The effect of Regulatory Capital on Credit Default Swap
Usage and Bank Lending
Nicholas John Wester
10254285
29
thof June 2015
BSc Economics & Finance
Bachelor thesis
Supervisor: Prof. Tomislav
Ladika
Abstract
This paper studies the impact of regulatory capital, using the Basel frameworks, on bank lending and risk taking behavior in the United States. A difference in differences method is used to estimate the difference in the riskiness of loans before and after Basel II implementation between banks that are active in the credit default swaps market and those that are not. Furthermore, a panel data regression is used to estimate the change in loan riskiness following from a change in credit default swaps usage. This paper finds no significant increase in the relative riskiness of bank lending for banks that trade in credit default swaps, moreover no statistical evidence is found that a change in credit default swaps usage causes a change in loan riskiness.
Contents 1. Introduction ……….4-5 2. Theoretical framework………...5-7 2.1 Basel I………...5-6 2.2 Basel II………..6-7 2.3 Basel III……….7 3. Literature review……….7-9
4. Data and sample description………...9-11 4.1 databases………9-10 4.1 manipulations………10-11
5. Methodology and results………12-17 5.1 summary statistic……….12-13 5.2 Method………...13-14 5.3 Results………...15-17 5.3.1 Difference in differences……….15-16 5.3.2 Panel data………..17
6. Conclusion and limitations……….18-19 6.1 conclusion………..18 6.2 Limitations and suggestions for further research..……….18-19
References………20-21
1. Introduction
Capital regulation has been a topic of interest for many years, first negotiations towards standardized regulation dates back to the early 1980s following from the Latin American debt crisis (Santos, 2001). Regulatory capital has reformed over the years so as to counter inefficiencies and evolvement of the financial system over time. A recent development is credit default swaps, an insurance against default, and its characteristic impact on a bank its capital requirements.
In 1988, the first Basel agreements were set in an attempt to regulate capital reserves of banks so as to increase stability, and protect them against shocks in the economy. In these agreements, capital requirements were set, accounting for certain levels of risk based on the Risk Weighted Assets. Banks, however, are able to decrease their capital requirements by buying a certain type of credit derivatives called credit default swaps, CDS. CDS allows banks to hedge against default, it acts as an insurance where the seller of the CDS takes over risk and protects against default. A consequence is that the credit rating of the loan is represented by the credit rating of the insurer as opposed to the borrower. This paper examines the relationship between risky loans and CDS usage following from changes in regulatory capital.
The market for CDS has grown substantially over time and grown into a multi- trillion dollar industry. In this sample CDS net notional amount, guarantor position minus beneficiary position, to up to 2,48% of total assets, moreover notional beneficiary amounts to up to 41,31% of total assets as seen in the following table and thus is an interesting topic of research.
Previous research on CDS seems to be divided into different camps, on the one hand it is argued that these credit derivatives are a useful tool for banks to transfer credit risk (Acharya and Johnson, 2007). On the other hand, some papers state that the use of CDS, have contributed to the recent financial crisis, as it might affect risk-taking behavior (Stulz, 2009). Previous research, however, seem to focus more on the effects on the firm level than the bank level and focuses mainly on the
Min
Max
Mean
std deviation
Net notional
-‐97400000
211000000
8987813
32800000
Guarantor
0
5000000000
314000000
805000000
Beneficiary
0
5190000000
323000000
828000000
Net notional/assets
-‐0,41%
2,48%
0,09%
0,30%
Guarantor/assets
0
39,84%
2,66%
6,67%
Beneficiary/assets
0
41,31%
2,57%
6,90%
bank lending activity. This paper instead focuses on bank level, the main question that will be addressed is how regulatory capital affects credit default swap usage on loan riskiness. This paper will focus on the difference in a bank its risk level between firms that are active in the credit default swap market and banks that are not active in this market.
Using a difference in differences approach this paper will analyze the
difference in coefficient before and after Basel II implementation for relative position of credit default swaps on high-risk loans. We further test for the difference in the fraction of high-risk loans to total loans between banks that are active in the credit default swap market and banks that are not. This paper finds a positive relationship in loan riskiness and being active in the credit default swap market. We further find weak statistical evidence on the increase in credit default swap usage and high-risk loans. Moreover a panel data regression is performed to estimate whether a change in the usage of credit default swaps is followed by a change in the fraction of high-risk loans. This paper found no statistical evidence
The structure of this paper will be as follows; in the following section (2) important theory will be presented, followed by a review of outcomes and theories of previous research. In section 3 a description of the data, along with manipulations necessary for the regressions, will be given. Hereafter, in section 4, the research method will be discussed. In section 5 the results of the research will be presented, and finally a conclusion will be given on the research along with suggestions for further research.
2. Theoretical framework
2.1 Basel IIn 1988 the first Basel accord was issued by the Basel committee on Bank
Supervision, hence its name, Basel I. This accord was first introduced in response to the early 1980 Latin American debt crisis, and aimed to regulate international banks by setting minimum capital requirements. The accord is an attempt to force banks to hold capital such that they are able to bear situations of economic distress.
Moreover, it is aimed to set international rules as otherwise banks that do not comply gain a competitive advantage compared with banks in other countries. Lower capital levels are prefered by banks as capital reserves can otherwise be used to generate returns, i.e. they pose an opportunity cost (History of the Basel Committee).
Weighted Assets, of which at least 4% was required to be Tier 1 capital or equity capital and disclosed reserves. Assets are categorized in four different categories each corresponding to a predetermined weight based on the credit risk of the before mentioned asset has thus named Risk Weighted asset. The assets are placed in the following categories: 0% (risk free), 20% (money-market loans), 50% (mortgages) and 100% (commercial and consumer loans). It gained support from the G10, the 10 biggest economies and was fully implemented by 1992.
