummy equity ratio variable is coded 1 if the equity-ratio of the bank i in the year t is above the yearly median equity ratio and zero otherwise. 13 All estimations are available upon request. 14 In our models, the equations are: 11 12

P(Yi,t = DECR) = 1 − P(Yi,t = CONST ) = P(Yi,t = INCR) =

exp(˛1 + ˇX i,t−1 ) 1 + exp(˛1 + ˇX i,t−1 ) exp(˛1 + ˇX i,t−1 )

1 + exp(˛1 + ˇX i,t−1 ) exp(˛2 + ˇX i,t−1 )

1 + exp(˛2 + ˇX i,t−1 )

.

−

.

(11) exp(˛2 + ˇX i,t−1 )

1 + exp(˛2 + ˇX i,t−1 )

.

(12)

(13)

Please cite this article in press as: Battaglia, F., et al. Securitized banking and the Euro financial crisis: Evidence from the Italian banks risk-taking. Journal of Economics and Business (2014), http://dx.doi.org/10.1016/j.jeconbus.2014.02.003

G Model JEB-5675; No. of Pages 16

ARTICLE IN PRESS F. Battaglia et al. / Journal of Economics and Business xxx (2014) xxx–xxx

15

All regression results using the ordered logit model are qualitatively similar to those obtained employing the Eq. (1). This suggests that our findings are not driven by the selected modeling technique.

6. Conclusions This paper aims at testing the relation between asset securitization and the bank’s equity. The empirical analysis shows several results. First, consistently with Hansel and Krahnen (2007), Nijiskens and Wagner (2011), and Udhe and Michalak (2010) (and contrary to Wu et al., 2011), we find that securitization activity contributes to increase the issuing banks’ risk profile, both in terms of systematic and idiosyncratic risk. However, while the effects on the systematic risk are mainly influenced by the current securitization activity, the effects on the idiosyncratic risk are mainly influenced by the previous securitizations. In addition, the findings show that the financial turmoil started in 2007 caused an increase of both risk measures for securitizing banks.

References Acerbi, C., & Tasche, D. (2002). On the coherence of expected shortfall. Journal of Banking and Finance, 1487–1503. Acharya, V., Pedersen, L., Philippe, T., & Richardson, M. (2010). Measuring systemic risk. NYU: Department of Finance. Acharya, V., & Yorulmazer, T. (2007). Cash-in-the-market pricing and optimal resolution of bank failures. Bank of England (Bank of England working papers 328). Adrian, T., & Brunnermeier, M. (2008). CoVAR. Federal Reserve Bank of New York (Working paper). Agostino, M., & Mazzuca, M. (2011). Empirical investigation of securitization drivers: The case of Italian banks. European Journal of Finance, 17(8), 623–648. Allen, F., & Carletti, E. (2006). Credit risk transfer and contagion. Journal of Monetary Economics, 53, 89–111. Allen, F., & Gale, D. (2004). Financial intermediaries and markets. Econometrica, 72, 1023–1061. Altunbas, Y., Gambacorta, G., & Marques-Ibanez, D. (2010). Does monetary policy affect bank risk-taking?. (BIS working paper, No. 298). Ambrose, B. W., LaCour-Little, M., & Sanders, A. B. (2005). Does regulatory capital arbitrage or asymmetric information drive securitization? Journal of Financial Services Research, 28(1–3), 113–133. Bank for International Settlements. (2009). (79th annual report, Basel). Barth, M. E., Landsman, W. R., Lang, M., & Williams, C. D. (2006). Accounting quality: International accounting standards and US GAAP. SSRN: http://ssrn.com/abstract=897241 Accessed December 2009 Battaglia, F., & Mazzuca, M. (2013). Does securitization affect banks’ liquidity risk? The case of Italy. In J. A. Batten, P. Mackay, & N. Wagner (Eds.), Advances in risk management (pp. 127–147). London: Palgrave MacMillan Series. Berger, A., Bouwman, C. H. S., Kick, T., & Schaeck, K. (2010). Bank risk taking and liquidity creation following regulatory interventions and capital support. Deutsche Bundesbank (Discussion Paper Series 2, No. 05/2010). Berger, A. N., Clarke, G. R. G., Cull, R., Klapper, L., & Udell, G. F. (2005, June). Corporate governance and bank performance: A joint analysis of the static, selection, and dynamic effects of domestic, foreign, and state ownership. (World Bank Policy Research working paper 3632). Berger, A. N., Molyneux, P., & Wilson, J. O. S. (2010). The Oxford Handbook of Banking (pp. 599–625). Oxford: Oxford University Press. Campbell, J. Y., Lettau, M., Malkiel, B. G., & Xu, Y. (2001). Have individual stocks become more volatile? An empirical exploration of idiosyncratic risk. Journal of Finance, 56(1), 1–43. Cardone-Riportella, C., Samaniego-Medina, R., & Trujillo-Ponce, A. (2010). What drives bank securitization? The Spanish experience. Journal of Banking & Finance, 34(11), 2639–2651. Casu, B., Clare, A., Sarkisyan, A., & Thomas, S. (2011). Does securitization reduce credit risk taking? Empirical evidence from US bank holding companies. European Journal of Finance, 17(9–10), 769–788. Casu, B., Clare, A., Sarkisyan, A., & Thomas, S. (2013). Securitization and bank performance. Journal of Money, Credit and Banking, 45(8), 1617–1658. Cerrato, M., Choudhry, M., Crosby, J., & Olukuru, J. (2012). Why do UK Banks Securitize?. (SSRN working paper). Chow, G. (1960). Tests of equality between sets of coefficients in two linear regressions. Econometrica, 28(3), 591–605. De Jonghe, O. (2010). Back to the basics in banking? A micro-analysis of banking system stability. Journal of Financial Intermediation, 19(July (3)), 387–417. Duffie, D. (2007). Innovations in Credit Risk Transfer: Implication for Financial Stability. (Stanford University working paper). Elsinger, H., Lehar, A., & Summer, M. (2006). Using market information for banking system risk assessment. International Journal of Central Banking, 2(1), 137–165. Engle, R. F., & Brownlees, C. T. (2010). Volatility, Correlation and Tails for Systemic Risk Measurement. Stern School of Business, New York University (Manuscript). Estrella, A. (2002). Securitization and the efficacy of monetary policy. Economic Policy Review, (May), 243–255. Federal Reserve Bank of New York. European Central Bank. (2008a). Financial stability review. European Central Banks Monthly Bulletin, (June). European Central Bank. (2008b). Securitization in the euro area. European Central Banks Monthly Bulletin, (February).

