*An earlier version of this paper appears in the Journal of Credit Risk, Vol. 10, No. 1, 2014.
1. Overview
A major source of firm funding and liquidity, credit lines can pose significant credit risk to the underwriting banks. By the end of 2019 Q1, the unused commitment across all depository institutions in the U.S. was USD$8 trillion, almost 44% of these institutions’ aggregate assets (USD$18 trillion).1
Understanding credit line usage patterns is important to banks, for many reasons and different applications. Exposure at default (EAD), the usage estimation conditional upon default, enters into the regulatory capital calculation under Basel II, together with probability of default (PD) and loss given default (LGD). Economic capital calculations require the assessment of the loss risk associated with credit line drawdown. Credit line usage also factors into banks’ liquidity management, as simultaneous drawdown on the credit lines by different borrowers can significantly affect both the asset side and the liability side of the balance sheet. Basel III’s liquidity coverage ratio, intended to ensure that banks have adequate funding sources relative to the total demand of cash outflows, includes credit line drawdown. Current Expected Credit Losses (CECL), introduced by Financial Accounting Standards Board (FASB) in 2016 and effective starting December 15, 2019, requires estimation of expected losses over the life of the loans, for both on-balance sheet and off-balance sheet items. Institutions thus need reliable estimation of EAD for both funded part and unfunded part of the loans throughout the life of the loans. For the funded part, institutions can look at the usage model for EAD estimation. Usage can be used to estimate the exposure. For the unfunded part, one way is to use credit conversion factor (CCF)2 to convert off-balance-sheet exposure to EAD. Understanding CCF behavior empirically is important for institutions to make reasonable CCF assumptions. Basel IV, which requires financial institutions to implement by 2022, makes more conservative restrictions on CCF. In addition, usage pattern is an important component in pricing these lines, as the credit line valuation requires the estimation of the expected drawdown. Further, credit line usage behavior can potentially provide additional information about borrowers’ financial distress and defaults.
Our study yields a number of interesting findings, with significant and practical implications. The sample covers valid usage information from 2000–2017. We pool the dataset from multiple U.S. financial institutions. We study the extent to which borrowers draw down their credit lines and link usage and LEQ to line-level and firm-level variables.
We find that defaulted firms draw down credit lines heavily when approaching default. The mean usage and median usage ratios for defaulted firms are 81% and 97%, respectively, which suggests the risky nature of these facilities. A large portion of the firms draw down almost the full amount of their credit lines. Results suggest that troubled firms draw down credit lines long before default; the mean (median) usage ratio two years prior to default is 76% (92%). Troubled firms also have more additional draw down in the last year before default. The mean LEQ and median LEQ for defaulted firms at default quarter are 47% and 48%, respectively. In comparison, non-defaulted firms, on average, show lower usage levels than defaulted firms. The mean and median usage ratios for non-defaulted firms are 52% and 56%, respectively. The mean and median LEQ for non-defaulted firms are 30% and 14%, respectively. After controlling for other factors such as default probability, the difference in usage between defaulted and non-defaulted firms becomes smaller, suggesting that it can be explained by the two groups’ different risk profiles. We observe the same behavior for LEQ.
Our results show that usage increases with default probability. Risky borrowers tend to draw down lines more. At the same time, firms with an internal “non-pass” grade one year prior to default tend to make fewer additional drawdowns from the unused part, suggesting banks’ monitoring of credit lines reduces line usage. We further find that line usage is related to loan purpose type, collateral type, commitment size and usage one quarter prior. Usage ratios are higher during times of recession.
The rest of the paper is organized as follows. Section 2 reviews the existing studies on credit line usage. Section 3 describes the data. Section 4 discusses usage measurements. Section 5 presents our results. Section 6 concludes.