Although Basel had introduced improvements regarding bank regulation, it also had some significant shortfalls. However, the Basel committee stated that is was always intended to progress over time. One of which was a critique of the risk
categories, which were insufficiently differentiated, the simplicity of Basel I implied that it was possible to work around the rules (Jones, 2000). It was possible to reduce level of required capital. Moreover Jones (2000) states that regulatory capital
arbitrage, the act of lowering regulatory measures of risk without reducing actual risk, is a possible flaw in the Basel I framework.
2.2 Basel II
The shortcomings of Basel I lead to a new capital framework of which a proposal was issued in June 1999, named Basel II Capital Accord. This revised version of the first Basel Accord consists of three pillars.
i. Minimum Capital Requirements
ii. Supervisory review of an institution’s capital adequacy
iii. Disclosure requirements of banks.
The main improvement in Basel II lies in the way regulatory capital reflects risk and thus the capital requirements. Improvements arise from a higher sensitivity to the risk the bank is subject to by its underlying assets. Another important driver for the new framework was to better deal with newly developed financial innovations since Basel I.
The change in the capital framework from Basel I to Basel II is mostly represented in the calculations of the risk-weighted assets. Relative to Basel I the capital requirements remains unchanged at 8% of total risk-weighted assets. Basel II aims to use a standardized approach, where in Basel I the risk weights were
assigned based on basic borrower categories, money market, mortgages or commercial and consumer loans, the new approach assigns risk weights on the basis of credit rating. Moreover an internal ratings-based approach was introduced,
where banks before relied on for example Standard & Poor based ratings, banks now had to assess risk of default on their own, so to form their own view and form a model applicable to its own situation (Lee, 2014). A possible downside is that banks might underestimate its own risk exposure.
There were some shortcomings in the Basel II agreement. The standardized approach was based on the credit ratings. However, credit ratings may change during economic distress, which can change the risk weights assigned to banks assets. As risk weights increase during economic downfall, capital requirements increase. Hence, the banks are forced to raise its capital reserves, called
procyclicality. Kashyap and Stein (2004) argue that raising capital during a recession could be very costly, and can possibly strengthen the recession. Daníelsson,
Embrechts, Goodhart, Keating, Muennich, Renault and Shin (2001) also state that Basel II is procyclical, and furthermore claim that its models underestimate the riskiness of assets.
2.3 Basel III
It soon became apparent that Basel II had some flaws and that there was need for improvement. The fall of the Lehman Brothers had confirmed this idea. In response to the financial crisis, discussions for a new Basel agreement commenced. Again reform was required with respect to regulatory capital, with a solution to the procyclicality that followed from the second Basel agreement. The new capital
structure aims for an increase common equity from two percent in Basel II to four and a half percent in Basel III. An increase in tier 1 capital from four percent to six
percent. Finally, a mandatory capital buffer of two and a half percent is required. Moreover, a counter-cyclical buffer of up to two and a half percent can be demanded following from high credit growth. Basel III will be implemented gradually initiating in 2013 the common equity ratio will amount to three and a half percent of risk-weighted assets, and minimum tier 1 capital will amount to four and a half percent. From 2019, Basel II will be fully implemented. An overview of implementation is provided in 1 in the appendix.
3. Literature Review
Risk management is an important aspect in the banking sector. However it is unclear how risk management affects banks their risk-taking behavior. There has yet to be an
answer on how risk management tools affect the riskiness of banks, Santemero and Trester (1998) for example state that the availability of risk management tools do not directly suggest that a bank is less risky, or vice versa. Research is divided into two different camps, some studies found evidence that banks that use risk management are found to be riskier. Cebenoyan and Strahan (2004) for example found that banks with enhanced abilities in credit risk management might have a higher leverage ratio, i.e. lend to riskier borrowers relative to other banks. However Ellul and Yerramilli (2012) disagree, they state that better risk management improves control and thus lead to lower risk.
Calem and Rob (1998) find a positive relationship between increases in minimum capital requirements and the levels of risk in banks, i.e. risk increases as capital requirements increase. Which seems weird given that minimal capital
requirements are set to decrease leverage ratio, and hence decrease risk. According to Blum (1999), who also finds that stricter capital regulation could increase a bank its risk level, an explanation might be that stricter regulation leads to less profit. This lower profit might entail that cost of bankruptcy decreases, and banks might become more reluctant towards their level of risk. Moreover lower profits are obviously not preferable, increasing levels of risk might increase expected profits. Gale (2010) points out another possible explanation for higher risk taking. As capital requirements increase and, therefore, less capital are available for lending purposes, the profits decrease. Gale points out that as a response banks might increase their interest rates on loans, to account for the lower profits. This action might entail that because of the higher interest rates on average riskier borrowers are attracted, increasing the risk exposure. This is in agreement with Kashyap, Stein and Hanson (2010) who find that an increase in minimum capital requirements of ten percentage points causes an increase in loan interest rates of 25-45 percentage points.
It is argued that credit derivatives such as default swaps may play a significant role in risk management for banks. JP Morgan first introduced credit default swaps in 1994, it was intended to sell off credit risk involved with Exxon Mobil to the European Bank of Reconstruction and Development, after which they were able to increase their lending to Exxon Mobil (Tett, 2009). The main idea of credit default swaps is to sell off risk, where the beneficiary pays an agreed upon fee to the guarantor, in order for the guarantor to take over risk. Another way of selling off risk is to sell off loans. However the main difference is that when credit default swaps are bought the buyer remains to be in control of the loan. An important side effect of these credit default swaps is that it decreases capital requirements. When issuing a loan, banks are required to hold capital based on the risk-weighted assets. These
risk weights are in its turn based on the credit rating of the borrower, the lower the credit rating, the higher the capital that must be held by the bank, to account for economic distress. In case a bank buys a credit default swap on a loan, the credit rating of the beneficiary, the seller, is represented in the bank its books. In some cases, this decreases capital requirements as these credit ratings may fall into a lower category of the risk-weighted assets.
