Understanding and Predicting LGD on Bonds & Loans



A Theoretical Model for Ultimate Loss-Given-Default - Understanding Liquidity and the Returns on Defaulted Debt

Michael Jacobs, Jr.[1]

Office of the Comptroller of the Currency

Draft: October 2007

J.E.L. Classification Codes: G33, G34, C25, C15, C52.

Keywords: Recoveries, Default, Loss Given Default, Financial Distress, Bankruptcy, Restructuring, Credit Risk, Entropic Methods, Bootstrap Methods, Forecasting

Abstract

In this study we empirically investigate the determinants of, and build alternative predictive econometric models for, ultimate loss-given-default (LGD) on bonds and loans. We utilize an extensive sample of S&P and Moody’s rated defaulted firms in the period 1985-2004, 650 bankruptcies and out-of-court settlements of distress that are largely representative of the US large corporate loss experience, for which we have the complete capital structures and can track the recoveries on all instruments to the time of resolution. We extend prior work by modeling LGD at the firm and instrument level, both separately and jointly, and show that a simultaneous equation approach incorporating feedback between these levels, and theoretically motivated cross-equation parameter restrictions, has certain desirable properties relative to univariate approaches. We consider most of the previously considered determinants of LGD, confirm many of the stylized facts and findings of the literature, and reveal some new ones. In regard to contractual, capital structure, industry and macroeconomic determinants, our results are broadly consistent with the existing literature in Keisman (2000), Carey and Gordy (2004, 2005) and Acharya (2003, 2007). In a departure from the extant literature, we find the economic and statistical significance of firm specific (financial ratios), debt (price of traded debt at default) and equity market (cumulative abnormal returns on equity prior to default - CARs) variables. Alternative econometric models are built in various classes spanning fully to non-parametric (beta link generalized linear model-BLGLM, Kullback-Leibler relative entropy-KLRE and generalized beta kernel density models-BDKE). Across various univariate specifications (with and without LGD at default, CARs or financial ratios), the BLGLM model is found to perform better than a linear regression model with a transformed dependent variable, in terms of the quality of estimated parameters as well as overall model performance metrics. The models are validated rigorously through a resampling experiment in a rolling out-of-time and out-of-sample framework, in which the distributions of discriminatory (areas under ROC curves and Kolmogorov-Smirnov statistics) and predictive accuracy (McFadden Pseudo-Rsquareds and Hoshmer-Lemeshow) statistics are compared. While the BLGLM model is found to consistently outperform the KLRE and BKDE models out-of-sample, results regarding the relative superiority of the bivariate versus univariate estimations is inconclusive.

1. Introduction and Summary

Loss given default (LGD) [2], the loss severity on defaulted obligations, is a critical component of risk management, pricing and portfolio models of credit. This is among the three primary determinants of credit risk, the other two being the probability of default (PD) and exposure of default (EAD). However, LGD has not been as extensively studied, and is considered a much more daunting modeling challenge in comparison to, PD. Starting with the seminal work by Altman (1968), and after many years of actuarial tabulation by rating agencies, predictive modeling of default rates is currently in a mature stage. The focus on PD is understandable, as traditionally credit models have focused on systematic components of credit risk which attract risk premia, and unlike PD, determinants of LGD have been ascribed to idiosyncratic borrower specific factors. However, now there is an ongoing debate about whether the risk premium on defaulted debt should reflect systematic risk, in particular whether the intuition that LGDs should rise in worse states of the world is correct, and how this could be refuted empirically given limited and noisy data (Carey and Gordy, 2003).

The recent heightened focus on LGD is evidenced the recent flurry of research into the relatively neglected area of LGD (Acharya et al (2005), Carey and Gordy [2003, 2004, 2005], Altman et al [2001, 2003, 2004, 2006], Gupton et al [2000, 2001], Araten et al [2003], Frye [2000 a,b,c], Jarrow [2001]). This has been motivated by the large number of defaults and near simultaneous decline in recovery values observed at the trough of the last credit cycle circa 2000-2002, regulatory developments such as Basel II (BIS 2003, 2004, 3005) and the growth in credit markets. However, obstacles to better understanding and predicting LGD, including dearth of data and the lack of a coherent theoretical underpinning, have continued to challenge researchers. In this paper, we hope to contribute to this effort by synthesizing advances in financial theory, econometric methodology and data acquisition in order to build an empirical model of LGD that is consistent with a priori expectations and stylized facts, internally consistent and amenable to rigorous validation. In addition to answering the many questions that academics have, we further aim to provide a practical tool for risk managers, traders and regulators in the field of credit.

LGD may be defined variously depending upon the institutional setting or modeling context, or the type of instrument (traded bonds vs. bank loans) versus the credit risk model (pricing debt instruments subject to the risk of default vs. expected losses or credit risk capital). In the case of bonds, one may look at the price of traded debt at either the initial credit event[3] , the market values of instruments received at the resolution of distress[4] (Keisman et al, 2000; Altman et al, 1996) or the actual cash-flows incurred during a workout[5]. When looking at loans that may not be traded, the eventual loss per dollar of outstanding balance at default is relevant (Asarnow et al, 1995; Araten et al, 2003). There are two ways to measure the latter – the accounting LGD refers to nominal loss per dollar outstanding at default[6], while the economic LGD refers to the discounted cash flows to the time of default taking into consideration when cash was received.[7] The former is used in setting reserves or a loan loss allowance, while the latter is an input into a credit capital attribution and allocation model.

An aspect of LGD modeling deserving of special attention, until recently neglected altogether or grossly simplified, is the distributional characterization of this random variable (or stochastic process). While the available theory and evidence suggests it to be stochastic and predictable with respect to other variables, LGD has been treated as either deterministic or as an exogenous stochastic process in most extant credit models.[8] Such assumptions are made for tractability and in practical application results in understated capital, mispricing and unrealistic dynamics of model outputs. We will contribute to resolving such deficiencies by attempting to model the ex ante distribution of LGD as a function of empirical determinants – contractual features, firm capital structure, borrower characteristics, debt / equity market and systematic factors. However, we will not directly model the interdependency between LGD and other parameters of interest, such as probabilities of default, by either estimating a structural or reduced form model in which they are determined simultaneously[9].

