Ebook Exposure at Default of Unsecured Credit Cards
Credit cards, home equity lines of credit, and revolving lines of credit are examples of revolving retail exposures, whereas mortgages, auto loans, and home equity loans are examples of term loans. There are many differences between revolving exposures and term loans for example, revolving exposures are open-ended, while the term loans are close-ended; borrowers pay interest only on funds drawn from the revolving credit; and a qualifying revolving exposure (QRE) is unsecured and unconditionally cancelable by the lender to the fullest extent permitted by federal law. In terms of repayment, the interest and principal payment of term loans are usually equal monthly installments over the life of the loan, whereas revolving credits allow the consumer to repay any amount at any time as long as the preestablished minimum monthly payment is met.
Revolving retail credit products offer convenience and financial flexibility that term loans lack. They can provide borrowers access to funds when deterioration in credit quality prevents them from borrowing through other credit channels. Agarwal, Ambrose, and Liu (2006) studied home equity line utilization at and after origination and found that borrowers with greater expectations of a decline in future credit quality originate credit lines to preserve financial flexibility. Furthermore, borrowers with higher FICO scores (a measure of credit risk) tend to have higher credit utilization at origination, consistent with the theoretical models predicting that borrowers with lower credit quality signals preserve flexibility by utilizing a lower amount of credit at origination relative to borrowers with higher credit quality signals. Agarwal and others (2006) also found that borrower credit line utilization increases in response to drops in borrower’s FICO scores, consistent with the theoretical “credit risk” prediction of Strahan (1999).
Using a sample of Spanish corporate credit lines, Jiménez, Lopez, and Saurina (2007) found that among a wide variety of loan-level, firm-level, lender-level, and macro factors that affect line utilization rates, the most important factors are a firm’s default experience, the age of the credit line, and the length of the banking relationship. In particular, firms that default on their credit lines during the sample period have significantly higher line utilization rates, and these rates increase as the default year approaches. In addition, the age of the credit line and the length of the banking relationship are negatively related to utilization rate.
The empirical findings of Agarwal and others (2006) and Jiménez and others (2007) may have some implications for exposure at default (EAD) under the Basel II Capital Accord. Since credit line utilization increases significantly as borrower credit quality deteriorates, the amount outstanding might become significantly higher in the event of default, resulting in an EAD much higher than outstanding at the time of capital calculation. In other words, without considering the correlation between the borrower’s probability of default and the corresponding EAD, economic capital models may underestimate the impact of credit losses. It is also necessary to evaluate revolving retail exposures during downturn periods when the credit quality of many borrowers is deteriorating and aggregate default rates are significantly higher than average.
However, it is not clear from Agarwal and others (2006) and Jiménez and others (2007) whether the increase in credit line utilization as borrower credit quality deteriorates is caused mainly by an increase in drawn amount by the borrower or, rather, a cutback of the credit line by the lender. On one hand, as borrowers approach default, their financial conditions deteriorate and they are likely to tap the undrawn line as a source of additional funds, resulting in higher EAD. On the other hand, lenders may cut back credit lines to reduce potential losses if they observe deterioration of borrower credit quality; as a result, higher credit utilization does not necessarily imply proportional increase in EAD.
As illustrated in figure 1, the behavior of both borrowers and lenders while they “race to default” jointly determines EAD. Under scenario A, the lender does not cut back the credit line, the initial utilization is 50 percent ($500 out of $1,000), utilization at default increases by 20 percentage points to 70 percent ($700 out of $1,000), and EAD is $700. Under scenario B, the borrower exhibits the same drawdown behavior; however, the lender quickly reduces the credit line to $600 in observation of the deteriorating credit quality of the borrower. As a result, under scenario B, although utilization at default is 100 percent (which is higher than under scenario A), the EAD is only $600 (less than the $700 under scenario A). To date, there is no empirical study that shows whether and how often banks cut back lines on revolving retail credits, and whether the “race to default” differs on accounts of different credit constraint and delinquency status. Our empirical analysis in section 3 sheds light on this.
This study focuses on EAD, one of the key risk parameters in Basel II minimum risk-based capital requirements for credit risk. EAD is a bank’s expected gross dollar exposure (including net accrued but unpaid interest and fees) for a facility upon the borrower’s default. For fixed exposures, such as bullet or term loans, EAD is simply the amount outstanding at the time of capital calculation (plus accrued but unpaid interest and fees). For variable exposures, such as lines of credit, EAD is the current outstanding plus an estimate of additional drawdowns and accrued but unpaid interest and fees up to the time of default.
