The relation between securitization and bank risk in
the after-crisis period October 2010 – December
2015
Bachelor’s Thesis
Berry Kramer
10553711
29 June 2015
BSc Economics and Business,
Specialization Economics and Finance
2
Abstract
Securitization is seen as one of the main causes of the financial crisis of 2007-2009. This thesis investigates whether there is a relation between securitization and bank risk. I
investigate if the risk of securitizing US holding companies, which is measured by their stock beta’s, is significantly higher than the risk of non-securitizing US holding companies in the period from October 2010 until December 2015. I use this particular period to investigate a post-crisis period until the most recent quarter that is available in the data. On average, the securitizing holding companies show a significantly higher stock beta than the
non-securitizing holding companies. This indicates that there is a positive relation between securitization and bank risk. The results from the regression analysis both with and without the six largest holding companies show no significant relation between securitization and bank risk after controlling for total assets. This is evidence that the relation between securitization and bank risk is weaker after the crisis.
3
1. Introduction
Securitization is seen as one of the main causes of the financial crisis. In the period before the financial crisis, securitization of loan and bond portfolios was becoming more popular
amongst banks. The amount of securitization outstanding has increased from 5731.625 million dollar to 13997.294 million dollar in the period from 2001 until 2008 (Ben Salah & Fedhila, 2012). Franke and Krahnen (2007) state that “the volume of Collateralized Loan Obligations (CLOs) and Collateralized Bond Obligations (CBOs) strongly increased in the United States and Europe”. The main mechanism of securitization is the transfer of balance sheet assets to a certain special purpose vehicle (SPV). The SPV issues securities to investors, the proceeds of these issues will finance the purchase of the assets. This should transfer balance sheet risk to the capital market, which releases capital for other purposes (Le et al., 2016).
Casu et al. (2011) raise concern about the increasing volume of securitization activity and speculate that this can increase systemic risk. The relation between securitization and
systemic risk is named by Baur and Joossens (2006). They say that banks transfer risk through securitization, for instance through a CDO issuance. This risk transfer poses a risk, because if the risk is incorrectly distributed then the transfer leads to a concentration of risk for some market participants. This can increase the correlation between these market participants, which increases risk for the whole financial system (systemic risk increases). On the opposite, Nijskens and Wagner (2011) say that if a credit risk transfer is correctly distributed, then banks can reduce individual bank risk. However, banks respond by increasing their risk-taking through other methods. For example, “by increasing their lending, by reducing their monitoring and screening efforts or by leveraging up their capital structure” (Nijskens & Wagner, 2011). They indicate that risk transfer activities increased risk in some parts of the financial system, which has probably caused the financial crisis. The influence of bank size on risk-taking is shown by the empirical findings of Uhde and Michalak (2010) who state that “the increase in systemic risk is more relevant for large banks that repeatedly engage in securitization”. This is supported by Jiangli and Pritsker (2008) who observe that asset size is related to banks engaging in securitization activities.
Baur and Joossens (2006), Uhde and Michalak (2010) and Ben Salah and Fedhila (2012) find a positive relation between securitization and bank risk. On the other hand, Jiangli and
Pritsker (2008) and Casu et al. (2011) find a negative relation between securitization and bank risk. Le et al. (2016) investigate if there is a change in the relationship between securitization and bank risk between a period before and a period after the crisis. They find that the relation changes, because in the period before the crisis there is a positive relation and in the period after the crisis there is no relation. A possible reason for this is the Dodd-Frank Act. This regulation, implemented after the crisis in 2010, might have diminished risk-taking because “This requirement calls for retaining at least 5% of the risk in the assets the sponsors
securitize on their balance sheets” (Le et al., 2016). Another reason might be the introduction of accounting standard FAS 166 in 2009. Because of this standard, banks require to recognise and disclose risks which are linked to securitization. These regulations can possibly have a
4
negative effect on the relation between securitization and bank risk after the crisis. I will investigate whether there is a relation between securitization and bank risk after the crisis. I will do this by investigating whether securitizing US holding companies had higher stock beta’s than non-securitizing US holding companies in the post-crisis period from October 2010 until December 2015. I hereby follow Nijskens and Wagner (2011) who use the stock beta’s as a measure of bank risk. An increase in the stock beta can either be because of an increase in the bank’s individual risk or an increase in systemic risk. The higher the stock price is correlated to the market, which is reflected by the stock beta, the higher is the risk of a bank to a collapse of the financial system.
This thesis consists of a literature review, methodology and regression results. In the literature review, I will discuss how securitization can in theory be related to bank risk and present the findings of previous literature. In the methodology, I will describe the data and discuss the methods I use. Finally, I will present and discuss my regression results.
2. Literature Review
Several studies have investigated the effect of securitization on bank risk. Some studies have found results that show a decrease in bank risk, while other studies have found an increase of bank risk. How could securitization increase or decrease bank risk?
Jiangli and Pritsker (2008) state that “securitization can be understood as a credit-derivative that transforms the risk profile of the asset side of the banks’ balance sheet”. Keeping the liability side fixed, securitization can decrease insolvency risk on the asset side because the securitized assets can be used as an insurance against bank insolvency. Uhde and Michalak (2010) also look at the risk of the asset side. They argue that banks finance their new assets with liquid capital from cash and synthetic transactions. Depending on the risk level of these new assets, securitization can increase banks’ overall risk exposure.
Jiangli and Pritsker (2008) argue that securitization allows a bank to choose its exposure to credit risk of the underlying pool of loans. They say that “depending on the rewards for different types of risk exposures, and the exact structure of the transaction, securitization can be used to increase or decrease the bank’s insolvency risk” . On the other hand, Baur and Joossens (2006) argue that a credit risk transfer via securitization affects the bank’s systemic risk. Credit risk remains in the financial system because CDO tranches are sold to other banks. This suggests that banks invest in the same underlying collateral pool as their
investors. This can increase the correlation between the banks, which can increase systemic risk.
Franke and Krahnen (2007) state that securitization and reinvestment have a positive impact on banks’ default risk. Banks mostly securitize their loan portfolios to increase their set of investment opportunities. The income from securitization activities can be used to invest in risk-free assets or to repay outstanding debt. This would decrease banks’ overall risk. It is
5
more likely that banks use these securitization proceeds to expand their loan business. Moreover, the banks will take a first-loss position in these new loans and thereby take the default risk. Uhde and Michalak (2010) also say that banks take a first-loss position in the issuance of senior tranches. While securitization transfers the tail risk of these senior tranches out of the balance sheets, most of the default risks remain. Franke and Krahnen (2007) add that “even though the total loan portfolio of the bank is now better diversified, the overall risk of the bank is likely to be higher than before securitization”.