Higher levels of available capital are preferable as it increases lending and thus returns, therefore decreasing risk-weighted assets is profitable. Credit default swaps as a tool for decreasing capital requirements was supported by several institutions. The Federal Reserve Board, for example, made a statement saying that banks should have the ability to use credit derivatives for capital relief (Shan, Tang & Yan, 2014). The Federal Reserve board would later issue a proposal on how credit derivatives held by banks should be incorporated into the capital requirements. Moreover, the ISDA, the International Swaps and Derivative Association, in a paper also stated that credit derivatives should be included in capital regulations. This proposal was later accepted by the Basel Committee and incorporated into the 2004 Basel II agreement (Shan, Tang and Yan, 2015). AIG, an insurance company, stated that 72% of its credit default swaps guarantor positions are used for capital relief (Shan, Tang and Yan, 2015).
Yet there are some problems in Basel II still with regards to credit derivatives. Basel II, for example, allows for mismatches in maturity and in facility amount. Thus, only part of the loan can be protected by credit default swaps. Another problem arising in the regulatory aspect of capital is that credit default swaps are allowed to be traded cross-sector, for example, insurance companies, meaning that non-banking firms are able to buy or sell credit default swaps. Non-non-banking firms are however not subject to banking regulations and thus are not required to hold reserve capital for the loans they guarantee protection on (Stulz, 2009).
4. Data and sample description.
4.1 databasesThe data used in this study has been collected from several databases. The first database, Loan Pricing Corporation Dealscan database, provides data on loans issued in the United States. The second database is the Bank Holding Companies database, which provides data on Credit default swaps positions from FRY-9 reports.
The last database is COMPUSTAT, a Standard & Poor database containing fundamental information on publicly held companies in the United States.
From the Loan Pricing Corporation’s Dealscan database all bank loans were retrieved, which account for most bank loans issued to firms in the United States. As Standard & Poor credit ratings were retrieved from COMPUSTAT, only the loan observations issued to firms available in the COMPUSTAT database are kept. We then restricted the sample to loans issued between 2006 and 2013, as this is the time period to be examined in this paper. Furthermore, loan observations were excluded if the lead arranger of the loan was not included in the collected credit default swaps data. This provides us with a dataset consisting of 21226 loan observations, of which there are 37 unique lead arrangers, and 3671 unique borrowers.
Credit ratings from Standard & Poor on the borrower firms in the loan dataset are retrieved from the COMPUSTAT database. Annual reported credit ratings are obtained and match the credit rating on each loan observation restricted to borrower firm and the year in which the loan was issued. We also match the notional amount of credit default swaps on which the lender act as guarantor restricted by year and lender id. Furthermore, the lenders total assets are matched based on the year with the credit default swap. Lastly, net income and market value are added.
4.2 Manipulation
To manipulate the data into a form with which we are able to perform regressions, a statistical program called Stata is used. A variable is created that divides the credit default swaps positions by the amount of assets of that bank. As banks holding more assets, subsequently are able to gold more credit default swaps. By dividing
positions by assets, positions will be represented relative to its assets.
In order to use the credit ratings in the sample, values must be assigned to the ratings. A new variable is generated called sp_rating which assigns a value of 1 to 22, where AAA gets assigned a 1, AA+ gets assigned a 2, AA gets assigned a 3, etc. till SD which gets assigned 22. Then weight categories are generated based on the Basel framework in which the highest rating category includes all credit ratings from AAA (1), till A- (7), this dummy variable is named low_risk_weight, and is assigned 1 if the credit rating of the borrower firm lies in this risk category for each loan observation. The next category includes credit ratings ranging from BBB+ (8), till BB- (13), this dummy variable is assigned a 1 for each observation that lies in this
range and is named medium_risk_weight. The last category, dummy variable
high_risk_weight, is assigned a 1 for all other credit risk, B+ (14) till SD (22).
Furthermore restricted by year and quarter the fractions of loans in each risk weight with respect to total loans are calculated for each lender. In the time frame of our sample the average low-risk loans account for 41,93 percent of total loans, medium-risk loans account for 44,22 percent, and subsequently high-risk loans amount to 13,84 percent of total loans. The variables are named
frac_low_risk_loans, Frac_medium_risk_loans and Frac_high_risk_loans.
Subsequently medium-risk loans and high-risk loans are added in a new variable, which is called Frac_medium_plus_high.
To account for the change from Basel I to Basel II, a dummy variable is created named basel_II_effective, it assigns a 1 to all observations from 2009, the year in which Basel II became fully effective. It assigns a zero for all variables before 2009. Basel III phases in initiating in 2013, although as we cannot have data on the full implementation of Basel III as this will not be until 2019. As we only have one year of Basel III, we will not include this in the analysis.
A distinction is made between banks that are active in credit default swap trading, and those that are not. A dummy variable is created which denotes a value of one if the bank trades credit default swaps, and zero otherwise. This paper defines that a bank is active in the credit default swap market if at any point in the analysis it buys or sells credit default swaps.