In this study we empirically investigate the determinants of, and build predictive econometric models for, ultimate LGD. We utilize an extensive sample of 650 S&P and Moody’s rated defaulted firms[10] (1985-2004), for which we have the complete capital structures, which is highly representative of the US large corporate loss experience over the last two decades. We contribute to the literature along several dimensions. First, we extend prior work by modeling LGD at the firm and instrument level, both separately and jointly. In particular, a simultaneous equation estimation of LGD at the instrument and obligor levels is shown to have some desirable properties relative to univariate approaches, both in terms of quality of parameter estimates and model performance diagnostics. Second, as compared to prior work we consider both new variables as well as previously considered variables, in a unified framework. The latter include contractual (e.g., collateral rank, debt cushion), capital structure (e.g., proportion of secured and bank debt), firm specific (e.g., leverage and cash flow), industry (e.g., industry profitability), macroeconomic (e.g., aggregate default rate and the equity market return) and debt/equity market determinants (e.g., price of debt at default, cumulative abnormal returns on equity). We confirm many of the stylized facts and findings of the literature in regard to the determinants of LGD, and find in addition the independent significance of a macroeconomic factor, equity returns and the price of traded debt at default in explaining the LGD. Alternative econometric models are built in various classes spanning fully-parametric (beta link generalized linear model), semi-parametric (Kullback-Leibler relative entropy) to non-parametric methods (generalized beta kernel density frameworks). The model is validated rigorously through resampling experiment in a rolling out-of-time and out-of-sample framework, in which we compare both alternative classes of models on a univariate basis, as well as univariate vs. bivariate estimates. We compare distributions of discriminatory (areas under ROC curves and Kolmogorov-Smirnov statistics) and predictive accuracy (McFadden Pseudo-Rsquareds and Hoshmer-Lemeshow) statistics.

This paper is organized as follows. Section 2 reviews the literature, including both the treatment of LGD in credit models (both theoretical and practical), as well as direct empirical evidence. Section 3 presents the econometric methodologies employed herein. Section 4 discusses model performance metrics that we use to evaluate the models, encompassing both discriminatory (or classification) accuracy and predictive (or calibration) accuracy measures. Section 5 describes the data used in this study and discusses various summary statistics. Section 6 presents the estimation results, at both the instrument and obligor levels, and univariate vs. Bivariate models, for the econometric approaches introduced in Section 3. Section 7 discusses the results of the model validation exercise, out-of-time and out-of sample bootstrapping exercises for the model performance statistics introduced in Section 4. Finally, Section 8 concludes and discusses directions for future research.

2. Review of the Literature

In this section we will examine the way in which different types of theoretical credit risk models have treated LGD – assumptions, implications for estimation and application. We will then turn to the empirical evidence, both on the estimation of LGD, and on the performance of these various models. Finally, we will look at some of the stat-of-the-art and vendor models of LGD, and how they have attempted to incorporate lessons learned.

2.1 Treatment of LGD in Theoretical Credit Risk Models

Credit risk modeling was revolutionized by the approach of Merton (1974), who built a predictive theoretical model in the option pricing paradigm of Black and Scholes (1973), which has come known to be the structural approach. Equity is modeled as a call option on the value of the firm, with the face value of zero coupon debt serving as the strike price, which is equivalent to shareholders buying a put option on the firm from creditors with this strike price. Given this capital structure, log-normal dynamics of the firm value and the absence of arbitrage, closed form solutions for the default probability and the spread on debt subject to default risk can be derived. The LGD can be shown to depend upon the parameters of the firm value process as is the PD, and moreover is directly related to the latter, in that the expected residual value to claimants is increasing (decreasing) in firm value (asset volatility or the level of indebtedness). Therefore, LGD is not independently modeled in this framework; this was addressed in much more recent versions of the structural framework (Frye [2000], Dev et al [2002[, Pykhtin [2003]), to be discussed in more detail later.

Extensions of Merton (1974) relaxed many of the simplifying assumptions. Complexity to the capital structure was added by Black and Cox (1976) and Geske (1977), with subordinated and interest paying debt, respectively. The distinction between long- and short-term liabilities in Vasicek (1984) was the precursor to the KMV( model. However, these models had limited practical applicability, the standard example being evidence of Jones, Mason and Rosenfeld (1984) that these models were unable to price investment grade debt any better than a naïve model with no default risk. Further, empirical evidence in Franks and Touros (1989) showed that the adherence to absolute priority rules (APR) assumed by these models are often violated in practice, which implies that the mechanical negative relationship between expected asset value and LGD may not hold. Longstaff & Schwartz (1995) incorporate into this framework a stochastic term structure with a PD-interest rate correlation. Other extensions include Kim at al (1993) and Hull & White (1995), who examine the effect of coupons and the influence of options markets, respectively.

Partly in response to this, a series of extensions ensued, the so-called “second generation” of structural form credit risk models (Altman [2003, 2006]). The distinguishing characteristic of this class of models is the relaxation of the assumption that default can only occur at the maturity of debt – now default occurs at any point between debt issuance and maturity when the firm value process hits a threshold level. The implication is that LGD is exogenous relative to the asset value process, defined by a fixed (or exogenous stochastic) fraction of outstanding debt value. This approach can be traced to the barrier option framework as applied to risky debt of Black and Cox (1976).