Based on previous empirical research, Casu et al. (2011) argue that there is a positive relation between banks engaging in securitization and the risk of these issuing banks. Securitization causes banks to insert illiquid assets into liquid funds, which could enlarge the use of credit and banks are more likely to hold riskier assets. This can increase bank risk. Baur and Joossens (2006) mention a similar reason, they suggest that banks will invest in assets with a higher return and higher risk because of a credit risk transfer.
These arguments can explain whether securitization increases or decreases bank risk. Several studies have empirically investigated if there is a relationship between securitization and bank risk.
A few articles show a negative relationship between securitization and bank risk. Jiangli and Pritsker (2008) for example examine the effect of several forms of asset securitization on insolvency risk, profitability and leverage ratios of banks. They use data on US bank holding companies from 2001 to 2007. They use three approaches to analyse the effect of
securitization, particularly mortgage securitization, on US bank holding companies. The first approach assumes that insolvency risk has a relation to the construction of the balance sheet. This relationship is used to calculate the bank’s insolvency risk. The method places the securitized loans on the balance sheet and predicts the bank’s insolvency risk assuming the securitizer did not securitized its loans. The second approach they use is comparing the average performance of large securitizing banks to non-securitizing banks with a similar size. Jiangli and Pritsker (2008) use an instrumental variable regression which includes the variable bank size as their third approach. They use bank size because it significantly affects the likelihood that a bank is securitizing. Using these approaches, they find significant evidence that “mortgage securitization reduces insolvency risk and increases bank leverage and bank’s profitability”.
Casu et al. (2011) study whether securitization had an impact on bank’s credit-risk taking behaviour, using data on US holding companies in the period from 2001 until 2007. Their results show that securitization has a significant negative effect on bank’s credit risk taking. According to Casu et al. (2011) this indicates that “banks with a greater amount of their assets securitized will be more risk-averse in their activities”.
A few articles have found a positive relationship between securitization and bank risk. Baur and Joossens (2006) investigate whether securitization increases the systemic risk of banks by looking at Collateralized Debt Obligations (CDOs). They performed an empirical analysis
6
with monthly data from 2000 to 2005 to find a relation between the issuance of CDOs and systemic risk. They find a positive correlation between extreme joint movements of equity returns in the banking sector and CDO issuance, in Europe. This can be seen as a significant positive relation between systematic risk and CDO issuance. Franke and Krahnen (2007) also examine the issuance of CDOs . They investigate the impact of CDO-transactions around its announcement date on the stock betas of banks. The bank´s stock beta measures the effect on the systemic risk in the stock market. Their empirical results show that default risk tends to increase if the banks use the income from securitization to further engage in securitized loans. On average, their empirical results show an increase of the stock beta at the time of a
securitization announcement. According to them, this indicates that banks use the decrease in risk from securitization to take new risks.
Nijskens and Wagner (2011) also investigate bank’s stock beta’s. They focus on bank risk by looking at the bank´s share prices in a period before the crisis. They are searching for a relationship between credit risk transfer activities and bank risk. Bank risk is captured by the bank’s beta. Nijskens and Wagner (2011) state that “the bank’s beta can be split up into the standard deviation relative to the market’s standard deviation (individual risk) and its
correlation with the market (systemic risk)”. They investigate a sample of 52 banks around the world which are issuing Collateralized Loan Obligations (CLO). They find a significant increase of 0.21 of the stock beta’s of CLO banks. This increase in the beta is due to higher systemic risk only, they find no significant increase in individual bank risk. They conclude that there is a positive relation between credit risk transfer activities and systemic risk. Uhde and Michalak (2010) also investigate the relationship between securitization and systemic risk. They use a dataset of 592 synthetic and cash securitization transactions by 54 European banks from 1997 until 2007. They separate the banks into a group of securitizing banks and a group of non-securitizing banks. A bank is non-securitizing if it does not securitize in the whole period. They compare the differences in their model´s beta coefficients between the two groups. The results show that there is a positive relationship between securitization and systemic risk. This indicates that if a bank reinvests capital into risky assets, then systemic risk will also increase after securitizing these risky assets.
Ben Salah and Fedhila (2012) examine the effect of securitization on risk taking and banking stability. They use a dataset of 174 US commercial banks in the period from 2001 until 2008. They find that securitization has a negative impact on the quality of US banks’ loan portfolios and that securitization increases credit risk in the balance sheets of these banks. These two findings insinuate that credit risk increases when US banks securitize their loans. According to them, this does not confirm that banks become less stable. In contrast, they find a
significant positive effect of securitization on banking stability. This asymmetric effect is explained by Ben Salah and Fedhila (2012) who state that “different classes of securitized assets lead to heterogeneous effects on American banks’ risk”. The findings show that while mortgage securitization has a positive impact on banking stability, non-mortgage
securitization has a negative impact on banking stability. This is because banks have more incentive to monitor mortgage securitization. Therefore, the risk of engaging in mortgage securitization is lower than for non-mortgage securitization. This is in line with the findings of
7
Jiangli and Pritsker (2008), who have found empirical evidence that mortgage securitization lowers insolvency risk.
While the discussed papers provide evidence on the relationship between securitization and risk in the period before the crisis, the evidence on post-crisis period is limited. The only paper to investigate this relation in recent years is Le et al. (2016), who investigate whether the relation between securitization and bank risk changes after the financial crisis. They use a sample of securitization activities of 323 banks in the period from 2001 until 2012. In their variable construction for the regression they use several measures of credit risk. To estimate banks’ credit risk, Le et al. (2016) “focus on three balance sheet ratios: nonperforming loan ratios to total loans, net charge-offs to total assets and loan and lease loss provisions to total assets”. For securitization activities they use a dummy variable which is zero if a bank does not securitize during the period. According to them, securitization is connected to bank size. They control for bank size by adding the natural logarithm of total assets. Their findings show a change in the relationship between securitization and bank risk. Securitization increases bank risk in the period before the crisis, while no significant evidence has been found for this relationship in the period after the crisis.