Furthermore, some control variables are added. At the loan level, three variables representing the type of loan are created. The first one being the amount of credit lines relative to total loans, the second control variable is the relative amount of term loans tot total loans. The last of the control variables is the average maturities of the loan observations, which is measured in months. All variables are created for each bank restricted to each quarter. Concerning bank level, two more control variables are added. The first is the profitability of the firm, represented by the return on assets. Defined as net income divided by assets. The second variable is the market value of the bank measured in billions of US dollars.
5. Methodology and results
5.1 Summary statistics
The following tables represent a summary of the banks in the sample for banks that are active in the credit default swap market and banks that are not respectively. In the appendix a summary of the whole sample can be found in table 2.3. In the tables maturity is measured in months and market value is in billions of US dollars.
Table 2.1 summary of statistics for credit default swaps active banks
19 banks
Mean
Std Deviation
Min
Max
Fraction low risk
loans
0,4387
0,2033
0,0119
0,9479
Fraction med risk
loans
0,4308
0,1903
0,0075
0,9324
Fraction high risk
loans
0,1305
0,1260
0,0053
0,6768
CDS/Assets
0,0819
0,0977
0,0000
0,4131
Fraction term loans
0,3069
0,1692
0,0000
1,0000
Fraction credit lines
0,5716
0,1947
0,0000
1,0000
Average maturity
47,8085
11,3492
10,0000
81,2000
ROA
0,0060
0,0071
-‐0,0143
0,0291
Market value
50,2931
68,3828
0,0000
238,0207
Table 2.2 summary of statistics for banks not active in credit default swaps.
18 banks
Mean
Std Deviation
Min
Max
Fraction low risk
loans
0,5386
0,2939
0,0547
0,9150
Fraction med risk
loans
0,4228
0,2841
0,0803
0,8911
Fraction high risk
loans
0,0386
0,0562
0,0003
0,1945
CDS/Assets
0,0000
0,0000
0,0000
0,0000
Fraction term loans
0,3516
0,3000
0,0000
1,0000
Fraction credit lines
0,5677
0,3315
0,0000
1,0000
Average maturity
49,7610
20,6647
6,0000
152,0000
ROA
0,0060
0,0095
-‐0,0600
0,0171
Market value
6,1415
7,3664
0,0000
31,4904
Of the 37 banks in the sample, nineteen banks have been active in the credit default swap market, fifteen of which have both sold and bought credit default swaps, eighteen banks have not traded in credit default swaps. From the tables it can be seen that the mean fraction of high-risk loans is 13,05% for banks that are active in
de credit default swap market and 3,86% for banks that are not. The average maturity is 47,81 months for credit default swaps active banks and 49,76 for non-active banks. The average return on assets is 0,6% for both categories. The average market value is 50,29 billion US dollars, whereas the average market value is 6,14 billion US dollars.
5.2 Method
To find an answer if Basel gives an effect as to the use of credit default swaps and following (Shan, Tang and Yan, 2015) a difference in differences approach is used. A regression analysis will be performed where the dummy variable cds_active
represents the difference in credit default swap active banks and non-active banks. Accounting for this difference will allow for differences found in loan riskiness to be addressed to the credit default swap usage. The estimation will be as follows:
𝐹𝑟𝑎𝑐_ℎ𝑖𝑔ℎ_𝑟𝑖𝑠𝑘_𝑙𝑜𝑎𝑛𝑠
= 𝛼 + 𝛽!∗ 𝐶𝑑𝑠_𝑎𝑠𝑠𝑒𝑡𝑠 ∗ 𝑏𝑎𝑠𝑒𝑙_𝐼𝐼_𝑒𝑓𝑓𝑒𝑐𝑡𝑖𝑣𝑒 + 𝛽!
∗ 𝐵𝑎𝑠𝑒𝑙_𝐼𝐼_𝑒𝑓𝑓𝑒𝑐𝑡𝑖𝑣𝑒 + 𝛽!∗ 𝐶𝑑𝑠_𝑎𝑠𝑠𝑒𝑡𝑠 + 𝛽!∗ 𝐹𝑟𝑎𝑐_𝑡𝑒𝑟𝑚_𝑙𝑜𝑎𝑛𝑠 + 𝛽!
∗ 𝐹𝑟𝑎𝑐_𝑐𝑟𝑒𝑑𝑖𝑡_𝑙𝑖𝑛𝑒 + 𝛽!∗ 𝐴𝑣𝑔_𝑚𝑎𝑡𝑢𝑟𝑖𝑡𝑦 + 𝛽!∗ 𝐶𝑑𝑠_𝑎𝑐𝑡𝑖𝑣𝑒 + 𝛽! ∗ 𝑅𝑂𝐴 + 𝛽!∗ 𝑀𝑎𝑟𝑘𝑒𝑡𝑣𝑎𝑙𝑢𝑒 + 𝛽!∗ 𝑦𝑒𝑎𝑟 𝑓𝑖𝑥𝑒𝑑 𝑒𝑓𝑓𝑒𝑐𝑡𝑠 + 𝜀!
In this equation year fixed effect represents a dummy variable for every year that is examined, 2006 till 2013. Furthermore included is an interaction term
cds_assets*basel_II_effective, with which we are able to find the difference in credit
default swap usage prior and post Basel goes into effect. The Basel term in this interaction terms is a dummy variable. Its coefficient represents the change in coefficient for credit default swaps relative to assets before and after basel II implementation.
The high-risk loans in this estimation are defined as borrowers with a credit rating from B+ till SD. In a second analysis medium-risk loans and high-risk loans are merged into one category, Frac_medium_plus_high. As medium-risk loans represent relatively risky loans as well, this will enable to find a difference in the riskiness of loans. The analysis is shown in the following equation.