All structural models suffer from several common deficiencies. First, reliance upon an unobservable asset value process makes calibration to market prices problematic, inviting model risk. Second, the limitation of assuming a continuous diffusion for the state process implies that the time of default is perfectly predictable (Duffie and Lando [2000]). Finally, the inability to model spread or downgrade risk distorts the measurement of credit risk. This gave rise to the reduced form approach to credit risk modeling (Duffie and Singleton, 1999), which instead of conditioning on the dynamics of the firm, posit exogenous stochastic processes for PD and LGD. These models include (to name a few) Litterman & Iben (1991), Madan & Unal (1995), Jarrow & Turnbull (1995), Jarrow et al (1997), Lando (1998) and Duffie (1998). The primitives determining the price of credit risk are the term structure of interest rates (or short rate), and a default intensity and an LGD process. The latter may be correlated with PD, but it is exogenously specified, with the link of either of these to the asset value (or latent state process) not formally specified. However, the available empirical evidence (Lando and Turnbull [1997], Duffie and Singleton [1999]) has revealed these models deficient in generating realistic term structures of credit spreads for investment and speculative grade bonds simultaneously. A hybrid reduced – structural form approach of Zhou (2001), which models firm value as a jump diffusion process, has had more empirical success, especially in generating a realistic negative relationship between LGD and PD (Altman et al 2003, 2006).

The fundamental difference of reduced with structural form models is the unpredictability of defaults: PD is non-zero over any finite time interval, and the default intensity is typically a jump process (e.g., Poisson), so that default cannot be foretold given information available the instant prior. However, these models can differ in how LGD is treated. The recovery of treasury assumption of Jarrow & Turnbull (1995) assumes that an exogenous fraction of an otherwise equivalent default-free is recovered at default. Duffie and Singleton (1999) introduce the recovery of market value assumption, which replaces the default-free bond by a defaultable bond of identical characteristics to the bond that defaulted, so that LGD is a stochastically varying fraction of market value of such bond the instant before default. This model yields closed form expressions for defaultable bond prices and can accommodate the correlation between PD and LGD; in particular, these stochastic parameters can be made to depend on common systematic or firm specific factors. Finally, the recovery of face value assumption (Duffie [1998], Jarrow et al [1997]) assumes that LGD is a fixed (or seniority specific) fraction of par, which allows the use of rating agency estimates of LGD and transition matrices to price risky bonds.

It is worth mentioning the treatment of LGD in credit models that attempt to quantify unexpected losses analogously to the Value-at-Risk (VaR) market risk models, so-called credit VaR models (Creditmetrics™ [Gupton et al, 1997], KMV CreditPortfolioManager™ [KMV Corporation, 1984], CreditRisk+™ [Credit Suisse Financial Products, 1997], CreditPortfolioView™ [Wilson, 1998]). These models are widely employed by financial institutions to determined expected credit losses as well as economic capital (or unexpected losses) on credit portfolios. The main output of these models is a probability distribution function for future credit losses over some given horizon, typically generated by simulation of analytical approximations, as it is modeled as highly non-normal (asymmetrical and fat-tailed). Characteristics of the credit portfolio serving as inputs are LGDs, PDs, EADs, default correlations and rating transition probabilities. Such models can incorporate credit migrations (mark-to-market mode - MTM), or consider the binary default vs. survival scenario (default mode - DM), the principle difference being that in addition an estimated transition matrix needs to be supplied in the former case. Similarly to the reduced form models of single name default, LGD is exogenous, but potentially stochastic. While the marketed vendor models may treat LGD as stochastic (e.g., a draw from a beta distribution that is parameterized by expected moments of LGD), there are some more elaborate proprietary models that can allow LGD to be correlated with PD.

We conclude our discussion of theoretical credit risk models and the treatment of LGD by considering recent approaches, which are capable of capturing more realistic dynamics, sometimes called “hybrid models”. These include Frye (2000a, 2000b), Jarrow (2001), Jokivuolle et al (2003), Carey & Gordy (2003), Pykhtin (2003) and Bakshi et al (2001). Such models are motivated by the conditional approach to credit risk modeling, credited to Finger (1999) and Gordy (2000), in which a single systematic factor derives defaults. In this more general setting, they share in common the feature that dependence upon a set of systematic factors can induce an endogenous correlation between PD & LGD. In the model of Frye (2000a, 2000b), the mechanism that induces this dependence is the influence of systematic factors upon the value of loan collateral, leading to a lower recoveries (and higher loss severity) in periods where default rates rise (since asset values of obligors also depend upon the same factors). In a reduced form setting, Jarrow (2001) introduced a model of co-dependent LGD and PD implicit in debt and equity prices.[11]

2.2 Empirical Evidence on LGD and Tests of Credit Models

In this section we focus on the application of credit models and estimation of LGD from data. This ranges from simple quantification of LGD, calibration of credit models embedding LGD assumptions, and finally to empirical or vendor models of LGD.

There is a long tradition of actuarially estimating loss severities from bond or loan data, independent of any credit modeling framework and allied parameters such as PD. These have been conducted by academics, banks and rating agencies. The earliest studies relied exclusively on secondary market prices of bonds or loans. Altman & Kishore (1996) estimate LGDs for 300 defaulted senior secured and senior unsecured bonds from 1978-1995, yielding estimates ranging from 10% to 70% that could be statistically distinguished among various industry groups. Later studies looked at ultimate recoveries on defaulted loans or bonds, either the nominal or discounted price at emergence from bankruptcy. Altman and Eberhart (1994) and Fridson et al (Merrill Lynch 2000) provided evidence that more senior significantly outperformed more junior bonds in the post-default period. Altman and Kishore (1996) find statistically different LGDs across broad industrial sectors.