Le et al. (2016) name a few explanations for this change, one of them is the fact that banks suffered gigantic economic losses during the crisis. Other possible reasons are the new regulations for securitization that were implemented after the crisis. These regulations decrease banks’ incentive to take risk via securitization activities. One of these new
regulations is the Dodd-Frank Act, which was introduced in 2010. The Dodd-Frank Act could diminish risk taking, because parties that sponsored securitization require to retain a minimum of 5% of the risk in securitized assets on their balance sheets. In 2009, the accounting
standard FAS 166 was implemented. According to the Financial Accounting Standards Board (2009) the statement from FAS 166 improved its previous statement FAS 140. Financial reporting improved by removing special-purpose entities from the consolidation guidance and removed the possibility of sale accounting for mortgage securitization. Because of this standard, banks require to recognise and disclose risks which are affiliated with securitization. This can have a negative effect on the relation between securitization and bank risk. This insinuates that bank risk should be less affected by securitization.
The findings of Le et al. (2016) does not show a significant relationship between
securitization and bank risk after the crisis. Based on this result, they suggest that banks have less incentive to engage in risk-taking because of changes in their operating environment. To test if there is a relationship between securitization and bank risk, I will investigate if the stock beta’s of securitizing US holding companies are significantly larger than the stock beta’s of non-securitizing US holding companies in the post-crisis period from October 2010 until December 2015. If there is a significant difference in the average stock beta’s between securitizing and non-securitizing US holding companies, then it can be suggested that there is a positive relationship between bank risk and securitization.
8
3. Methodology
In this section, I will investigate whether the risk of securitizing US holding companies, as measured by their beta’s, is significantly larger than of non-securitizing US holding
companies. First, I will explain how the US holding companies are categorized. Second, I will describe the methods to calculate the beta’s. To test whether there is a significant difference of the stock beta´s, I will perform a T-Test. Lastly, I will perform regression analysis. Data and descriptive statistics
Securitizing and non-securitizing US holding companies
I have looked into the Y-9C reports of the US Holding Companies for the period from the last quarter of 2010 until the last quarter of 2015. If in this whole period a holding company participated at least once in a securitization activity, which is issuing one or more of the securitization activities loans, then I named it a “securitizing” holding company and a “non-securitizing” holding company otherwise.
I have chosen to investigate the time period from October 2010 until December 2015. I use this period because in October the last quarter of 2010 starts and the period ends at the last quarter of 2015, which is the most recent quarter available in the data. This is the period after the crisis, which started in 2007 and ended in 2009 (Bedendo & Bruno, 2012). I will
investigate this post-crisis period and examine if there is a difference with empirical research that investigated the pre-crisis period.
On the FFIEC Website is a list of all US Holding Companies with assets larger than 10 billion dollars on 31 December 2015. In total these are 109 companies. I will use these holding companies for my research. On Wharton Research Data Services I acquired the Y-9C reports of these US holding companies. Since June 2001, US banks are required to quarterly provide detailed information on securitization activities in the Y-9C reports. In this report,
Securitization activities are divided into seven categories: 1–4 family residential loans, home equity lines, credit card receivables, auto loans, other consumer loans, commercial and
industrial loans, and a category with all other loans, all leases, and all other assets (Casu et al., 2011). After studying the securitization activities in the Y-9C reports, complete securitization activities data was found for 78 US holding companies. From these holding companies, there are 31 securitizing and 47 are non-securitizing. After searching for daily stock returns over the whole period, only 27 securitizing and 32 non-securitizing holding companies remain for the sample. For 3 securitizing holding companies, the total asset data was incomplete. This was also the problem for 5 non-securitizing holding companies. I omitted three non-securitizing holding companies with the lowest amount of total assets to equalize the amount of both groups.
After this selection, 24 securitizing and 24 non-securitizing US holding companies were chosen.
9
Table 1: The list of 24 securitizing and 24 non-securitizing US holding companies
Securitizing US Holding Companies Non-Securitizing US Holding Companies
Bank of New York Mellon Corp. Northern Trust Corp.
Goldman Sacks Group Inc. American Express Company
Capital One Financial Corp. Fulton Financial Corp.
PNC Financial Services Group Inc. Comerica Inc.
Regions Financial Corp. Webster Financial Corp.
First BanCorp Bank of Hawaii Corp.
Suntrust Banks Inc. Zions BanCorp.
Keycorp SVB Financial Group
Umpqua Holdings Corp. CIT Group Inc.
M&T Bank Corp. Valley National BanCorp
Huntington BancShares Commerce BancShares
Bank of America UMB Financial Group
BB&T Corp. Prosperity BancShares
WinTrust Financial Corp. International BancShares Inc.
East West BanCorp First Citizens BancShares
First Horizon National Corp. Synovus Financial Corp.
Fifth Third BanCorp TrustMark Corp.
BancorpSouth Hancock Holding Company
Wells & Fargo Company TCF Financial Corp.
J.P. Morgan Chase & Co. State Street Corp.
Popular Inc. Old National BanCorp
BOK Financial Corp. Cullen/Frost Bankers
CitiGroup Inc. IberiaBank Corp.
10
Descriptive Statistics
I use US bank holding company data retrieved from the Compustat database at Wharton Research Data Services. I retrieved quarterly data for total assets, turnover and total liabilities from the last quarter of 2010 until the last quarter of 2015. I divide the total liabilities by the total assets to compute the leverage, which is the debt/asset ratio.