𝐹𝑟𝑎𝑐_𝑚𝑒𝑑𝑖𝑢𝑚_𝑝𝑙𝑢𝑠_ℎ𝑖𝑔ℎ = 𝛼 + 𝛽!∗ 𝑐𝑑𝑠_𝑎𝑠𝑠𝑒𝑡𝑠 ∗ 𝑏𝑎𝑠𝑒𝑙_𝐼𝐼_𝑒𝑓𝑓𝑒𝑐𝑡𝑖𝑣𝑒 + 𝛽!∗ 𝑏𝑎𝑠𝑒𝑙_𝐼𝐼_𝑒𝑓𝑓𝑒𝑐𝑡𝑖𝑣𝑒 + 𝛽!∗ 𝑐𝑑𝑠_𝑎𝑠𝑠𝑒𝑡𝑠 + 𝛽!∗ 𝑓𝑟𝑎𝑐_𝑡𝑒𝑟𝑚_𝑙𝑜𝑎𝑛𝑠 + 𝛽!∗ 𝑓𝑟𝑎𝑐_𝑐𝑟𝑒𝑑𝑖𝑡_𝑙𝑖𝑛𝑒 + 𝛽!∗ 𝑎𝑣𝑔_𝑚𝑎𝑡𝑢𝑟𝑖𝑡𝑦 + 𝛽!∗ 𝑐𝑑𝑠_𝑎𝑐𝑡𝑖𝑣𝑒 + 𝛽!∗ 𝑅𝑂𝐴 + 𝛽! ∗ 𝑚𝑎𝑟𝑘𝑒𝑡𝑣𝑎𝑙𝑢𝑒 + 𝛽!∗ 𝑦𝑒𝑎𝑟 𝑓𝑖𝑥𝑒𝑑 𝑒𝑓𝑓𝑒𝑐𝑡𝑠 + 𝜀!
Furthermore, it is tested whether a change in credit default swaps usage relative to assets is caused by a change in the fraction of high-risk loans relative to total loans. The same control variables are used as in the difference in differences method, however cds_active is omitted, as a change in credit default swaps is not relevant for non-credit default swaps active banks. It is estimated using a panel data regression represented by the following equation:
𝐹𝑟𝑎𝑐_ℎ𝑖𝑔ℎ_𝑟𝑖𝑠𝑘_𝑙𝑜𝑎𝑛𝑠!"
= 𝛼! + 𝛽!∗ 𝑐𝑑𝑠_𝑎𝑠𝑠𝑒𝑡𝑠!"+ 𝛽!∗ 𝑅𝑂𝐴!"+ 𝛽!∗ 𝑀𝑎𝑟𝑘𝑒𝑡𝑣𝑎𝑙𝑢𝑒!"+ 𝛽!
∗ 𝐵𝑎𝑠𝑒𝑙_𝐼𝐼_𝑒𝑓𝑓𝑒𝑐𝑡𝑖 𝑣𝑒 + 𝛽!∗ 𝐹𝑟𝑎𝑐_𝑡𝑒𝑟𝑚_𝑙𝑜𝑎𝑛𝑠!"+ 𝛽!
∗ 𝐹𝑟𝑎𝑐_𝑐𝑟𝑒𝑑𝑖𝑡_𝑙𝑖𝑛𝑒𝑠!"+ 𝛽!∗ 𝑚𝑎𝑡𝑢𝑟𝑖𝑡𝑦!" + 𝜀!
This analysis is repeated where the independent variable represents the fraction of medium-risk plus high-risk loans. The equation is then as follows:
𝐹𝑟𝑎𝑐_𝑚𝑒𝑑𝑖𝑢𝑚_𝑝𝑙𝑢𝑠_ℎ𝑖𝑔ℎ!"
= 𝛼! + 𝛽!∗ 𝑐𝑑𝑠_𝑎𝑠𝑠𝑒𝑡𝑠!"+ 𝛽!∗ 𝑅𝑂𝐴!"+ 𝛽!∗ 𝑀𝑎𝑟𝑘𝑒𝑡𝑣𝑎𝑙𝑢𝑒!"+ 𝛽!
∗ 𝐵𝑎𝑠𝑒𝑙_𝐼𝐼_𝑒𝑓𝑓𝑒𝑐𝑡𝑖 𝑣𝑒 + 𝛽!∗ 𝐹𝑟𝑎𝑐_𝑡𝑒𝑟𝑚_𝑙𝑜𝑎𝑛𝑠!"+ 𝛽!
5.3 Results
5.3.1 difference in differences
The regression outcomes are presented in the table below. In this table year fixed effects and the constant term are omitted for briefness, a full table of the outcomes can be found in table 3.1 and 3.2 for fraction of high-risk loans and fraction of medium-risk plus high-risk loans respectively.
Frac_high_risk_loans Frac_medium_plus_high
Cds_assets*basel_II_effective
0,1808
0,9465*
(0,2734)
(0,4900)
Fraction term loans
0,6580***
0,9404**
(0,1719)
(0,3081)
Fraction credit lines
0,3316**
0,6671**
(0,1531)
(0,2743)
Average maturity
0,0004
-‐0,0041
(0,0018)
(0,0032)
CDS/Assets
-‐0,1640
-‐0,8859**
(0,2502)
(0,4484)
Basel II effective
0,0151
-‐0,2398***
(0,0462)
(0,0827)
ROA
-‐3,0471
-‐5,7969
(2,8930)
(5,1855)
Market value
-‐8,62e-‐06
-‐4,39e-‐05
(0,0002)
(0,0003)
CDS active
0,1431***
0,1226*
(0,0397)
(0,0712)
*, **, *** Represent significance levels at 10%, 5% and 1% respectively.