Credit line usage literature generally falls into two main categories. The first category studies usage and LEQ measurement from a bank-risk management perspective. This category of literature provides statistics on usage and LEQ measurements and findings on how they vary with line-level, borrower-level, or macro-level variables. The results are useful to banks for their capital calculation, risk management practices, as well as satisfying regulatory requirements. Asarnow and Marker (1995) study a sample of syndicated loans in the Citibank Bank Loan Index and find that average line usage is higher for better rated firms. Bacham and Yang (2019) uses Moody’s Credit Research Dataset and empirically proves that higher usage is associated with higher default risk. All else equal, term loans default risk is similar to revolving lines with 40-60% usage. Araten and Jacobs (2001) is one of the most cited papers. They examine a sample of 399 defaulted borrowers using Chase Bank’s internal data and report an average loan equivalent exposure (LEQ) of 32%. They also find that LEQ is higher for better rated firms. They do not observe a relationship between LEQ and either commitment size or maturity. Jiménez et al. (2009a,b) look at the EAD and usage for a large sample of Spanish firms. They report an average one-year LEQ of 48%. Defaulted firms have higher usage than non-defaulted firms. Lines with smaller commitment, shorter maturity or with collateral have higher LEQ. Jacobs (2010) examines line usage data on 720 defaulted borrowers with Moody’s or Standard and Poor’s ratings and reports a one-year LEQ of 38%. He also finds that the EAD is lower for adversely rated firms. Bag and Jacobs (2012) finds that higher LEQ level assumption will increase the volatility of the EAD distribution. Pan (2009) shows that utilization percentage, remaining life of the borrower, age of the borrower, industry, and risk ratings are important factors that impact LEQ. Emery et al. (2008) studies the debt structure evolution during the three years prior to default for 291 defaulted firms. They find that revolver drawdowns increase as firms approach default. The average usage rate at default is approximately 70%. Moral (2006) and Taplin et al. (2007) discuss the empirical challenges in EAD measurements. Barakova and Parthasarathy (2012) finds that how conservative EAD prediction is in different models impacts final EAD estimates a lot. Across models, EAD estimation is higher when firm defaults are harder to predict, such as when the firm has good ratings, low line utilization, or when the credit cycle changes. Gürtler, Hibbeln, and Usselmann (2018) tries to estimate expected loss from empirically estimate CCF, LEQ, exposure at default factor (EADF), and EAD on a retail portfolio of a European bank and finds that CCF is the most preferable.3
In this study, we focus on the credit line usage of unlisted, middle-market firms. We gather credit line data from 19 U.S. financial institutions. All participates in Moody’s Credit Research Database (CRD), a consortium that collects credit risk data. These institutions range from USD$10 billion to USD$2 trillion in total asset size. The quarterly line information spans 2000 – 2017. The dataset covers revolving lines of credit; we exclude term loans from the dataset.4
Our dataset is unlike those used in other studies. First, it is large and derived from multiple institutions. As of the end of 2017, the unused commitment of the credit lines used in the final dataset totaled about USD$200 billion. Even though the dataset may represent a small portion of all the credit lines extended by financial institutions in the U.S.,5 its broad coverage across industries, size cuts and time makes it a database well-suited to studying credit line utilization behavior for middle-market borrowers. Second, it contains information on both defaulted and non-defaulted firms; most existing studies rely only upon defaulted names. Third, our dataset contains line information linked to borrower characteristics. Fourth, the dataset covers both economic downturns and expansion times, which allows us to better understand credit line usage behavior during cycles. Utilizing this unique dataset, we examine the extent to which borrowers draw down their credit lines and the characteristics of those firms with high usage. We study how line usage changes with banks’ internal ratings, collateral, loan purpose, commitment size and through economic cycles. The dataset is not dominated by a single financial institution. The largest share of any financial institution in the dataset is about 29% in terms of the number of facilities. The sample consists of 4.97 million quarterly observations on 53k defaulted loans and 321,000 non-defaulted loans from June 2000 through December 2017. Line records with a negative balance amount or those with a commitment amount less than USD$5,000 are deleted. We exclude lines that have never been utilized, because they could be inactive lines that are kept in the loan accounting system. Including inactive lines may introduce downward bias in usage estimates. Our approach of excluding them leads to a more conservative estimate of the usage measurements.