Table 2: Descriptive statistics of all US holding companies
Variable Group Obs Mean Median St. Dev. Min Max
Total assets ($100,000) Securitizing 24 483189.365 124474.143 734921.869 13178.129 2381388.429 Non-Sec 24 41832.085 21305.829 51869.582 9826.197 229742.810 All 48 262510.725 32126.082 561570.157 9826.197 2381388.429 Turnover ($100,000) Securitizing 24 6027.920 1595.214 8924.534 177.658 26504.286 Non-Sec 24 780.410 258.317 1725.691 135.709 8498.143 All 48 3404.165 424.171 6889.428 135.709 26504.286 Leverage (Debt/Asset Ratio) Securitizing 24 0.887 0.890 0.016 0.847 0.913 Non-Sec 24 0.887 0.890 0.024 0.813 0.927 All 48 0.887 0.890 0.020 0.813 0.927 Figure 1: Average Total Assets
$0 $100.000 $200.000 $300.000 $400.000 $500.000 $600.000 Securitizing Non-Securitizing Avg. Total Assets ($100,000)
11
Figure 2: Average Turnover
Figure 3: Leverage (Debt/Asset Ratio)
Figure 1 shows that the average total amount of assets is around $450 billion larger for the securitizing US holding companies compared to the non-securitizing US holding companies. This is most probably caused by the huge amount of assets of the six largest securitizing US holding companies. For example J.P. Morgan Chase & Co., Bank of America and Citigroup Inc., who possess respectively $2381 billion, $2168 billion and $1884 billion. Figure 2 shows an average turnover difference of around $500 million, which can also be directed to the large turnovers of the largest holding companies. Figure 3 shows a really small difference in
leverage of 0.0006, which is almost negligible. $0 $1.000 $2.000 $3.000 $4.000 $5.000 $6.000 $7.000 Securitizing Non-Securitizing Avg. Turnover ($100,000) 0,8864 0,8866 0,8868 0,8870 0,8872 0,8874 0,8876 Securitizing Non-Securitizing
12
Bank market Beta’s
I will follow Nijskens and Wagner (2011) to use the stock beta’s of the US holding companies as a measure of bank risk. Bank risk can be idiosyncratic or systemic risk, and beta measures its systemic component. A higher beta due to systemic risk indicates an increase in the risk that the financial system collapses. By calculating the beta’s of the US holding companies, I can investigate whether systemic risk is higher for securitizing or non-securitizing holding companies. Formally, I will test the following hypothesis:
𝐻𝐻0: 𝛽𝛽������ = 𝛽𝛽1𝑠𝑠𝑠𝑠𝑠𝑠 ����������� 1𝑛𝑛𝑛𝑛𝑛𝑛−𝑠𝑠𝑠𝑠𝑠𝑠
𝐻𝐻1: 𝛽𝛽������ > 𝛽𝛽1𝑠𝑠𝑠𝑠𝑠𝑠 ����������� 1𝑛𝑛𝑛𝑛𝑛𝑛−𝑠𝑠𝑠𝑠𝑠𝑠
Following Franke and Krahnen (2007), a higher beta is expected for the securitizing US holding companies compared to the non-securitizing US holding companies.
To calculate the beta of each bank, the CAPM model is used: 𝑅𝑅𝑖𝑖 = 𝛽𝛽0+ 𝛽𝛽1𝑅𝑅𝑚𝑚+ 𝜀𝜀𝑡𝑡
Where 𝑅𝑅𝑖𝑖 are the daily stock returns of holding company i , 𝑅𝑅𝑚𝑚 are the daily stock returns of the S&P500 (market portfolio), 𝜀𝜀𝑡𝑡 is the error term, and 𝛽𝛽1 = 𝐶𝐶𝑛𝑛𝐶𝐶(𝑅𝑅𝑖𝑖,𝑅𝑅𝑚𝑚)
𝑉𝑉𝑉𝑉𝑉𝑉(𝑅𝑅𝑚𝑚)
The daily stock returns of the US Holding Companies for the period from 1 October 2010 until 31 December 2015 are acquired from Wharton Research Data Services. They retrieve their data from the Centre for Research in Security Prices (CRSP). I also retrieved the daily returns on the S&P500 index from the CRSP database.
The resulting market beta’s of the sample of banks are presented in Table 3. It can be seen that almost all of them exceed 1, which suggests that the stock returns of almost all the US holding companies are more volatile than the returns of the S&P500. This indicates that the non-diversifiable risk for the holding companies is larger than for the benchmark.
13
Table 3: Beta estimation results
Securitizing Non-Securitizing
Company Names 𝜷𝜷𝑺𝑺𝑺𝑺𝑺𝑺 Company Names 𝜷𝜷𝑵𝑵𝑵𝑵𝑵𝑵−𝑺𝑺𝑺𝑺𝑺𝑺
Bank of New York Mellon Corp.
1.3259 Northern Trust Corp. 1.1614
Goldman Sacks Group Inc.
1.3098 American Express
Company
1.0746 Capital One Financial
Corp.
1.2564 Fulton Financial Corp. 1.2078
PNC Financial Services Group Inc.
1.1951 Comerica Inc. 1.3160
Regions Financial Corp. 1.6756 Webster Financial Corp. 1.4872
First BanCorp 1.4339 Bank of Hawaii Corp. 0.9825
Suntrust Banks Inc. 1.5138 Zions BanCorp. 1.4352
Keycorp 1.4034 SVB Financial Group 1.4299
Umpqua Holdings Corp. 1.3149 CIT Group Inc. 1.1031
M&T Bank Corp. 1.0453 Valley National BanCorp 1.0526
Huntington BancShares 1.3927 Commerce BancShares 0.9971
Bank of America 1.6775 UMB Financial Group 1.2250
BB&T Corp. 1.1955 Prosperity BancShares 1.2347
WinTrust Financial Corp. 1.1042 International BancShares
Inc.
1.6002
East West BanCorp 1.3108 First Citizens BancShares 0.9893
First Horizon National Corp.
1.3518 Synovus Financial Corp. 1.5073
Fifth Third BanCorp 1.3697 TrustMark Corp. 1.2331
BancorpSouth 1.3323 Hancock Holding
Company
1.2412
Wells & Fargo Company 1.2692 TCF Financial Corp. 1.2420
J.P. Morgan Chase & Co. 1.3663 State Street Corp. 1.3166
Popular Inc. 1.4180 Old National BanCorp 1.4299
BOK Financial Corp. 1.0235 Cullen/Frost Bankers 1.0437
CitiGroup Inc. 1.7063 IberiaBank Corp. 1.0841
Morgan Stanley 1.7926 New York Community
BanCorp
0.9372
Avg. Beta 1.3660 Avg. Beta 1.2138
St. Dev 0.1981 St. Dev 0.1822
14
Figure 4: Average Beta
Figure 4 shows that the average beta of the securitizing US holding companies is around 0.15 larger than the average beta of the non-securitizing US holding companies.
T-Test
To confirm the hypothesis, a T-Test is used to estimate whether the stock beta of the securitizing holding companies is significantly larger than the stock beta of the non-securitizing holding companies.