As seen in the table a positive coefficient is found for the effect of Basel
implementation on high-risk loans. This implies that after Basel II was implemented the fraction of high-risk loans increased relative to before. However, this effect has not been found significant at the ten percent level. When medium-risk loans and high-risk loans are merged, a negative coefficient is found for the effects of Basel II on the fraction of medium-risk plus high-risk loans. This effect has been found
significant at the one percent level, implying that loan riskiness decreased after Basel II.
Furthermore, positive coefficients are found for the dummy variable
the fraction of high-risk loans and medium-risk plus high-risk loans respectively. It implies that the fraction of high-risk loans has increased for firms that trade in credit default swaps.
The interaction term represents the difference in coefficient for cds_assets, credit default swap usage relative to assets, before and after Basel II. Its significance level is then tested in the following way:
𝐻!: 𝑐𝑑𝑠_𝑎𝑠𝑠𝑒𝑡𝑠 ∗ 𝑏𝑎𝑠𝑒𝑙_𝐼𝐼_𝑒𝑓𝑓𝑒𝑐𝑡𝑖𝑣𝑒! – 𝑐𝑑𝑠_𝑎𝑠𝑠𝑒𝑡𝑠 ∗ 𝑏𝑎𝑠𝑒𝑙_𝐼𝐼_𝑒𝑓𝑓𝑒𝑐𝑡𝑖𝑣𝑒!= 0 𝐻!: 𝑐𝑑𝑠_𝑎𝑠𝑠𝑒𝑡𝑠 ∗ 𝑏𝑎𝑠𝑒𝑙_𝐼𝐼_𝑒𝑓𝑓𝑒𝑐𝑡𝑖𝑣𝑒! – 𝑐𝑑𝑠_𝑎𝑠𝑠𝑒𝑡𝑠 ∗ 𝑏𝑎𝑠𝑒𝑙_𝐼𝐼_𝑒𝑓𝑓𝑒𝑐𝑡𝑖𝑣𝑒! ≠ 0
From the table it can be seen that a positive difference is found for the fraction of high-risk loans. However this difference has not been found significant at any level. For the fraction of medium-risk plus high-risk loans a positive difference has also been found, this difference is significant at the ten percent level.
When comparing the results of the regression analysis and the difference test, we find that the fraction of high-risk loans increased after Basel II initiation. A significant positive relationship is found in the fraction of high-risk loans and credit default swap active banks. Moreover, a positive difference is found in credit default swaps relative to assets in the interaction term, however, this result not significant. Therefor the difference is not found significantly different from zero, such that it cannot be stated that credit default swaps usage on the fraction of high-risk loans increased after Basel II implementation
In the second analysis a decrease in the fraction of medium-risk plus high-risk loans is found, which is significant at the one percent level. Moreover, a positive relationship is found with credit default swap active firms. Lastly, we find a positive difference for credit default swap relative to assets in the interaction term, which is significant at the 10 percent level. Implying that weak evidence is found that credit default swap usage increased for the fraction of medium-risk plus high-risk loans.
5.3.2 Panel data
The following table represents the coefficients found for the panel data regression. A more detailed table for the fraction of high-risk loans and the fraction of medium-risk plus high risk loans can be found in table 4.1 and table 4.2 respectively.
Frac_high_risk_loans Frac_medium_plus_high
Change in cds/assets
0,4548
-‐0,4558
(1,0751)
(2,0868)
ROA
2,0414
7,9911
(3,8633)
(7,4986)
Market value
0,0000
0,0000
(0,0000)
(0,0000)
Basel II effective
0,0427
0,2032
(0,0760)
(0,1475)
Fraction term loans
-‐0,2077
-‐0,8676
(0,3265)
(0,6337)
Fraction credit lines
-‐0,1582
-‐0,6393
(0,3040)
(0,5900)
Maturity
-‐0,0024
-‐0,0019
(0,0025)
(0,0048)
Constant
14,3539
54,2284
(29,2335)
(56,7409)
*, **, *** Represent significance levels at 10%, 5% and 1% respectively.
It can be seen in the table that a positive coefficient is found for the change in relative usage of credit default swaps on the change in the fraction of high-risk loans. Implicating that an increase in credit default swaps usage causes an increase in the fraction of high-risk loans. A negative value is found for the change in credit default swaps relative to assets for the fraction of medium-risk plus high-risk loans. Which implicates that an increase in relative credit default swaps usage causes a decrease in the fraction of medium-risk plus high-risk loans. However both these coefficients are not significant, therefore the values are not significantly different from zero.
6. Conclusion and limitations
6.1 Conclusion
The aim of this study has been to address the relationship between credit default swaps usage and bank lending. Credit default swaps are a financial derivative that is able to reduce the amount of regulatory capital a bank needs to hold by selling off risk. Capital reserves can be costly to a bank, credit default swaps can enable a bank to generate higher returns without holding excess capital reserves.
Previous research found that increases in regulatory capital increases risk levels in banks Rob and Calem (1998). As increases in regulatory capital, should imply decreases in bank risk, this increase might arise elsewhere. The use of credit default swaps decreases required reserve capital. The main question of this paper, therefore, is: Do credit default swap using banks issue more risky loans relative to non-credit default swap using banks? This paper aims to provide a question using a difference in differences approach, comparing the fractions of riskiness of loans before and after implementation of Basel II, between credit default swap using banks and non-credit default swap using banks.
Following theory, a positive relation between loan riskiness and the use of credit default swaps was expected on bank level. This paper found evidence supporting the positive relationship between credit default swaps usage and the relative amount of high-risk loans. However, the difference in relative positions of credit default swaps before and after Basel II has not been found significant. In our analysis of medium-risk loans plus high-risk loans combined again a positive relationship is found for credit default swaps using banks and loan riskiness. Weak statistical evidence implied an increase in the relationship between credit default swaps usage before and after higher capital regulation and loan riskiness.