We merge the line usage data with the default and financial statement information housed in the CRD. In this study, we present the results on line level. We seek to include all defaults consistent with Basel definitions, if the information is available. Once all defaults are detected, we aggregate the data to create a single default for each borrower posting a default event of 90 days past due (90DPD) or more severe. We then assign the defaulter a date of the earliest default event and the most severe default type detected over time. Table 1 presents the default types, ordered by default severity from the least severe to the most severe. Note, only 90DPD with a “non-pass” rating is included as a default event. 90DPD with a pass rating is likely to be technical default that does not progress to economic loss. We therefore exclude that as a default event. We also exclude “substandard” as a default event unless the credit is also 90DPD.6 U.S. regulatory bodies exclude the “substandard” classification in their definition of “unlikely to pay.” 7
Figure 1 presents the distribution of the commitment size.8 Most of the sample firms are small-and medium-sized enterprises (SMEs).9 The median commitment amount is USD$100,000; the 75th percentile of commitment size is at USD$768,000. Approximately 10% of the sample firms have a commitment amount greater than USD$5 million.
4. Usage Measurements
For defaulters, we measure EAD using two variables: usage is the percentage of the exposure expected to be drawn down in the event of default; LEQ is the percentage of additional drawdown over the remaining commitment amount in the event of default:
usage = balance at default / commitmentYear-1
LEQ = (balance at default - balanceYear-1) / (commitmentYear-1 - balanceYear-1)
LEQ is commonly used to estimate EAD.10 The LEQ measurement is a useful concept in portfolio modeling. It allows us to model the undrawn portion of a credit line separately from the drawn portion. Common practice assumes 100% usage on the drawn portion of the line (between now and the default horizon) and LEQ for the undrawn portion, should the exposure default.
Nevertheless, the LEQ has limitations. For lines with a total commitment almost fully drawn down (small, unused commitment), LEQ may not be meaningful. For example, a line with an unused commitment of US$1 and an additional drawdown of US$3 will have an LEQ of 300%. In addition, LEQ can be negative. Negative LEQs may result from the decrease in balance amount or from an overdraw in year -1. We follow previous studies to exclude observations with extreme LEQs (less than zero or greater than 120%).11 Finally, the LEQ is not independent of current usage. A line almost fully drawn may have a lower LEQ than a line that has a large unused commitment, simply because there is little the borrower can draw down from the former. The usage measurement captures the total usage at default, rather than the additional drawdown. Both measurements are useful.
The usage measurements for non-defaulters are defined similarly. Usage refers to the percentage of the total commitment amount expected to be drawn. LEQ refers to percentage of additional drawdown over the undrawn amount:
usage = balance / commitmentYear-1
LEQ = (balance - balanceYear-1) / (commitmentYear-1 - balanceYear-1)
To avoid the impact of extreme values, we exclude observations less than zero or greater than 120% when reporting results on usage ratio, and observations less than zero or greater than 120% when reporting results on LEQ ratio. An alternative approach is to set the usage or LEQ ratio to zero when it is less than zero and set it to 120% when it exceeds 120%. Our treatment has little impact on usage ratio, as most lines have a usage ratio between 0% and 120%. The treatment leads to a more conservative estimate of LEQ than the alternative approach because there are more lines with negative LEQ.
Note, in this study, we focus on the one-year horizon. In other words, we measure the current drawn amount relative to the commitment or drawn amount one year prior. Discussions with bank practitioners reveal that usage measurements at the one-year horizon are most relevant to them.
First, we study usage statistics. Table 2 presents summary statistics. Panel (a) reports the usage and LEQ for defaulted firms, and Panel (b) reports the usage and LEQ for non-defaulted firms. When computing the usage and LEQ measurements on defaulted firms, we look for the usage information for the quarters falling into the (-90, +90) day window, with day 0 being the default date. We then select the quarter with the largest balance amount within this six-month window. This method builds some conservatism into the usage measurement.