LGD in bank loans have been studied by banks, rating agencies and academics. In the ZETA™ model of Altman, Haldeman and Narayanan (1977), a 2nd generation of the Altman (1967) Z-Score PD estimation model, loan LGD estimates were based on a workout department survey (1971-75), in which the conclusion was that discounted post-restructuring recoveries on unsecured bank loans equal 30% plus accrued interest. Bank studies focusing on internal loan data include Citigroup (Asarnow & Edwards, 1995), Chase Manhattan Bank (1996) and JP Morgan Chase (Araten et al, 2003). Average LGDs of 35% (861 large corporate obligors 1979-1993), 36% (412 large corporate obligors 1986-1993) and 40% (3800 wholesale loans 1982-2000) for were found in the Citigroup, Chase and JPMC studies, respectively. Finally, among those conducting historical analysis, some recent rating agency studies are worthy of note. Moody’s (Hamilton et al, 2001) reports an implied mean LGD of 30.3% (47.9%) for 121(181) senior secured (senior unsecured) secondary market loan prices a month after default. Various consortia of banks have published composite loss severity statistics from member banks, including Loan Pricing Corporation[12] (2001; LPC) and the Risk Management Association (2000; RMA)[13]. While these have the advantage of containing some middle market firms, the degree of segmentation by borrower and instrument characteristics is limited in these, and there are issues associated with normalizing data across banks. Standard and Poors (Keisman et al, 2000) presents empirical results from the LossStatsTM database regarding ultimate LGDs. Analysis of 264 (690) bank loans (senior unsecured bonds) from 1987-1996 yields an average LGD of 16% (34%). This study also documents the independent influence of position in the capital structure (i.e., the proportion of debt above or below a claimant in bankruptcy), apart from collateral and seniority, in determining loss severities. Emery (2003) and Altman and Fanjul (2004) compare LGDs, as inferred from the prices of the traded instruments at default in a Moody’s database, on bank loans and bonds, respectively. A comparison of results reveals that loans experienced lower loss severity when controlling for seniority.[14] In a Moody’s report, Cantor et al (2003) document similar findings for corporate bonds as Altman and Fanjul (2004). Additionally, Altman and Fanjul (2004) document a differential LGD by rating at origination, such that “fallen angels” of the same seniority have significantly lower LGDs.[15] Finally, a recent rating agency study by Moody’s (Cantor and Varma, 2004) examines the determinants of ultimate LGD for North American corporate issuers over a period of 21 years (1983-2003), looking at many of the variables considered here ( seniority, security, initial default event, firm-specific, industry-specific, and macroeconomic factors).

Several recent empirical studies of LGD by academics have put more structure around this exercise, either by building predictive econometric models, or by attempting to directly test models, in either case attempting to capture the LGD-PD correlation. Frye (2000b) examines the LGD-PD correlation in 1982-1997 using the Moody’s Default Risk Service [DRSTM], finding a significant negative relationship at various levels of aggregation, consistent with the recent market experience in the 2000-2002 turn in the credit cycle. Jarrow (2001) develops a hybrid structural-reduced model, in which PDs and LGDs are functions of the underlying state of the macroeconomy, which allows the identification of recovery and default intensities from traded bonds using both observed debt and equity prices; further, his approach can accommodate risk premium into the estimation, critical for the high yield bond market. Bakshi et al (2001) extend the reduced form approach by allowing a flexible correlation between PD, LGD and the risk-free rate. Imposing a negative correlation between PD and LGD, in a calibration exercise using BBB- corporate bonds, they find that a 4% increase in the (risk-neutral) hazard rate is associated with a 1% increase in the expected LGD. Hu and Perraudin (2002) also examine this relationship with Moody’s DRSTM for the 1983-2000 period, standardizing the data by transforming all borrowers to senior unsecured, thereby isolating the influence of systematic factors. They find LGD-PD correlations on the order of 0.2. Jokivuolle and Peura (2003) present an option theoretic model for bank loans in which a firm’s asset value process is correlated with collateral, the latter being the only stochastic factor determining recovery (so that an exogenous PD can be assumed and the firm’s asset value process need not be estimated), and they are able to produce a positive correlation between PD and LGD. However, note that these studies look at LGD as implied from the prices of traded debt at or prior default, as opposed to ultimate LGD.

Several studies empirically study ultimate LGD. Carey and Gordy (2003) analyze the correlation of LGD and PD in a combined database of defaults in 1970-1999[16]. Examining aggregated quarterly data at the obligor level, while they find almost negligible correlation with PD over the entire sample, in 1988-1998 there is a significant relationship that is in line with Frye’s (2000b) results[17]. However, they document that LGDs tend to rise more in recessionary periods than they fall during expansions, suggesting that more is at play than a macroeconomic factor influencing the value of collateral.[18] In the Araten et al (2003) bank study, unsecured U.S. large corporate borrower level LGDs are regressed average Moody’s All-Corporate default rate for the period 1984-1999 on an annual basis, yielding an r-squared of 0.2, in line with Carey and Gordy (2004) for their restricted sample, yet significantly higher than Hu and Peraudin (2002). Altman et al (2001, 2005) find an that LGDs increase as the credit cycle worsens, going from 75% in 2001-2002 to 55% in 2003 as default rates decreased beneath their long run average of 4.5%.[19] This is verified by Keisman (2003), who finds that LGDs of all seniorities rise during this stress period, in the S&P LossStatsTM database for the period 1982-1999. However Altman (2005) finds that a systematic variable has no effect on LGD when bond market conditions (e.g., supply-demand imbalances) are accounted for. However, Acharya et al (2003, 2007) examine the same data and period as Keisman (2003), and while they verify that seniority and security are key determinants of LGD, in addition they find industry specific factors influencing LGD independently of the macroeconomic state and bond market conditions seen in Altman (2005). In particular, they find elevated LGDs in distressed industries (fewer redeployable assets, greater leverage and lower liquidity), after controlling for firm specific, contractual and systematic factors. This constitutes a test of the Schleifer and Vishny (1992) “fire-sale” hypothesis – an industry equilibrium phenomenon in which macro and bond market variables are spuriously significant due to omitting an industry factor.

We make note of the influence of this evidence regarding the PD-LGD correlation on the Basel II guidelines, the BIS Accord (2004) paragraph 468 on downturn LGD and the additional guidance offered in BIS (2005). This requires advanced IRB banks to not only capture all relevant risks regarding possible cyclical variability in LGD, but at the same time states that bank estimates of long-term ultimate LGDs having no such systematic variations may be acceptable. There was somewhat of a negative reaction to the inherent conservativeness in BIS (2004), partly motivating the clarification in BIS (2005); this was partly strengthened by some of the evidence that the PD-LGD correlation may not be economically significant (Araten et al 2003). The most recent thinking on this topic is that banks should have the latitude to assess the add-on to there average LGDs, accompanied by a rigorous and well documented process for assessing downturn effects by each asset class. Miu and Ozdemir (2006) argue that banks can incorporate conservatism into cyclical LGDs estimated in a point-in-time framework, without an LGD-PD correlation; however, they estimated commensurate in creases in credit capital to compensate for this.