Moreover in the panel data regression no evidence is found that an increase in credit default swaps usage causes an increase the riskiness of loans. Taking into account all findings this paper is not able to provide a definitive answer to the relationship between credit default swaps and risk levels of banks.
6.2 Limitations and suggestions for further research.
There were several limitations encountered in this paper. The first limitation is missing data, due to the relatively small amount of observations the model might be
inferior in estimating coefficients. For further research, it is encouraged to use a larger sample.
Another limitation of this research is that we were unable to retrieve credit default swap data matching the loan observations. Instead, we matched quarterly positions on credit default swaps and matched the loan observation to the following quarter. For further research it advised to examine the effects of changes in credit default swaps usage on the riskiness of loans in a different quarter. Due to time limitations this paper was not able to do so.
References
Acharya, V. V., & Johnson, T. C. (2007). Insider trading in credit derivatives. Journal of Financial Economics, 84(1), 110-141.
Blum, J. (1999). Do capital adequacy requirements reduce risks in banking?. Journal of Banking & Finance, 23(5), 755-771.
Calem, P., & Rob, R. (1999). The impact of capital-based regulation on bank risk-taking. Journal of Financial Intermediation, 8(4), 317-352.
Cebenoyan, A. S., & Strahan, P. E. (2004). Risk management, capital structure and lending at banks. Journal of Banking & Finance, 28(1), 19-43.
Chesney, M., Stromberg, J., & Wagner, A. F. (2012). Managerial incentives to take asset risk. Swiss Finance Institute Research Paper, (10-18).
Danielsson, J., Embrechts, P., Goodhart, C., Keating, C., Muennich, F., Renault, O., & Shin, H. S. (2001). An academic response to Basel II.
Gale, D. (2010). Capital regulation and risk sharing. International Journal of Central Banking, 6(4), 187-204.
History of the Basel Committee. (n.d.). Retrieved from
http://www.bis.org/bcbs/history.htm
Jones, D. (2000). Emerging problems with the Basel Capital Accord: Regulatory capital arbitrage and related issues. Journal of Banking & Finance, 24(1), 35-58.
Kashyap, A. K., & Stein, J. C. (2004). Cyclical implications of the Basel II capital standards. Economic Perspectives-Federal Reserve Bank Of Chicago, 28(1), 18-33.
Kashyap, A. K., Stein, J. C., & Hanson, S. (2010). An analysis of the impact of ‘substantially heightened’capital requirements on large financial institutions. Booth School of Business, University of Chicago, mimeo.
Lee, E. (2014). Basel III and Its New Capital Requirements, as Distinguished From Basel II. The Banking Law Journal, 131(1), 27-69.
Santomero, A. M., & Trester, J. J. (1998). Financial innovation and bank risk taking. Journal of Economic Behavior & Organization, 35(1), 25-37.
Santos, J. A. (2001). Bank capital regulation in contemporary banking theory: A review of the literature. Financial Markets, Institutions & Instruments, 10(2), 41-84.
Shan, S. C., Tang, D. Y., & Yan, H. (2014). Regulatory capital and bank lending: The role of credit default swaps. Working paper.
Shan, S. C., Tang, D. Y., & Yan, H. (2015). When Is CDS Trading Innocuous?. Available at SSRN 2608776.
Stulz, R. M. (2009). Credit default swaps and the credit crisis (No. w15384). National Bureau of Economic Research.
Tett, G. (2009). Fool's gold: How the bold dream of a small tribe at JP Morgan was corrupted by Wall Street greed and unleashed a catastrophe. Simon and Schuster.
Appendix.
1. Basel III evolution
Table 1.
2. Summary statistics
In these tables summaries of statistics are presented. In the tables average maturity is measured in months. Market value is measured in billions of US dollars.
Table 2.1 summary of statistics for credit default swaps active banks
19 banks
Mean
Std Deviation
Min
Max
Fraction low risk
loans
0,4387
0,2033
0,0119
0,9479
Fraction med risk
loans
0,4308
0,1903
0,0075
0,9324
Fraction high risk
loans
0,1305
0,1260
0,0053
0,6768
CDS/Assets
0,0819
0,0977
0,0000
0,4131
Fraction term loans
0,3069
0,1692
0,0000
1,0000
Fraction credit lines
0,5716
0,1947
0,0000
1,0000
Average maturity
47,8085
11,3492
10,0000
81,2000
ROA
0,0060
0,0071
-‐0,0143
0,0291
Market value
50,2931
68,3828
0,0000
238,0207
Table 2.2 summary of statistics for banks not active in credit default swaps.
18 banks
Mean
Std Deviation
Min
Max
Fraction low risk
loans
0,5386
0,2939
0,0547
0,9150
Fraction med risk
loans
0,4228
0,2841
0,0803
0,8911
Fraction high risk
loans
0,0386
0,0562
0,0003
0,1945
CDS/Assets
0,0000
0,0000
0,0000
0,0000
Fraction term loans
0,3516
0,3000
0,0000
1,0000
Fraction credit lines
0,5677
0,3315
0,0000
1,0000
Average maturity
49,7610
20,6647
6,0000
152,0000
ROA
0,0060
0,0095
-‐0,0600
0,0171
Market value
6,1415
7,3664
0,0000
31,4904
Table 2.3 Summary of statistics for whole sample
Variable
Mean
Std Deviation
Min
Max
Fraction low risk
loans
0,4193
0,2348
0,0119
0,9479
Fraction med risk
loans
0,4422
0,2235
0,0075
0,9324
Fraction high risk
loans
0,1384
0,1451
0,0003
0,8782
CDS/Assets
0,0481
0,0850
0,0000
0,4131
Fraction term loans
0,3465
0,2441
0,0000
1,0000
Fraction credit
lines
0,5426
0,2713
0,0000
1,0000
Average maturity
50,4072
18,8704
3,0000
324,0000
ROA
0,0058
0,0085
-‐0,0600
0,0291
3. Regression outcome.
Table 3.1
Regression outcome with dependent variable the fraction of high-risk loans
Frac. high risk
loans
Coefficient Std. Err.
t-‐value
P>t
[95%
Conf.