5.2 Drivers Behind Usage
We next examine a set of variables that may further help explain the usage pattern of credit lines. These variables include a firm-level probability of default measure, risk grade assigned by lenders, collateral type, and total commitment size. We also look at how usage ratios change across economic cycles.
FIRMS WITH HIGHER DEFAULT RISK HAVE HIGHER USAGE
Firms with higher default risk are likely to draw down credit lines more. Such firms may have liquidity issues, and high leverage or bad financial performance may make it difficult for them to acquire additional funding other than drawing down their credit lines. Higher risk may also stem from aggressive expansion, which, in turn, implies higher credit line usage. We examine how usage relates to a firm’s default risk as measured by Moody’s Analytics RiskCalc™ EDF™ (Expected Default Frequency) credit measures.14 The EDF credit measure is a probability of default measurement based upon financial statement ratios as well as equity market information. The RiskCalc EDF considers financial ratios that measure the profitability, leverage, liquidity, debt coverage, growth, operating efficiency and size of a firm.15 Note, only 33% of the firms in the sample have EDF measures; the rest cannot be matched to a financial statement in order to calculate EDF measures.
In Table 5, we divide the sample firms into ten buckets, grouped by RiskCalc EDF credit measures, with an equal number of firms in each bucket. EDF measures are calculated with information available in the previous year’s financial statement. Group 1 has the lowest default risk as measured by the RiskCalc EDF measure, and Group 10 has the highest default risk. We compute the mean usage and LEQ for defaulted firms and non-defaulted firms separately for each bucket. We also report the EDF measure range. As expected, EDF measures for defaulters are, in general, higher than those of non-defaulters. Results show that, for defaulted firms, higher EDF measures are related to higher usage. The average usage for Group 1 is 36%, whereas the average usage for Group 10 is 75%. We also see an increase in usage and LEQ for non-defaulted firms as the EDF measure increases. The increase in the defaulters’ LEQ measurements with the EDF level is very slow. LEQ increases from 23% for Group 1 to 41% for Group 10, but LEQ for Group 9 and Group 10 are the same. One possible reason is that high-risk firms already have high usage prior to default.
FIRMS WITH NON-PASS GRADE HAVE HIGHER USAGE
A bank’s internal rating captures the bank’s own assessment of the borrower’s creditworthiness. Such ratings are used in credit approval, portfolio monitoring, provision and allowance calculation and capital allocation. A bank’s internal rating process is subject to regulatory oversight. Therefore, most banks have internal rating systems that parallel the credit risk rating scale used by bank regulators.17 Table 6 assigns firms to two groups by their internal rating one year prior: “pass” grade and “non-pass” grade. We observe that firms with “non-pass” ratings have higher usage than those with “pass” ratings. However, especially for defaulted companies, firms with “pass” ratings have higher LEQ than those with “non-pass” ratings.
Results suggest that banks use internal ratings to monitor credit line usage, especially for firms that eventually default on their loans. “non-pass” rating can indicate a higher risk than “pass” rating, so the higher usage for firms with “non-pass” ratings is in line with the increasing usage for higher EDF, presented in last session. However, firms with “non-pass” ratings can be on the radar of credit reviewers and thus have difficulties to draw down the lines more from the unused commitment. So LEQ for firms with “non-pass” rating is lower than those with “pass” rating, especially for defaulted companies, which are more likely to be monitored closely by banks.
In the regression analysis, we further examine whether having a “non-pass rating” is associated with higher usage ratios, after controlling for other variables. We discuss the multivariate regression results in the next section.
USAGE AND LEQ DIFFERENT FOR DIFFERENT COLLATERAL TYPES
Collateral posting is another variable that may affect credit line usage. If a firm has both secured and unsecured lines, it has incentives to draw down the unsecured line more. At the same time, banks may monitor line usage of unsecured lines more closely because the LGD would be higher. Alternatively, banks may grant unsecured lines to borrowers with lower risk or longer lending relationships. Table 7 shows that for non-defaulters, unsecured lines or second lien lines have the lowest usage and lowest LEQ. In the regression analysis, we further examine whether or not the relationship between collateral posting and usage measurements holds, after controlling for other variables.