Finally, we may mention vendor models of LGD, which incorporate approaches taken in the academic and agency literature, in addition applying proprietary methodologies and data sources. S&P (Friedman & Sandow, 2003) applies a Kulback-Leibler maximum entropy non-linear regression model to the LossStatsTM database, which incorporates Bayesian style prior information (point masses at 0 and 100%) to produce predictive densities of LGD. Moody’s LoosCalc2™ (Gupton, 2004) applies an econometric based models (local regression) of a Gaussian transformed LGD variable to KMV’s proprietary LGD database

3. Theoretical Model

The model that we propose is an extension of Black and Cox (1976). The baseline mode features perpetual corporate debt, a continuous and a positive foreclosure boundary. The former 2 assumptions remove the time dependence of the value of debt, thereby simplifying the solution and comparative statics. The latter assumption allows us to study the endogenous determination of the foreclosure boundary by the bank, as in Carey and Gordy (2007). We extend the latter model by allowing the coupon on the loan to follow a stochastic process, accounting for the effect of illiquidity. Note that in this framework, we assume no restriction on asset sales, so that we do not consider strategic bankruptcy, as in Leland (1994) and Leland and Toft (1996).

Let us assume a firm financed by equity and debt, normalized such that the total value of perpetual debt is 1, divided such that there is a single loan with face value [pic]and a single class of bonds with a face value of [pic]. The loan is senior to that bond, and potentially has covenants which permit foreclosure. The loan is entitled to a continuous coupon at a rate c, which in the baseline model we take as a constant, but may evolve randomly. Equity receives a continuous dividend, having a constant and a variable component, which we denote by [pic], where [pic] is the value of the firm’s assets at time t. We impose the restriction that [pic], where r is the constant risk-free rate. The asset value of the firm, net of coupons and dividends, follows a Geometric Brownian Motion with constant volatility [pic]:

[pic] (3.1)

Where we denote the fixed cash outflows per unit time as:

[pic] (3.2)

Default occurs at time t and is resolved after a fixed interval [pic]. AT this point dividend payments cease, but the loan coupon continues to accrue through the settlement period. At the point of emergence, holders of the loan receive [pic], or the minimum of the legal claim or the value of the firm at emergence. We can value the firm at resolution using the standard Merton (1974) formula:

[pic] (3.3)

Where [pic] is the standard normal distribution function and the arguments are:

[pic] (3.4)

Denote the total legal claim at default by

[pic] (3.5)

This follows from the assumption that the coupon c on the loan with face value [pic] continues to accrue at the contractual rate throughout the resolution period[pic], whereas the bond with face value [pic] does not. In order to value recovery on the firm, we modify the Merton equations in (3.2) and (3.3) to the following:

[pic] (3.6)

Where we replace [pic] with D in the arguments to the Gaussian distribution functions in (3.4):

[pic] (3.5)

It follows that the economic loss-given-default (LGD) to all creditor classes (or the estate level loss rate) may be expressed as:

[pic] (3.7)

The superscript Q reminds us that this is under risk-neutral probability measure and the subscript D represents the entire legal claim for both bond-holders and stock-holders. This represents the complement of the recovery on the firm (6) per unit of total legal claim D, discounted at the risk-free rate. The risk-neutral LGD on the loan is:

[pic] (3.8)

Finally, to close-out the baseline, model, as data-sets of recoveries measure such at emergence from bankruptcy, we must replace the risk-free rate with the drift of the asset-value process[pic] under the physical probability measure P:

[pic] (3.9)

Similarly, the LGD on the firm and on the bond under physical measure are given by:

[pic] (3.10)

[pic](3.11)

1. Extension 1: Stochastic Coupon Rate on the Bond

Now let us generalize this model by allowing the coupon rate on the loan to follow an Orhnstein-Uhlenbeck process:

[pic] (3.1.1)

Where [pic]is the speed of mean-reversion, [pic]is the long-run mean, [pic]is the constant diffusion term, and [pic] is a standard Weiner process having instantaneous correlation with the source of randomness in the firm-value process, given heuristically by [pic]. It can be shown that the expression for the economic LGD in (9) is modified to:

[pic] (3.1.2)

Where the time-dependent volatility is given by:

[pic] (3.3.3)

This is a well-known result (see Bjerkskund, 1991). [pic]

2. Extension 2: Optimal Foreclosure Boundary

Thus far we have taken the solved for the LGD under the assumption that the senior bank creditors foreclose on the Bank when the value of assets is Vt, where t is the time of default. However, this is not realistic, as value fluctuates throughout the bankruptcy or workout period, and we can think that there will be some foreclosure boundary (denoted [pic]) below which foreclosure is effectuated. Furthermore, in most cases there exists a covenant boundary, above which foreclosure cannot occur, but below which it may occur as the borrower is in violation of a contractual provision. For the time being, let us ignore the latter complication, and focus on the optimal choice of [pic]by the bank.

In the general case of time dependency in the loan valuation equation [pic], following Black and Cox (1976), we have the following 2nd order partial differential equation:

[pic] (3.2.1)

This is subject to the boundary conditions:

[pic] (3.2.2)

[pic] (3.2.3)

Following Carey and Gordy (2007), we modify (3.2.3) such that the value of the loan at the threshold is not a constant, but simply equal to the recovery value of the loan at the default time t;

[pic] (3.2.4)

Second, we remove the time dependency in the value of the perpetual debt. This implies the new boundary condition:

[pic] (3.2.5)

This gives rise to second-order ordinary differential equation in lieu of (3.2.1):

[pic] (3.2.6)

The general solution to (3.2.6) is given by:

[pic] (3.2.7)

Where the function [pic] is related to the confluent hypergeometric function [pic] according to:

[pic] (3.2.8)

And the constants are defined as:

[pic]

References

Acharya, Viral V., Bharath, Sreedhar T. and Anand Srinivasan, Understanding the recovery rates on defaulted securities, London Business School, Working Paper, 2003.