Interval]
Cds_assets*basel
_II_effective
0,1808
0,2734
0,6600
0,5090
-‐0,3599
0,7216
Frac_term_loans
0,6580
0,1719
3,8300
0,0000
0,3180
0,9979
Frac_credit_lines
0,3316
0,1531
2,1700
0,0320
0,0289
0,6343
Avg_maturity
0,0004
0,0018
0,2400
0,8090
-‐0,0031
0,0039
Cds_assets
-‐0,1640
0,2502
-‐0,6600
0,5130
-‐0,6588
0,3309
Basel_II_effective
0,0151
0,0462
0,3300
0,7450
-‐0,0762
0,1064
ROA
-‐3,0471
2,8930
-‐1,0500
0,2940
-‐8,7694
0,2675
Marketvalue
0,0000
0,0000
-‐0,0600
0,9560
0,0000
0,0000
Cds_active
0,1431
0,0397
3,6000
0,0000
0,0645
0,2216
_Iyear_2007
0,0064
0,0414
0,1500
0,8780
-‐0,0755
0,0882
_Iyear_2008
0,0049
0,0591
0,0800
0,9350
-‐0,1121
0,1218
_Iyear_2009
-‐0,0207
0,0536
-‐0,3900
0,7000
-‐0,1267
0,0854
_Iyear_2010
-‐0,0343
0,0413
-‐0,8300
0,4080
-‐0,1161
0,0475
_Iyear_2011
-‐0,0294
0,0408
-‐0,7200
0,4720
-‐0,1101
0,0513
_Iyear_2012
-‐0,0507
0,0401
-‐1,2600
0,2090
-‐0,1301
0,0287
_Iyear_2013
0,0000
(omitted)
_cons
-‐0,4062
0,1302
-‐3,1200
0,0020
-‐0,6637
-‐0,1486
Table 3.2
Regression outcome with dependent variable the fraction of medium-‐risk plus high-‐
risk loans.
Frac medium
plus high
Coefficient
Std. Err.
t
P>t
[95%
Conf.
Interval]
Cds_assets*basel_
II_effective
0,9465
0,4900
1,9300
0,0560
-‐0,0228
191579,0
000
Frac_term_loans
0,9404
0,3081
3,0500
0,0030
0,3310
154971,0
000
Frac_credit_lines
0,6671
0,2743
2,4300
0,0160
0,1245
1209729,
0000
Avg_maturity
-‐0,0041
0,0032
-‐1,2900
0,2010
-‐0,0104
0,0022
Cds_assets
-‐0,8859
0,4484
-‐1,9800
0,0500
-‐1,7729
0,0011
Basel_II_effective
-‐0,2398
0,0827
-‐2,9000
0,0040
-‐0,4034
-‐0,0762
ROA
-‐5,7969
5,1855
-‐1,1200
0,2660
-‐16,0536
4,4598
Marketvalue
0,0000
0,0000
1,5700
0,1190
0,0000
0,0000
Cds_active
0,1226
0,0712
1,7200
0,0880
-‐0,0183
0,2634
_Iyear_2007
-‐0,0855
0,0742
-‐1,1500
0,2510
-‐0,2322
0,0612
_Iyear_2008
-‐0,0887
0,1060
-‐0,8400
0,4040
-‐0,2983
0,1210
_Iyear_2009
-‐0,1916
0,0961
-‐1,9900
0,0480
-‐0,3817
-‐0,0016
_Iyear_2010
-‐0,0894
0,0741
-‐1,2100
0,2300
-‐0,2360
0,0571
_Iyear_2011
-‐0,0036
0,0731
-‐0,0500
0,9610
-‐0,1482
0,1410
_Iyear_2012
-‐0,1027
0,0720
-‐1,4300
0,1560
-‐0,2451
0,0396
_Iyear_2013
0,0000
(omitted)
_cons
0,2088
0,2334
0,8900
0,3720
-‐0,2528
0,6704
4. Panel data regressions
Table 4.1 Panel data regression on change in the fraction of high-risk loans.
Change in frac high-‐
risk loans
Coefficient Std. Err.
t-‐value
P>t
[95% Conf. Interval]
Change in
cds_assets
0,4548
1,0751
0,4200
0,6730
-‐1,6852
2,5948
ROA
2,0414
3,8633
0,5300
0,5990
-‐5,6484
9,7311
Marketvalue
0,0000
0,0000
0,4000
0,6900
0,0000
0,0000
Basel_II_effective
0,0427
0,0760
0,5600
0,5760
-‐0,1085
0,1940
Frac_term_loans
-‐0,2077
0,3265
-‐0,6400
0,5270
-‐0,8576
0,4422
Frac_credit_lines
-‐0,1582
0,3040
-‐0,5200
0,6040
-‐0,7632
0,4468
Avg_maturity
-‐0,0024
0,0025
-‐0,9800
0,3310
-‐0,0073
0,0025
Constant
14,3539
29,2335
0,4900
0,6250
-‐43,8340
72,5418
Table 4.2 Table 4.1 Panel data regression on change in the fraction of medium-risk plus high-risk loans.