We also test how the usage ratio changes with collateral type when there is a collateral posting (Table 7). Collateral types include agriculture, equipment and machinery, all assets, real estate, inventory and accounts receivables, and cash and securities. For defaulters, usage and LEQ are high for all collateral types. For non-defaulters, usage and LEQ are highest when the collateral types are real estate or agriculture.
The fourth variable we look at is the total commitment size. Large firms typically are able to secure higher commitment amounts from lenders. Therefore, commitment size can be a proxy for the size of the borrower. In Table 8, we divide the sample into five groups based on the size of the commitment. Larger lines are drawn down less, for both defaulted firms and non-defaulted firms. One explanation for this finding is that larger firms have more financing channels, while small firms usually rely on their banks for funding. Therefore, smaller firms draw down more from the credit lines. The other explanation is that banks may monitor lines with larger commitment size more closely.
PRO-CYCLICAL NATURE OF LINE USAGE
We further look at how line usage changes through economic cycles. Previous studies showed mixed results on usage cyclicality. Firms tend to draw down credit lines to help weather economic downturns. At the same time, banks typically tighten credit during recessions. Fewer growth opportunities, or firm downsizing, may also lead to lower credit line usage during recessions. Evidence on line usage cyclicality has implications for a bank’s liquidity and capital management. Particularly, in recent stress-testing exercises, the EAD estimation is one of the key components of expected loss and capital ratio calculation. A usage ratio or LEQ ratio is typically applied to the total or remaining commitments of credit lines to forecast EAD. Therefore, whether and how these two ratios change over economic cycles are central to bank stress-testing exercises.
Figure 4 plots the mean usage ratio by quarter for defaulted and non-defaulted firms separately. The mean usage ratios for both defaulted and non-defaulted firms are high around March 2001 and around June 2009, suggesting higher credit line utilization during economic downturns.
Note, for non-defaulters there seems to be a downward trend in usage ratio. As we have different financial institutions contributing data at different time periods, the trend in usage ratio may reflect differences across institutions. In the regression analysis detailed in the next section, we add a fixed bank effect control to account for such differences.
Figure 5 plots the mean LEQ ratio by quarter for defaulted and non-defaulted firms separately. The mean LEQ ratio is high around March 2001 and around June 2009, suggesting more credit line drawdown during economic downturns.
In the next section, we further examine the relationship between usage or LEQ ratio and an indicator of recession in a multivariate regression setting.
A closer look at the balance amount and commitment amount around the end of 2008 reveals that the increases in usage and LEQ ratios mostly come from additional drawdown rather than a cut in a commitment amount. The results suggest that banks were one of the liquidity providers for private firms at the peak of the recession.
5.3 Regression Analysis
The previous sections analyze each of the factors that may explain credit line usage patterns. In this section, we conduct regression analysis to understand how these factors are associated with the usage measurements in a multivariate scenario.
We pool together both defaulters and non-defaulters into one sample. The dependent variable is either usage or LEQ. The first independent variable is a dummy variable equal to 1 for a defaulted firm-quarter and equal to 0 otherwise. The second independent variable is the EDF measure, in deciles. We place firms into ten equally sized groups ranked by EDF measure. The first group contains the lowest EDF/lowest risk and the tenth group contains the highest EDF/highest risk. The third variable is a dummy variable equal to 1 if the line is unsecured and equal to 0 otherwise. The fourth variable measures whether or not the internal rating one-year prior is a pass grade. In the univariate analysis, we observe that firms with “pass” ratings one year prior have lower usage but higher LEQ than those with “non-pass” ratings. We add an interaction variable between the pass dummy and the default dummy to capture any interactive effects if any. The last variable is the commitment size measured on a scale of 1 to 5, with 1 referring to the group with smallest commitment size. We also add the usage one year prior and the squared term of the usage one year prior as independent variables. The squared term captures the potential non-linear relationship between usage one year prior and the dependent variable.