, Does industry-wide distress affect defaulted firms – evidence from creditor recoveries, Journal of Political Economy, Forthcoming, 2007.

Altman, Edward I., 1968, Financial ratios, discriminant analysis and the prediction of corporate bankruptcy, Journal of Finance, 23, 589-609.

, R. Haldeman and P. Narayanan, 1977, Zeta( analysis: a new model to predict corporate bankruptcies, Journal of Banking and Finance, 1:1, July, 29-54.

and A. Eberhart, 1994, Do seniority provision protect bondholders’ investments?, Journal of Portfolio Management, Summer, 67-75.

and Vellore M. Kishore, 1996, Almost everything you wanted to know about recoveries on defaulted bonds, Financial Analysts Journal, Nov/Dec.

, Resti, Andrea and Andrea Sironi, 2001, “Analyzing and Explaining Default Recovery Rates”, ISDA Research Report, London, December 2001.

, Resti, Andrea and Andrea Sironi, 2003, Default recovery rates in Credit risk modeling: a review of the literature and empirical evidence, NYU Solomon Center Working Paper #, 2003.

and G. Fanjul, 2004, “Defaults and Returns in the High Yield Bond Market: Analysis through 2003”, NYU Salomon Center Working Paper Series#S-0304, Summer, 67-75.

, Brady, Brooks, Resti, Andrea and Andrea Sironi, 2004, The link between default and recovery rates: theory, empirical evidence and implications, Journal of Business 78(6), 2005.

and S. Ramayanam, 2006, “The High Yield Bond Default and Return Report: 3rd Quarter 2006 Review”, NYU Salomon Center Special Report,November.

, 2006, Default recovery rates and LGD in credit risk modeling and practice: an updated review of the literature and empirical evidence, Working Paper, NYU Salomon Center, November 2006.

Araten, Michel, Jacobs, Jr., Michael and Peeyush Varshney, Measuring Loss Given Default on Commercial Loans for the JP Morgan Chase Wholesale Bank: An 18 Year Internal Study, RMA Journal, May 2004, 28-35.

Asarnow, Elliot and David Edwards, 1995, Measuring loss on defaulted bank loans: a 24-year study, Journal of Commercial Lending, Vol. 77, No.7, pp. 11-23.

Bakshi, G., Madan, Dilip and Frank Zhang, 2001, Understanding the role of recovery in default risk models: empirical comparisons and implied recovery rates, Federal Reserve Board, Finance and Economics Discussion Series, 2001-37.

Basel Committee on Banking Supervision, 2003, “The New Basel Accord”, Consultative Document, Bank for International Settlements (BIS), April.

Basel Committee on Bank Regulation, 2004, “International Convergence on Capital Measurement and Capital Standards”, BIS, June.

, 200%, “Guidance on Paragraph468 of the Framework Document”, BIS, July.

Berger, J., Statistical Decsion Theory and Bayesian Analysis, New York, Springer-Verlag.

Black, Fischer and John C. Cox, 1976, Valuing corporate securities: some effects of bond indenture provisions, Journal of Finance, 31, 637-659.

Bouezmarni, T. and J.M. Rolin, 2001, Consistency of beta kernel density

function estimator, Canadian Journal of Statistics.

Cantor, R., D. Hamilton and P.R. Varma, 2003, “Recovery Rates on Defaulted Corporate Bonds and Preferred Stocks”, Moody’s Investor Service, December.

and P.R. Varma, 2004, “Determinants of Recovery Rates on Defaulted Bonds and Loans for North American Corporate Issuers: 1983-2003”, Moody’s Investor Service, December.

Carey, Mark and Michael Gordy, 2003, Systematic risk in recoveries on defaulted debt, Federal Reserve Board, Working Paper.

, 2004, Measuring systematic risk in recoveries on defaulted debt 1: firm-level ultimate LGDs , Federal Reserve Board, Working Paper.

, 2005, Uncovering recoveries, Federal Reserve Board, Working Paper, February.

, 2007, The bank as grim reaper: debt decomposition and recoveries on defaulted debt, Federal Reserve Board, Working Paper.

Chase Manhattan Bank, 1996, “Loss Given Default Study”, Portfolio Management, Memorandum, September.

Cover, Thomas M. and Joy A. Thomas. Elements of Information Theory. Wiley, New York, 1991.

Credit Suisse Financial Products, 1997, “Creditrisk+ - A Credit Risk Management Framework”, Technical Document.

Davison, A.C. and D.V. Hinkley. 1997. Bootstrap Methods and their Application. Cambridge University Press, Cambridge.

Dev, Ashish and Michael Pykhtin, 2002, Analytical approach to credit risk modeling, Risk, March, S26-S32.

Duffie, Darrell, 1998, Defaultable term structure models with fractional recovery of par, Graduate School of Business, Stanford University.

and Kenneth J. Singleton, 1999, Modeling the term structures of defaultable bonds, Review of financial studies, 12, 687-720.

and David Lando, 2000, Term structure of credit spreads with incomplete accounting information, Econometrica, ###-###.

Eberhart, A. C., Moore, W. T. and R. T. Rosenfeldt, 1989, Security pricing and deviations from absolute priority rule in bankruptcy proceedings, Journal of Finance 45, 747-769

Efron, B., 1979, Bootstrap methods: another look at the jackknife, American Statistician 7, 1-23.

and R, Tibshirani, 1986, Bootstrap methods for standard errors, confidence intervals, and other Measures of statistical accuracy, Statistical Science 1, 54-75.

Emery, K, 2003, Moody’s Loan Loss Database as of November 2003, Moody’s Investor Service, December.

Finger, Chris, 1999, Conditional approaches for CreditMetrics™ Portfolio Distributions, CreditMetrics™ Monitor, April.

Firth, D. (1991), “Generalized Linear Models,” in Statistical Theory and Modelling, ed. Hinkley, D.V., Reid, N., and Snell, E.J., London: Chapman and Hall.