The T-Test formula:
𝑇𝑇 = 𝛽𝛽������ − 𝛽𝛽𝑆𝑆𝑠𝑠𝑠𝑠 ������������𝑁𝑁𝑛𝑛𝑛𝑛−𝑆𝑆𝑠𝑠𝑠𝑠 𝑆𝑆�1𝑛𝑛 +𝑚𝑚1
𝛽𝛽𝑆𝑆𝑠𝑠𝑠𝑠
������ = Average Beta of the Securitizing Holding Companies 𝛽𝛽𝑁𝑁𝑛𝑛𝑛𝑛−𝑆𝑆𝑠𝑠𝑠𝑠
������������ = Average Beta of the Non-Securitizing Holding Companies n = Number of Securitizing Holding Companies
m = Number of Non-Securitizing Holding Companies The formula to calculate the standard deviation S:
𝑆𝑆2 = (𝑛𝑛 − 1)𝑆𝑆𝑋𝑋2+ (𝑚𝑚 − 1)𝑆𝑆𝑌𝑌2
𝑛𝑛 + 𝑚𝑚 − 2
Degrees of freedom = n + m – 2 = 24 + 24 – 2 = 46
I will assume that the stock beta’s are normally distributed. 1,10 1,15 1,20 1,25 1,30 1,35 1,40 Securitizing Non-Securitizing Average Beta
15
Table 4: Known Variables for the T-Test
Known Variables Value
𝛽𝛽1𝑠𝑠𝑠𝑠𝑠𝑠 ������ 1.3660 𝛽𝛽1𝑛𝑛𝑛𝑛𝑛𝑛−𝑠𝑠𝑠𝑠𝑠𝑠 ����������� 1.2138 𝜎𝜎𝑠𝑠𝑠𝑠𝑠𝑠 0.1981 𝜎𝜎𝑛𝑛𝑛𝑛𝑛𝑛−𝑠𝑠𝑠𝑠𝑠𝑠 0.1822 𝜎𝜎𝑠𝑠𝑠𝑠𝑠𝑠2 0.0392 𝜎𝜎𝑛𝑛𝑛𝑛𝑛𝑛−𝑠𝑠𝑠𝑠𝑠𝑠2 0.0332 n 24 m 24
Table 5: T-Test Results
Calculated Variables Value
𝛽𝛽1𝑠𝑠𝑠𝑠𝑠𝑠 ������ − 𝛽𝛽����������� 1𝑛𝑛𝑛𝑛𝑛𝑛−𝑠𝑠𝑠𝑠𝑠𝑠 0.1522 S 0.1903 t-value 2.7718*** p-value 0.0040 *** if P < 0.01
According to the T-Test, the average stock beta of the securitizing US holding companies is significantly higher than the average stock beta of the non-securitizing US holding companies. This indicates that there is a positive relation between securitization and the stock beta after the crisis.
The T-Test does not control for bank size, while beta and bank’s securitization decision might be related to bank size.
16
The following figures show the linear relation between the beta and total assets for both groups individually and for all holding companies as well.
Figure 5-10: Linear relation between beta and total assets
Fig 5: all Fig 6: securitizing
Fig 7: non-securitizing Fig 8: securitizing, without 6 largest companies
Fig 9: all, without 6 largest companies Fig 10: 6 largest companies
0 0,5 1 1,5 2 $0 $1.000.000 $2.000.000 $3.000.000 Beta Total Assets ($100,000) 0 0,5 1 1,5 2 $0 $1.000.000 $2.000.000 $3.000.000 Beta Total Assets ($100,000) 0 0,5 1 1,5 2 $0 $100.000 $200.000 $300.000 Beta Total Assets ($100,000) 0 0,5 1 1,5 2 $0 $200.000 $400.000 Beta Total Assets ($100,000) 0 0,5 1 1,5 2 $0 $100.000$200.000$300.000$400.000 Beta Total Assets ($100,000) 0 0,5 1 1,5 2 $0 $1.000.000 $2.000.000 $3.000.000 Beta Total Assets ($100,000)
17
Figure 11: Linear relation between beta and total assets, comparing securitizing to non-securitizing without the 6 largest non-securitizing holding companies
Figure 5 shows a positive relation for both the securitizing and non-securitizing US holding companies. This insinuates that if total assets increase, the stock beta will also increase. The trend line of Figure 6 shows a similar positive linear relation between the beta and total assets. This can insinuate that the stock beta increases if total assets increase, which suggests that bank size is positively related to the stock beta of a securitizing holding company. However, figure 7 shows a slightly decreasing, almost horizontal trend line. This shows that total assets does not relate to the stock beta of a non-securitizing holding company.
Therefore, I conclude that the result obtained in figure 5 is driven by the securitizing holding companies. This is probably because of the large differences in asset size between the two groups. The six largest securitizing holding companies have assets differing from $800 billion to $2400 billion, while the largest holding company of the non-securitizers is $230 billion. This suggests that the trend lines of figure 5 and 6 are driven by these large holding
companies. Therefore, I will compare holding companies of comparable size to obtain a more accurate linear relation between beta and total assets. This means that I will omit the six largest securitizing holding companies from the figures 5 and 6.
The figures 8,9 and 11 do not include the six largest securitizing holding companies. Figure 8 shows no relation between bank size and risk and figure 9 shows a slightly positive relation. This can support the conclusion that the trend lines in figures 5 and 6 are driven by the six largest securitizing holding companies. Figure 10 shows no relation for the six largest
securitizing holding companies. The figures 8 and 10 both show no relationship between bank size and risk, while there is a positive relation in figure 6 which uses both large and small securitizing holding companies. This could indicate that holding companies with a large amount of total assets are riskier than small holding companies. Figure 11 shows that the
0 0,2 0,4 0,6 0,8 1 1,2 1,4 1,6 1,8 $0 $100.000 $200.000 $300.000 $400.000 Beta Total Assets ($100,000) Securitizing Non-Securitizing Linear (Securitizing) Linear (Non-Securitizing)
18
trend line of the securitizing holding companies is higher than the trend line of the non-securitizing holding companies. This shows that the beta for non-securitizing holding companies lays on a higher level than the beta of non-securitizing holding companies. This supports the result from the T-Test.