We add fixed-effect variables in the regression models. The fixed effect specification is intended to control for unobservable factors such as bank policies or industry practice that may have an impact on usage or LEQ. Such factors can be specific to a particular quarter, to the financial institutions that contribute the data or to the industry sectors to which the borrowers belong. The fixed effect model is often known as “within” regression. In other words, it measures the variation of the dependent variable within a group. We run two sets of regression models for usage and LEQ ratios, respectively. The first includes fixed bank, sector and quarter effect. The second omits the quarter effect and adds a dummy variable to indicate recession quarters and an interaction term between the recession dummy and the default dummy. The coefficient on the recession dummy measures whether the usage ratio or LEQ ratio becomes higher or lower during times of recession. The coefficient on the interaction term further measures the difference in the cyclical behavior of the usage or LEQ ratio between defaulted and non-defaulted firms.
Note, we did not add a fixed-firm effect into the regression models because our primary interest in this study is the cross-firm variation of line usage.
Finally, we adjust the standard error to allow for clustering at the borrower firm level, as firms may have similar credit line contracts rolling over multiple times. The standard errors and the p-values in Table 9 are reported after the correction.
- The coefficient on the default dummy is 6% for the usage regression and 14% for the LEQ regression. After controlling for other variables, the usage difference between defaulted firms and non-defaulted firms becomes smaller than that observed in Table 2 and Figure 3. It implies that such a difference can be explained by the other variables in the regression.
- Firm risk as measured by EDF measure has a significant impact on usage and LEQ. Firms in decile 10 have 10% more usage and 5% more LEQ than those in decile 1. It shows that firms with higher default risk tend to draw down more from credit lines. The coefficient on the pass grade dummy is positive for LEQ and negative for usage. In other words, pass grade firms have lower usage. And the coefficient on the interaction term between pass and default is not significant at 95% significance level.
- The coefficient on the collateral dummy is negative. All else being equal, firms with unsecured lines have lower usage and LEQ than those with secured lines.
- The coefficient on the commitment size is negative, consistent with the finding that lines with larger commitment get drawn down less.
- The coefficient on the usage in last year is positive and highly significant, showing that usage increases with usage one year prior within the specified interval [0%, 120%]. Firms with high usage levels tend to maintain the high level. LEQ increases (decreases) with usage one year prior when the latter is smaller (larger) than 66% (0.98/(0.74*2)). The additional drawdown as measured by LEQ tends to be larger where usage is already high, except for those firms that have used the majority of their credit line capacity.
6. Conclusion
This study is the first to provide large sample evidence on the credit usage for U.S. middle-market borrowers. Previous studies generally focused on a small sample from one institution, collected data on publicly listed firms or utilized non-U.S. datasets. Our findings confirm the high EAD documented in previous studies. We show that usage varies with firm-level and line-level characteristics and economic cycles.
We find that for defaulted firms EAD is indeed high. Usage reaches a high level approximately two years before default. Non-defaulted firms, on average, have lower usage levels. For both defaulters and non-defaulters, usage is higher if default risk is higher. This finding raises the question of whether or not banks charge sufficient spreads to cover the higher usage by riskier borrowers: a subject that warrants further study.
Usage measurements are also related to commitment size, collateral type and banks’ internal ratings. Our findings suggest that banks monitor lines with poorer internal ratings, larger commitment size, or without collateral more closely. We observe that usage ratios increased during the most recent recession.
1 Data are from Federal Deposit Insurance Corporation (FDIC).
2 Institutions can calculate the EAD of the unfunded part of the loan using either loan equivalent (LEQ) or CCF. Basel uses term LEQ and CCF interchangeably. OCC defines CCF as the balance at default to balance 12 months prior to default. Empirically and conceptually the two measures should lead to the same EAD. In this paper, we adopt the Basel definition of CCF and use the term LEQ, which is directly calculated from credit line information.