Fitchrisk Management, North American Loan Loss Database, October 201.

Franks, Julian and Walter Torous, 1989, An empirical investigation of US firms in reorganization, Journal of Finance 44, 747-769.

, 1994, A comparison of financial recontracting in distressed exchanges and Chapter 11 reorganizations, Journal of Financial Economics, 35, 349-370.

Fridson, Martin S., Garman, Christopher M. and Kathryn Okashima, 2000, Recovery rates: a search for meaning, Merrill Lynch & Company, High Yield Strategy.

Friedman, Craig and Sven Sandow, 2003a, Ultimate recoveries, Risk, 69-73.

, 2003b, Model performance measures for expected utility maximizing investors, International Journal of Theoretical and Applied Finance, June.

Frittelli, Marco, 2000, Representing sublinear risk measures and pricing rules, Mimeo, July.

Frye, John, 2000a, Collateral damage, Risk, April, 91-94.

, 2000b, “Collateral Damage Detected”, Federal reserve Bank of Chicago, Emerging Issues Series, October, 1-14.

, 2000c, Depressing recoveries, Risk, November, 108-111.

, 2003, A false sense of security, Risk, August, 63-67.

Geske, Robert, 1977, The valuation of corporate liabilities as compound options, Journal of Financial and Quantitative Analysis, 12, 541-552.

Gordy, Michael, 2000, A comparative anatomy of credit models, Journal of Banking and Finance, January, 119-149.

Grünwald, P.D. and A.P. Dawid, 2004, Game theory, maximum entropy, minimum discrepancy, and robust Bayesian decision theory. Annals of Statistics 32(4), pp. 1367-1433.

Gupton, G.M., Gates, D. and Lea V. Carty, 2000, Bank loan loss given default, Moody’s Investor Services, Global Credit Research, November.

, Finger, C. and Mickey Bhatia, 1997, “Creditmetrics™ Technical Document”, J.P. Morgan & Co., New York

and Roger M. Stein, 2005, “LossCalcTM V2: Dynamic Prediction of LGD”, Moody’s KMV.

Hamilton, D.T., Gupton, G.M. and Alexandra Berthault, 2001, Default and recovery rates of corporate bond issuers: 2000, Moody’s Investor Services, February.

Hosmer, D. W. and S. Lemeshow, 1980, A goodness-of-for test for the multiple logistic regression, Communications in Statistics, A10, pp. 1043-1069.

, Hosmer, T., Lemeshow, S. and le Cessie, S., 1997, A comparison of goodness-of-fit tests for the logistic regression model, Statistics in Medicine 16, pp. 965-980.

and S., Lemeshow, 2000, Applied Logistic Regression: Second Edition. John Wiley & Sons, New York.

Hotchkiss, E. S., 1993, The liquidation/reorganization choice of firms entering chapter 11, (New York University, NY).

Hu, Yen-Ting and William Perraudin, The dependence of recovery rates and defaults, Bank of England, 2002.

Hull, John and Alan White, 2002, The impact of default risk on options and other derivative securities, Journal of Banking and Finance, 19, 299-322.

Jaynes, E.T., 1957, Information Theory and Statistical Mechanics, Physical Review, 106, May, pp. 620-30.

Jarrow, Robert A., 2001, Default parameter estimation using market prices, Financial Analysts Journal 55:5, 75-92.

, Lando, David and Stuart R. Turnbull, 1997, A Markov model for the term structure of credit spreads, Review of Financial Studies 10, 481-523.

and Stuart R. Turnbull, 1995, Pricing derivatives on securities subject to credit risk, Journal of Finance 50, 53-86.

Jokivuolle, Esa and Samu Peura, 2003, Incorporating collateral value uncertainty in loss given default estimates and loan-to-value ratios, European Financial Management, Vol. 9, September, 299-314.

Jones, E.S., Mason E. and E. Rosenfeld, 1984, Contingent claims analysis of corporate structures: an empirical investigation, Journal of Finance 39, 611-627.

Keisman, David and Karen van de Castle, 2000, Suddenly structure mattered: insights into recoveries of defaulted debt, Corporate Ratings – Commentary, Standard and Poors.

Kim, I.J., Ramaswamy, S. and S. Sundaresan, 1993, Does default risk in coupons affect valuation of corporate bonds?: A contingent claims model, Financial Management, 22:3, 117-131.

Lafferty, G. and J. Lebanon, 2001, Boosting and maximum likelihood for exponential models. Technical Report CMU-CS-01-144, School of Computer Science, CMU

Loan Pricing Corporation / Fitchrisk Management, North American Loan Loss Database, October 2001.

Longstaff, Francis A. and Eduardo S. Schwartz, 1995, A simple approach to valuing risk fixed and floating rate debt, Journal of Finance, 89, 789-819.

Maddala, G.S. , 1983, Limited Dependent and Qualitative Variables in Finance ( Cambridge University Press, Cambridge).

, 1991, The perspective on the use of limited-dependent and qualitative variables models in accounting research, The Accounting Review 66: 788-807.

Mallick, B.K. and A.E. Gelfand, Generalized linear models with unknown link functions, Biometrika 1994 81(2):237-245.

McCullagh, P. and J. A. Nelder. 1989. Generalized Linear Models. New York: Chapman & Hall.

Modigliani, Frank and Merton H. Miller, 1958, The cost of capital, corporation finance, and the theory of investment, American Economic Review 48, 261-297.

Ohlson, James A., 1980, Financial ratios and the probabilistic prediction of bankruptcy" Journal of Accounting Research, Vol. 18, No. 1, Spring.

Pagan A. and A. Ullah, 1999, Nonparametric Econometrics, Cambridge

University Press, Cambridge.

Pykhtin, Michael, 2003, Unexpected recovery risk, Risk, August, 74-78.

Renault, O. and O. Scalliet, 2003, On the way to recovery: a nonparametric bias free estimation of recovery rate densities, Mimeo, May 2003

Risk Management Association, 2000, Loss in the Event of Default Study.