Please cite this article in press as: Battaglia, F., et al. Securitized banking and the Euro financial crisis: Evidence from the Italian banks risk-taking. Journal of Economics and Business (2014), http://dx.doi.org/10.1016/j.jeconbus.2014.02.003

G Model JEB-5675; No. of Pages 16 16

ARTICLE IN PRESS F. Battaglia et al. / Journal of Economics and Business xxx (2014) xxx–xxx

Franke, G., & Krahnen, J. P. (2005). Default risk sharing between banks and markets: The contribution of collateralized debt obligations. In M. Carey, & R. Stulz (Eds.), The Risks of Financial Institutions. University of Chicago Press (NBER working paper, forthcoming). Gray, D., Merton, R. C., & Bodie, Z. (2009). New Framework for Measuring and Managing Macrofinancial Risk and Financial Stability. Cambridge, MA: Harvard Business School (Working paper No. 09-015). Hansel, D. N., & Krahnen, J.-P. (2007). Does Securitization Reduce Bank Risk? Evidence From the European CDO Market. Finance Department, Goethe University Frankfurt. Hartmann, P., Straetmans, S., & de Vries, C. (2005). Banking System Stability: A Cross-Atlantic Perspective. (NBER working paper 11698). Huang, X., Zhou, H., & Zhu, H. (2010, November). Systemic Risk Contributions of Large and Complex Financial Institutions. The Federal Reserve Board (Finance and economics discussion series: 2011-08 Screen Reader version). Jiangli, W., & Pritsker, M. (2008). The Impacts of Securitization on US Bank Holding Companies. Joint Forum. (2008). Credit Risk Transfer: Developments from 2005 to 2007. Basle Committee on Banking Supervision, International Organization of Securities Commissions, International Organization of Insurance Supervisors. Lehar, A. (2005). Measuring systemic risk: A risk management approach. Journal of Banking and Finance, 29, 2577–2603. Loutskina, E., & Strahan, P. E. (2009). Securitization and the Declining Impact of Bank Finance on Loan Supply: Evidence from Mortgage Acceptance Rate (p. 11983) (National Bureau of Economic Research working paper). Martínez-Solano, P., Yagüe-Guirao, J., & López-Martínez, F. (2009). Asset securitization: Effects on value of banking institutions. European Journal of Finance, 15(2), 119–136. Martin-Oliver, A., & Saurina, J. (2007). Why do Banks Securitize Assets?. Banco de Espana. Merton, R. C. (1980). On estimating the expected return on the market: An exploratory investigation. Journal of Financial Economics, 8(4), 323–361. Moody’s Investors service. (2008). 2007 Review and 2008 Outlook. International Structured Finance (Special report). Nijiskens, R., & Wagner, W. (2011). Credit risk transfer activities and systemic risk: How banks became less risky individually but posed greater risks to the financial system at the same time. Journal of Banking and Finance, 35, 1391–1398. Rajan, N. (2005). Has Financial Devolpment Made the World Riskier?. (National Bureau of Economic Research working paper 11728). Segoviano, M. A., & Goodhart, C. (2009). Banking Stability Measures. (IMF working papers 4). Shin, H. S. (2008). Securitization and Financial Stability. (mimeo). Udhe, A., & Michalak, T. C. (2010). Securitization and systemic risk in European banking: Empirical evidence. Journal of Banking and Finance, 34(12), 3061–3077. Tasche, D. (2000). Risk Contributions and Performance Measurement. (Working paper), TU Munich. Available at http://www.gloriamundl.org Wu, D., Yang, J., & Hong, H. (2011). Securitization and banks’ equity risk. Journal of Financial Services Research, 39(June (3)), 95–117. Yamai, Y., & Yoshiba, T. (2002). Comparative analyses of expected shortfall and value-at-risk (3): Their validity under market stress. Monetary and Economic Studies, Bank of Japan, 20, 181–238.

Please cite this article in press as: Battaglia, F., et al. Securitized banking and the Euro financial crisis: Evidence from the Italian banks risk-taking. Journal of Economics and Business (2014), http://dx.doi.org/10.1016/j.jeconbus.2014.02.003