3 CCF follows the OCC definition in this paper.
4 We work with the banks to separate revolving credit lines from term loans. The identifiers provided by banks include “revolver” “line of credit,” and “line.”
5 The closest measurement of aggregate unused commitment amount on the national level is the “other unused commitments” reported by FDIC Statistics on Depositary Institutions (see https://www5.fdic.gov/sdi/main.asp?formname=compare/). As of the end of 2017, the total amount under “other unused commitments” is USD$2.9 trillion, which includes commitments to extend credit through overdraft facilities or commercial lines of credit and retail check, credit, and related plans. Our data set represents about 7% (USD$200 billion divided by USD$2.9 trillion) of all the credit lines outstanding in the U.S. Nevertheless, our data set includes overdraft facilities and commercial lines of credit, but does not include retail check, credit and related plans. Therefore, the actual estimate should be higher than 7%.
6 A “substandard” asset is defined as “inadequately protected by the current sound worthiness and paying capacity of the obligor or by the collateral pledged, if any. Assets so classified must have a well-defined weakness or weaknesses that jeopardize the liquidation of the debt. They are characterized by the distinct possibility that the institution will sustain some loss if the deficiencies are not corrected.” This definition contrasts with an asset classified as “doubtful” that “has all the weaknesses inherent in one classified substandard with the added characteristic that the weaknesses make collection or liquidation in full, on the basis of currently known facts, conditions, and values, highly questionable, and improbable.” Please see FDIC (2012) for more details.
7 “Risk-Based Capital Standards: Advanced Capital Adequacy Framework,” Basel II; Final Rule.
8 We present the aggregate amount at the borrower level if the borrower has more than one line from a bank at any quarter.
9 Discussions with CRD participant banks reveal that the definition of SMEs may vary with bank size. Basel II characterizes an SME as a corporate with less than €50 million in turnover (see Annex 3 of “International Convergence of Capital Measurement and Capital Standards: A Revised Framework,” June 2004).
10 EAD = (commitmentyear-1 -balanceyear-1) x LEQ + balanceyear-1.
11 We set the upper limit to 120% to allow for overdraw of the lines. Setting the upper limit to 110% or 130% yields similar results.
12 Relax the current constraints on LEQ and allow LEQ to vary between -20% to 120%, the mean, 25th percentile, median, and 75th percentile LEQ for defaulted firms at default quarter are 37%, 0%, 18%, and 86%; for non-defaulted firms are 22%, 0%, 1%, and 43%. Median LEQ drops substantially when allowing negative LEQ. Theoretically, LEQ can be negative. We set lower limit of 0% to be on the more conservative side.
13 Other category includes loans with unknown purpose.
14 RiskCalc is Moody’s Analytics private firm probability of default model.
15 For more information on the RiskCalc model, please refer to Korablev, et al. (2012).
16 Larger companies, when compared to smaller companies, tend to have valid financial information to calculate EDF measures, and they have relatively lower usage.
17 See FDIC (2012) for the definitions of the regulatory ratings.
Acharya, V., Almeida, H., and Campello, M., “Aggregate Risk and the Choice Between Cash and Lines of Credit.” Journal of Finance, 68(5), 2059–2116, 2013.
Acharya, V., Almeida, H., Ippolito, F., and Perez-Orive, A., “Credit lines as monitored liquidity insurance: theory and evidence.” Journal of Financial Economics, 2014.
Araten, M. and Jacobs M., “Loan Equivalents for Revolving Credits and Advised Lines.” The RMA Journal, 34-39, 2001.
Asarnow, E. and Marker, J., “Historical Performance of the U.S. Corporate Loan Market: 1988-1993.” The Journal of Commercial Lending, Vol. 10 No. 2, 13-32, 1995.
Bacham, D. and Yang, L., “RiskCalcTM Small Business.” Moody’s Analytics, White Paper, 2019.
Bag, P. and Jacobs, M., “Parsimonious exposure‐at‐default modeling for unfunded loan commitments.” Journal of Risk Finance, 2012.