Rogueries, D. and R. Toffler, 1996, Neural networks and empirical research in accounting, Accounting and Business Research 26, 347-55.

Schleifer, A. and R. Vishny, 1992, Liquidation values and debt capacity: a market equilibrium approach, Journal of Finance, 47, 1343-1366.

Shannon, C.E. , 1948, A mathematical theory of communication, Bell System Technical Journal, Vol. 27, pp. 379-423, 623-656, July, October.

Silverman, B. W. , 1982, Kernel density estimation using the fast Fourier transform, Applied Statistics, Royal Statistical Society, Vol. 33.

Sivia, D. S. 1996. Data Analysis: a Bayesian Tutorial. Oxford University Press.

Vasicek, Oldrich A., 1984, Credit valuation, KMV( Corporation, March.

Venables, W.N. and B.D. Ripley, 1999, Modern Applied Statistics with S-Plus, 3rd Edition (Springer-Verlag, New York).

Wilson, Thomas C., 1998, “Portfolio Credit Risk”, Federal Reserve Bank of New York, Economic Policy Review, October, 71-82.

Weiss, Lawrence, 1990, Bankruptcy resolution: Direct costs and violation of priority of claims, Journal of Financial Economics 27, 285-314

Zhou, Chunsheng, 2001, The term structure of credit spreads with default risk, Journal of Banking and Finance 25, 2015-2040.

-----------------------

[1] Senior Financial Economist, Credit Risk Modelling, Risk Analysis Division, Office of the Comptroller of the Currency, 250 E Street SW, 2nd Floor, Washington, DC 20024, 202-874-4728, michael.jacobs@occ.. The views expressed herein are those of the authors and do not necessarily represent a position taken by of the Office of the Comptroller of the Currency or the U.S. Department of the Treasury.

[2] This is equivalent to one minus the recovery rate, or dollar recovery as a proportion of par, or EAD assuming all debt becomes due at default. We will speak in terms of LGD as opposed to recoveries with a view toward credit risk management applications.

[3] By default we mean either bankruptcy (Chapter 11) or other financial distress (payment default). In a banking context, this defined as synonymous with respect to non-accrual on a discretionary or non-discretionary basis. This is akin to the notion of default in Basel, but only proximate.

[4] Note that this may be either the value of pre-petition instruments received valued at emergence from bankruptcy, or the market values of new securities received in settlement of a bankruptcy proceeding or as the result of a distressed restructuring.

[5] Note that the former may viewed as a proxy to this, the pure economic notion.

[6] In the context of bank loans, this is the cumulative net charge-off as a percent of book balance at default (the net charge-off rate).

[7] There is debate surrounding the appropriate choice for a discount rate. Bank studies (Araten et al, 2003) have put forward arguments for a punitive rate as consistent with the “low end” of what buyers of distressed assets look for as overall return, the uncertainty of recoveries (with the standard deviation about equal to the average), the rates used by commercial loan pricing models, and consistency with “peer practice”. A competing argument among academics (Acharya et al, 2004) as well as practitioners (Friedman et al 2003) and that it is proper to discount ultimate recoveries using the coupon on the debt prior to default. Finally, some have argued in favor of discounting recoveries at the default risk-free Treasury term structure (Carey and Gordy, 2005). We do not address this issue directly in this paper – however, to the extent that one can jointly forecast time-to-resolution and the ultimate recovery, an implicit estimate of the proper actuarial discount rate can be formulated based upon this research.

[8] This is true especially away from the cutting edge or in off-the-shelf models, such as Creditmetrics or ERisk Abacus.

[9] Altman et al (2005) offers an extensive review of the theory and empirical evidence regarding the relationship between LGD and PD, and further documents the influence of the economic state on recovery values and hence the LGD estimate.

[10] Bankruptcies (Chapter 11 reorganizations and Chapter 7 liquidations) and out-of-court settlements (such as distressed exchanges).

[11] Jarrow (2001) also has the advantage of isolating the liquidity premium embedded in defaultable bond spreads.

[12] Subsequently Fitchrisk and Algorithmics.

[13][pic]%ABbd†ˆ«­Åà & 9 : ; p q { € ‚ † ‡ ‹ æÑËæÑæÑæÑæ¶Ñ¶Ñ§”z”hVhVDhVh"hƒH‚CJOJQJ^J[14]aJmH sH "há0oCJOJQJ^J[15]aJmH sH "hß

%CJOJQJ^J[16]aJmH sH 2jhß

%0JCJOJQJU[pic]\?^J[17]aJmH sH %hß

%CJOJQJ\?^J[18]aJmH sH hß

%OJQJ\?^J[19]mH sH (há0o5?CJOJQJ\?^J[20]aJmH LPC (RMA) find an average accounting LGD of 30% (27%) for 2,534 (977) loans, both in the 1998-2001period.

[21] In a comparable period, both authors find median LGDs on senior secured loans, senior secured bonds, unsecured loans and senior unsecured bonds were found to be 27.0%, 45.5%, 49.5% and 57.7%.

[22] Median LGDs of 49.5% and 66.5% for defaulted issuers originally investment and speculative grade, respectively.

[23] Moody’s DRSTM, S&P’s LossStatsTM and S&P’s CreditprioTM , and the Society of Actuaries private placements database.

[24] Correlations of 0.45 (0.80) for senior (subordinated) debt.

[25] Carey ands Gordy (2004) argue for a 2 stage approach to measuring LGD, first estimating an “estate LGD” at the obligor level, and then treating instrument level LGDs according to a contingent claims approach, as under the Absolute Priority Rule (APR) such recoveries can be viewed as collar options on residual value of the firm. However, they argue that the endogeneity of the bankruptcy decision will result in a measurement problem in the 1st stage borrower level. Furthermore, an extensive literature on violations of APR suggests a similar problem in the 2nd stage instrument level (Hotchkiss [1993], Eberhart et al [1989], Weiss [1990]).

[26] Altman (2006) reports that his model over-predicts LGD in recent years, which he speculates may be due to bubble conditions in the high-yield market.

................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download