The figures 8 and 10 show no relationship between bank size and risk, while figure 6 shows a positive relation for both large and small securitizing holding companies. This indicates that large banks are riskier than small banks, which means that bank size is positively related to risk. Figures 5 and 6 indicate that bank size is positively related to securitization. These two conclusions combined insinuate that there is a positive relation between securitization and risk. But it is possible that large holding companies both have high beta’s and are active in securitization. This is a reason why the regression analysis should control for total assets.
19
4. Regression Results
The previous section has shown that a T-Test not controlling for bank size might lead to misleading conclusions on the relationship between securitization and bank risk-taking. In this section, I use regression analysis to examine this relationship.
I follow Jiangli and Pritsker (2008) and Le et al. (2016) and control for bank size by adding the natural logarithm of total assets to the regression equation. The total assets of the US holding companies are the average of the total assets over the period from the last quarter of 2010 until the last quarter of 2015. I will also add two more control variables. First, the natural logarithm of the average turnover. Second, the leverage (debt/asset ratio). According to Jiangli and Pritsker (2008), there is a relation between leverage and securitization
decisions. I will follow Le et al. (2016) and capture banks’ securitization activities by adding a dummy variable which equals one if a holding company is securitizing and zero if a holding company is non-securitizing. I will perform eight regressions with different specifications to see which factors are significant in explaining systemic risk.
By adding the control variables, the regression equation (model) is the following: 𝛽𝛽𝑖𝑖 = 𝛼𝛼 + 𝛿𝛿1𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖 + 𝛿𝛿2 𝐴𝐴𝑆𝑆𝑆𝑆𝑖𝑖 + 𝛿𝛿3𝑇𝑇𝑇𝑇𝑅𝑅𝑖𝑖+ 𝛿𝛿4 𝐿𝐿𝑆𝑆𝐿𝐿𝑖𝑖
Dependent variable
𝛽𝛽𝑖𝑖 = The stock beta of US holding company i
Independent variables
𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖 = Dummy Variable which equals 1 if the holding company i is securitizing and equals 0
if the holding company i is non-securitizing
𝐴𝐴𝑆𝑆𝑆𝑆𝑖𝑖 = The natural logarithm of the average total assets of holding company i over the period
October 2010 until December 2015
𝑇𝑇𝑇𝑇𝑅𝑅𝑖𝑖 = The natural logarithm of the average turnover of holding company i over the period
October 2010 until December 2015
𝐿𝐿𝑆𝑆𝐿𝐿𝑖𝑖 = The leverage, which is the average debt / asset ratio of holding company i over the
period October 2010 until December 2015
Hypothesis
The regression analysis investigates if there is a significant difference between the stock beta’s of the US holding companies and securitization activities. There is a relation between securitization and the stock beta when the coefficient of the dummy variable of securitization activities is significantly different from zero. This results in the following hypothesis:
𝐻𝐻0: 𝛿𝛿1 = 0
20
Table 6: Regression results
β β β β β β β β Securitization 0.152*** 0.0935 0.0896 0.0863 0.105 0.152*** 0.0934 0.105 (0.0549) (0.0628) (0.0622) (0.0629) (0.0630) (0.0555) (0.0637) (0.0637) Ln(Assets) 0.0363* 0.185* 0.211* 0.0364* (0.0202) (0.105) (0.115) (0.0207) Ln(Turnover) -0.146 -0.170 0.0291 0.0289 (0.101) (0.110) (0.0197) (0.0200) Leverage -0.873 0.326 -0.0331 0.175 (1.453) (1.385) (1.369) (1.373) Constant 1.214*** 0.842*** 0.184 0.827 1.0428*** 0.925 0.871 0.888 (0.0388) (0.210) (0.502) (1.183) (0.122) (1.229) (1.202) (1.215) Obs 48 48 48 48 48 48 48 48 R2 0.1431 0.2004 0.2365 0.2428 0.1826 0.1442 0.2004 0.1829 Adjusted R2 0.1245 0.1649 0.1844 0.1724 0.1463 0.1061 0.1459 0.1272 * if P < 0.10 ** if P < 0.05 *** if P < 0.01
The numbers in this table are the coefficients. The numbers between the brackets are the standard errors of coefficient estimates.
The first regression shows that the stock beta of a securitizing holding company is significantly higher by 0.152 (at a 1% significance level) than the stock beta of a non-securitizing holding company. This is the same result as the T-Test. The constant represents the average beta of the non-securitizing firms. This constant is 1.214 and is significant at a 1% significance level. This regression is relevant, because it shows that without any control variables, the beta of a securitizing holding company is significantly higher than the beta of a non-securitizing holding company. This is in line with the hypothesis for the T-Test and the hypothesis for the outcome of the regression.
This result is comparable to the result of Nijskens and Wagner (2011), who have found a significant increase of 0.21 of the beta of banks participating in Collateralized Loan Obligations. This result is also in line with Franke and Krahnen (2007), whose empirical evidence shows an increase of the stock beta at the moment of a securitization announcement. Based on the results, I can conclude that there is positive relation between securitization and
21
bank risk. This is in line with pre-crisis research like Baur and Joossens (2006) and Uhde and Michalak (2010) who also find a positive relationship between securitization and systemic risk. Ben Salah and Fedhila (2012) show a similar result, they find a positive relation between securitization and credit risk. This result is contrary to Le et al. (2016), who find no relation between securitization and bank risk after the crisis.
In the second regression, the variable Ln(Assets) is added. An increase of one unit of
Ln(assets) results in an increase of 0.0363 of the stock beta at a 10% significance level. This indicates that total assets is positively related to the stock beta, which is also confirmed by Le et al. (2016). It is also in line with the conclusions from the figures 5 until 11. This regression shows no significant evidence that securitization is related to the beta. The stock beta of a securitizing holding company is 0.0935 higher than the stock beta of non-securitizing holding company, without significance. This result is in line with Le et al. (2016), who shows that there is no significant relation between securitization and bank risk in the period after the crisis. This outcome is different from pre-crisis results. Nijskens and Wagner (2011) found a significant increase of the stock beta of CLO banks, after controlling for total assets.
In the third regression, the variable Ln(Turnover) is added. When controlled for total assets, an increase of one unit of Ln(Turnover) results in a decrease of -0.146 of the stock beta. This coefficient is not significant. The coefficient of Ln(Assets) is similar to the result obtained in the second regression. The stock beta of a securitizing holding company is higher by 0.0896 than the stock beta a non-securitizing holding company, without significance. This regression does not have a large impact on the results, but it does have an added value. The adjusted R-squared increases from 0.1649 in the second regression to 0.1844 in the third regression. This shows that Ln(Turnover) plays a role in estimating the beta, when the model also controls for total assets.