Barakova, I. and Parthasarathy, H., “Modeling Corporate Exposure at Default.” SSRN Working Paper 2235218, Presented Quant Forum 2012, WEFAI/IBEFA, 2013.
Campello, M., Giambona, E., Graham, J., and Campbell, H., “Liquidity Management and Corporate Investment During a Financial Crisis.” Review of Financial Studies, 24(6), 1944–1979, 2010.
Demiroglu, C. and James, C., “The use of bank lines of credit in corporate liquidity management: a review of empirical evidence.” Journal of Banking and Finance. vol. 35, issue 4, 775-782, 2011.
Dwyer, D. and Eggleton, D., “Level and Rank Order Validation of RiskCalc V3.1 United States.” Moody’s Analytics, White Paper, 2009.
Emery, K., Gates, D., Marshella, T., and Ou, S., “Migration of Debt Structures and Revolver Usage as Firms Approach Default.” Moody’s Investor Service, Special Comment, 2008.
FDIC, “How Do Examiners Assign Loan Classification on Your Examination?” Federal Deposit Insurance Corporation, Presentation, http://www.fdic.gov/regulations/resources/ director/college/ny/materials/2012-loans.pdf, 2012.
Gao, P. and Yun, H., “Commercial Paper, Lines of Credit, and the Real Effects of the Financial Crisis of 2008: Firm-Level Evidence from the Manufacturing Industry.” Banque du Canada, Working Paper, http://www.banqueducanada.ca/wp-content/uploads/2010/09/gao.pdf, 2009.
Gürtler, M., Hibbeln, M., and Usselmann P., “Exposure at default modeling – A theoretical and empirical assessment of estimation approaches and parameter choice.” Journal of Banking and Finance, 2018.
Hau, A., “Pricing of Loan Commitments for Facilitating Stochastic Liquidity Needs.” Journal of Financial Services Research, 2011.
Huang, R., “How Committed Are Bank Lines of Credit? Evidence from the Subprime Mortgage Crisis?” Working Paper, http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1659986, 2009.
Ivashina, V. and Scharfstein, D., “Bank Lending During the Financial Crisis of 2008.” Journal of Financial Economics, 97(3), 319–338, 2009.
Jacobs, M., “An Empirical Study of Exposure at Default.” Journal of Advanced Studies in Finance, Vol. I, Issue 1(1), pp. 31-59, 2010.
Jiménez, G., Lopez, J., and Jesus Saurina, J., “Empirical Analysis of Corporate Credit Lines.” Review of Financial Studies, 22(12), 5069–5098, 2009a.
Jiménez, Gabriel, Lopez, J. and Saurina, J., “EAD Calibration for Corporate Credit Lines.” Journal of Risk Management in Financial Institutions, 2(2), 121–129, 2009b.
Korablev, I., Uliana M., Wang, X., Zhang, A., Zhao, J., and Douglas D., “RiskCalc U.S. 4.0.” Moody’s Analytics, White Paper, 2012.
Lins, K., Servaes, H. and Tufano, P., “What Drives Corporate Liquidity? An International Survey of Cash holdings and Lines of Credit.” Journal of Financial Economics 98(1), 160–176, 2008.
Moral, G., (2006). “EAD Estimates for Facilities with Explicit Limits.” The Basel II Risk Parameters, Engelmann, B., and Rauhmeier, R. (eds), pp. 197–242, Springer.
Pan, K., “The Loan Equivalency Factor.” The RMA Journal, 2009.
Qi, M., “Exposure at Default of Unsecured Credit Cards.” OCC Economics Working Paper 2009-2, 2009.
Sufi, A., “Bank Lines of Credit in Corporate Finance: An Empirical Analysis.” The Review of Financial Studies, Vol. 22, Issue 3, pp. 1057-1088, 2009.
Taplin, R., To, H., and Hee, J., “Modeling Exposure at Default, Credit Conversion Factors and the Basel II Accord.” Journal of Credit Risk, Summer, 75-84, 2007.