To quantify EAD, banks can estimate possible increases in exposures, which consist of additional drawdowns plus net accrued but unpaid interest and fees between the time of capital calculation and a potential default by the borrower over a fixed horizon for example, one year. Increases in exposure from the time of capital calculation to the time of default as a percent of undrawn amount at the time of capital calculation is often referred to as loan equivalent (or LEQ) by the industry. EAD can then be estimated as the current outstanding balance plus the estimated LEQ times the current undrawn, i.e., EAD($) = outstanding ($) + LEQ x [credit line ($)–outstanding($)]. In figure 1, LEQs of the defaulted loan in scenarios A and B are 40 percent and 20 percent, respectively, whereas the line utilization rates at default in scenarios A and B are 70 percent and 100 percent, respectively.
Besides LEQ, alternative approaches, such as credit conversion factor (CCF) and EAD factor (or EADF), can also be used to estimate EAD. CCF is the exposure at default as a percentage of the outstanding balance at the time of capital calculation, whereas EADF is the exposure at default as a percentage of the credit line at the time of capital calculation. The CCF estimation does not incorporate information on the credit line, and EADF does not incorporate information about the outstanding balance at the time of capital calculation. LEQ, on the other hand, incorporates both pieces of information, and therefore has the potential to provide the most accurate estimation of EAD. However, when a consumer maxes out the credit line (has zero undrawn), LEQ is not defined. Also when the borrower is getting close to maxing out the credit line, with a small undrawn amount, LEQ can be highly unstable even though it is defined. In this case, CCF and EADF can be used to estimate EAD instead. As such, while our empirical modeling and analysis mainly focuses on LEQ for accounts with undrawn amounts of more than $50, we also quantify CCF and EADF when accounts have zero or very small (<$50) undrawn lines or when accounts are closed or over limit.
In the new Basel II capital framework, probability of default (PD), loss given default (LGD), and exposure at default (EAD) are the key risk parameters that jointly determine the minimum credit risk capital required to cover unexpected credit losses of financial institutions. Thanks to credit scoring introduced in the 1980s and widely adopted in the mid-1990s, factors driving PD of retail credit exposures have been fairly well studied and understood. Financial institutions have commonly used PD models in underwriting and account management of retail credits. While there are few but increasing numbers of published studies on retail LGD, as surveyed in Qi and Yang (2009), we have not seen any published studies focusing on retail EAD, although there are very few empirical studies on wholesale EAD, as surveyed here.
In their technical appendix, Asarnow and Marker (1995) present some wholesale EAD analysis based on a small sample of 50 large corporate loans at Citibank from 1988 to 1993. They find that while lower credit quality borrowers have higher line utilization, they appear to have lower LEQ. Higher-quality borrowers may have higher LEQ because they are subject to fewer restrictions and covenants and less strict monitoring, and in times of trouble they can draw down available credit without interference from the bank. The Chase study by Araten and Jacobs (2001) uses a much larger sample (408 defaulted facilities observed at various times) from 1995 to 2000. They find that LEQ generally decreases with increasing risk, but results are not robust. They also find that LEQ is not differentiated by lending organization (e.g., middle market vs. large corporate), commitment type, commitment size, domestic vs. foreign, or industry. Moral (2006) provides a survey of pros and cons of various EAD estimation methods and illustrates these points using a set of defaulted facilities from a small and medium enterprise (SME) portfolio. Jacobs (2008) expands the previous empirical works on wholesale EAD by considering additional risk drivers, various measures of EAD risk, and in-sample and out-of-sample performance of alternative statistical models using updated and expanded data on 3,886 defaulted instruments for 683 borrowers from 1985 to 2007.
Most of the advanced IRB (internal-ratings-based) banks in the U.S. banking industry do not have EAD models or develop EAD segmentation using EAD-specific risk drivers, and common industry practice is to use average LEQ or CCF for each retail product or for each PD segment. Since little is known about retail EAD and its risk drivers, and the U.S. Basel II Final Rule is not specific about the approach to EAD, we try to fill this gap in the academic literature and in industry practices. To summarize our major findings up front, we use a large national sample of unsecured credit card defaults and find that a set of variables consisting of borrower attributes, account information variables, and macro factors are significant drivers of EAD. Models incorporating these factors show better predictive accuracy. We also find that in the “race to default,” borrowers are more likely to draw down additional funds but lenders rarely cut back credit limits, although lenders make it less easy to draw down additional funds as borrowers become more severely delinquent. The rest of the paper is organized as follows: Data description and summary statistics are provided in section 2. The dynamic behavior of borrowers and lenders as the borrowers approach default is analyzed in section 3. Correlation analysis between key variables is provided in section 4. Results of regression analysis and predictive accuracy of alternative models are reported in sections 5 and 6, respectively. Conclusions are drawn in section 7.
Download
PDF Ebook Exposure at Default of Unsecured Credit Cards
Posted in :