In the fourth regression, the model controls for Ln(Assets) and Ln(Turnover), the variable leverage is added. The coefficients of Ln(Assets) and Ln(Turnover) are similar to the second and third regression. If leverage increases with one unit, then the stock beta decrease with -0.873. This coefficient is not significant. The stock beta of a securitizing holding company is higher by 0.0896 than the stock beta a non-securitizing holding company, without
significance. The adjusted R-squared decreases from 0.1844 in the third regression to 0.1724 in the fourth regression. This shows that leverage does not play a role in estimating the beta. In the fifth regression, the only control variable is Ln(Turnover). An increase of one unit of Ln(Turnover) results in an increase of 0.0291 of the stock beta, without significance. The sign of Ln(Turnover) is different from the third and fourth regression. This can be caused by omitted variable bias. This occurs when the model leaves out one or more important variables. The coefficient for securitization activities is similar to the second, third and fourth
regression. This regression does not have a large impact on the results, but it does add value. The adjusted R-squared of the fifth regression compared to the first regression increases from 0.1245 to 0.1463. This shows that Ln(Turnover) plays a role in estimating the beta.
22
In the sixth regression, the only control variable is leverage. If leverage increases with one unit, then the stock beta increases with 0.326. This coefficient is not significant. The sign of the coefficient is now positive, which is different from the negative sign in the fourth regression. This can be caused by omitted variable bias. The coefficient of securitization is 0.152 at a 1% significance level. This indicates a significant higher stock beta for securitizing holding companies compared to non-securitizing holding companies, as expected by the hypothesis. However, the adjusted R-squared shows that this coefficient is not relevant. The adjusted R-squared decreases from 0.1245 to 0.1061 compared to the first regression. This shows that leverage does not play a role in estimating the beta.
The seventh regression controls for total assets and leverage. The coefficients of securitization activities and Ln(Assets) are similar to previous regressions. The coefficient of leverage is similar to the fourth regression. This regression does not have an impact on the results and does not add value. This is supported by the adjusted R-squared, which decreases from 0.1649 to 0.1459 compared to the second regression. The eight regression does not have any impact on the results and shows similar coefficients as in the previous regressions.
The regression analysis shows that Ln(Assets) is the most important control variable, which is consistent with the figures 5 until 11. The coefficient of Ln(assets) is significant in every regression that controls for bank size. While the regressions including Ln(Assets) show that there is no relationship between securitization and beta, most regressions that excluded Ln(Assets) suggested a positive relationship. Therefore, the regressions that exclude Ln(Assets) lead to misleading results, which is caused by omitted variable bias.
23
Regression analysis without the six largest securitizing holding companies
The following regression analysis consists of four regressions in which the six largest securitizing holding companies are omitted, which is the case in the figures 8 until 11.
Table 7: Regression Results
β β β β Securitization 0.101* 0.0972 0.0935 0.0895 (0.0542) (0.0597) (0.0597) (0.0602) Ln(Assets) 0.00441 0.120 0.147 (0.0285) (0.112) (0.117) Ln(Turnover) -0.108 -0.136 (0.101) (0.107) Leverage -1.170 (1.436) Constant 1.214*** 1.169*** 0.621 1.541 (0.0355) (0.293) (0.590) (1.276) Obs 42 42 42 42 R2 0.0795 0.0801 0.1070 0.1227 Adjusted R2 0.0565 0.0329 0.0365 0.0279 * if P < 0.10 ** if P < 0.05 *** if P < 0.01
The numbers in this table are the coefficients. The numbers between the brackets are the standard errors of coefficient estimates.
The first regression shows that the stock beta of a securitizing holding company is significantly higher by 0.101 (at a 10% significance level) than the stock beta of a non-securitizing holding company. This result fits together with figure 11 and is in line with Nijskens and Wagner (2011) and Franke and Krahnen (2007), who both show an increase in the stock beta which is related to securitization. This indicates that after omitting the six largest securitizing holding companies, the regression still shows the same relation between securitization and bank risk. This result is in line with pre-crisis research.
24
The second regression controls for the variable Ln(Assets). The coefficient of this variable is 0.00441 and is not significant. This result shows that there is a slightly positive (not
significant) relation between the beta and bank size, which is in line with figure 9. This could indicate that the result obtained in the second regression in table 6 is driven by the six largest securitizing holding companies. If this is the case, then this would be comparable to the conclusion of Uhde and Michalak (2010) who find that the increase in systemic risk is more relevant for banks with a larger bank size who continuously engage in securitization. The stock beta of a securitizing holding company is 0.0972 higher than the stock beta of a non-securitizing holding company. The p-value of this coefficient is 0.112 and is therefore not significant on a small margin above 10%. However, due to the small sample size, the result might be stronger. A larger sample size could increase accuracy and therefore the significance of the coefficient, which would increase the relevance of this result. This would indicate that after controlling for bank size there might be a positive relationship between securitization and bank risk. Besides that, the adjusted R-squared decreases from 0.0565 in the first
regression to 0.0369 in the second regression. This shows that Ln(Assets) does not play a role in estimating the beta.
The third and fourth regression add the variables Ln(Turnover) and leverage respectively. Both coefficients are not significant and do not have impact on the results. The adjusted R-squared increases from 0.0329 in the second regression to 0.0365 in the third regression. This suggests that Ln(Turnover) plays a minor role in estimating the beta. Leverage does not play a role in estimating the beta.
25
5. Conclusion
This paper investigated if the stock beta’s of securitizing US holding companies are higher than the stock beta’s of non-securitizing US holding companies in the period from October 2010 until December 2015. The aim for this is to investigate whether there is a relation between securitization and bank risk during this post-crisis period.
The results from the T-Test show that the average stock beta of the securitizing holding companies is significantly higher than the average stock beta of the non-securitizing holding companies. However, the regression results show no evidence that the stock beta’s are higher for the securitizing holding companies than the non-securitizing holding companies, when the regression controls for total assets. This indicates that there is no relation between
securitization and bank risk. This is in line with the research from Le et al. (2016), who find no significant relation between securitization and bank risk in the period after the crisis. The regression with all holding companies that only controls for total assets does show a
significant positive effect of total assets on the beta by 0.0363 (at a 10% significance level). This is in line with Jiangli and Pritsker (2008) and Le et al. (2016), who show that
securitization is related to bank size. This result also indicates that larger holding companies are riskier and more likely to securitize. On the contrary, table 7 does not show a significant relation between total assets and the stock beta. The regression results show that there is no significant relation between the turnover and leverage of the holding companies and the stock beta. The results after controlling for these variables remain the same.
The enormous economic losses that the banks suffered is named as one of the reasons why there is no relation between securitization and bank risk in the period after the crisis. Besides that, new regulations were implemented after the crisis. Regulations such as the Dodd-Frank Act and accounting standard FAS 166 could have diminished bank’s incentives to participate in securitization activities.
6. Implications
In this thesis, I investigated whether there is a relation between securitization and bank risk after the financial crisis. While the T-Test indicated a possible relationship, a regression analysis was further implemented to control for the possibility that this relationship is driven by bank size, which makes banks both riskier (Uhde and Michalak, 2010) and more likely to securitize (Jiangli and Pritsker, 2008). The results from the regression analysis show that there is no significant difference between the stock beta’s of securitizing and non-securitizing holding companies once bank size is controlled for. This finding is similar to the result obtained by Le et al. (2016). After controlling for bank size, they find a positive relation between securitization and bank risk in the period preceding the crisis while they find no relation between securitization and bank risk in the post-crisis period.
26
In light of the financial crisis, this finding can be seen as a positive result. Banks that securitize might not have a higher risk exposure to the financial system. This could indicate that the regulations which were implemented after the crisis could have reduced bank risk-taking. For instance the Dodd-Frank Act, which requires banks to keep a capital requirement of a minimum of 5% of the risk in securitized assets on the balance sheets. Another reason might be the accounting standard FAS 166. This accounting standard was implemented to reduce securitization by omitting special-purpose entities, like the SPV, from the
consolidation guidance and it removed the possibility of sale accounting for mortgage securitization. Other reasons why banks could have reduced their risk-taking could be the alertness of investors because of the financial crisis. Investors will focus more on
securitization activities in their due diligence. Also financial authorities will pay more attention to securitization activities when they monitor banks.
The regression analysis without the six largest securitizing holding companies shows no relation between bank size and risk. However, the regression analysis with all holding
companies does show a significant positive relation between total assets and the stock beta’s. This suggests that large banks are riskier. This is consistent with the conclusion of Uhde and Michalak (2010), who show that an increase in systemic risk is more relevant for large banks who continuously participate in securitization activities. Besides that, large banks are more likely to securitize. This is in line with the conclusion of Jiangli and Pritsker (2008) who show that as bank size increases, the likelihood that a bank is securitizing increases as well. The large bank size of securitizing holding companies reflects economies of scale in securitization. This implicates that large banks use securitization as an alternative funding source instead of deposits. They also conclude that large banks are more likely to participate in investments with higher expected losses than small banks.
Limitations
I started with US holding company data for the 109 companies with total assets larger than $10 billion in December 2015. Because of missing data on securitization activities, stock returns or total assets, a selection of only 48 companies is investigated. This sample size could be larger, which would create more accurate results. The sample size can be enlarged by looking at US holding companies with total assets less than $10 billion.
The regression model only controlled for total assets, turnover and leverage. Adding more control variables could have increased accuracy of the results. Besides that, it could have turned out that more variables are related to the stock beta.
Further Research
The stock beta is used as a measure of bank risk in this research. This measure is also used by Franke and Krahnen (2007) and Nijskens and Wagner (2011). In further research, other measures of risk can be used. Baur and Joossens (2006) for instance use CDO issuance as measure for credit risk transfer. A CDO issuance could provide more information on bank risk, because a CDO issuance is a direct transfer of risk. CDOs can increase the correlation between market participants when CDOs are bought by financial intermediaries. This can
27
increase systemic risk. Ben Salah and Fedhila (2012) use the ratio of risk weighted assets to total assets as a measure of bank credit risk. In further research, also another relationship can be investigated. Investigating whether there is a change in the stock beta’s of multiple securitizing US holding companies between a pre-crisis period and a post-crisis period could show if bank risk increased or decreased for securitizing US holding companies after the crisis.
28
7. References
Baur, D., & Joossens, E. (2006). The effect of credit risk transfer on financial stability (EUR Working Paper No 21521 EN). Retrieved from the Social Science Research Network website: http://dx.doi.org/10.2139/ssrn.881774
Bedendo, M., & Bruno, B. (2012). Credit risk transfer in U.S. commercial banks: What changed during the 2007–2009 crisis? Journal of Banking and Finance, 36, 3260-3273. doi: 10.1016/j.jbankfin.2012.07.011
Ben Salah, N. & Fedhila, H. (2012). Effects of Securitization on Credit Risk and Banking Stability: Empirical Evidence from American Commercial Banks. International Journal of
Economics and Finance, 4(5), 194-207. doi: 10.5539/ijef.v4n5p194
Casu, B., Clare, A., Sarkisyan, A., & Thomas, S. (2011). Does securitization reduce credit risk taking? Empirical evidence from US bank holding companies. The European Journal of
Finance, 17(9-10), 769-788. doi: 10.1080/1351847X.2010.538526
FASB (2009). Statement of Financial Accounting Standards No. 166. Retrieved on 1 June 2016 from the Financial Accounting Standards Board website:
http://www.gasb.org/jsp/FASB/Document_C/DocumentPage?cid=1176156241521&accepted Disclaimer=true
Franke, G., & Krahnen, J. P. (2007). Default risk sharing between banks and markets: the contribution of collateralized debt obligations. The Risks of Financial Institutions. 603-634. Jiangli, W., & Pritsker, M. (2008). The impacts of securitization on US bank holding
companies. Retrieved from the Social Science Research Network website: http://dx.doi.org/10.2139/ssrn.1102284
Le, H.T.T. , Narayanan, R.P., & Van Vo, T. (2016). Has the Effect of Asset Securitization on Bank Risk Taking Behavior Changed?. Journal of Financial Services Research, 49(1), 39-64 Nijskens, 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(6), 1391-1398.
Uhde, A., & Michalak, T.C. (2010). Securitization and systematic risk in European banking: Empirical evidence. Journal of Banking and Finance, 34(12), 3061-3077.
