Rating Market Value CDO

STRUCTURED FINANCE

Special Report

Moody’s Approach to Rating Market-Value CDOs

AUTHORS:

CONTENTS:

Yvonne Fu Falcone, Ph.D. • Assistant Vice President • (212) 553-1494 Jeremy Gluck, Ph.D. Managing Director (212) 553-3698 CONTACTS: Eileen Murphy Managing Director (212) 553-4061 Julie A. Culhane Investor Relations (212) 553-7941

Introduction Transaction Structure Factors Affecting Asset Price Volatility Volatility in a Portfolio Context The Role of Liquidity Measuring Portfolio Volatility Sources of Market Data Volatility and Correlation Adjustments and Liquidity Applying the Simulation Approach Role of the Collateral Manager Legal Concerns Summary Conclusion Exhibits

• • • • • • • • • • •

INTRODUCTION
As the market for collateralized debt obligations (CDOs) has mushroomed in recent years, the range of structures has widened considerably. For example, the underlying collateral in CDOs has evolved toward a variety of emerging market and developed country bond and loan types and away from an almost exclusive reliance on speculative-grade U.S. corporate bonds. At the same time, structures have changed to include “ramp-up” periods and a greater degree of dynamism. A minority of CDOs have been cast as market-value transactions, in which the credit enhancement is reflected in a cushion between the current market value of the collateral and the face value1 of the structure’s obligations. Within this framework, the collateral must normally be liquidated, either in whole or in part, if the ratio of the market value of the collateral to the obligations falls below some threshold. The liquidated collateral is used to pay down obligations, bringing the structure back into balance. In contrast, cash-flow transactions normally provide for the diversion of cash flows from junior to senior classes if certain tests that relate to the structure’s soundness are not met. Since the primary risk in a market-value transaction is that of a sudden decline in the value of the collateral pool, our analysis will focus on the price volatility of the assets that may be incorporated into these structures. We will see that this volatility can be reflected in a set of advance rates that represent adjustments to the value of each asset and are designed to provide a cushion against market risk.
1 To be more precise, the obligation is principal plus accrued interest.

April 3, 1998

THE STRUCTURE OF A MARKET-VALUE TRANSACTION
In large part, cash-flow CDOs have succeeded because they exploit the illiquidity of the highyield markets.2 The spreads on high-yield debt have historically more than compensated for the default risk associated with such debt, making it attractive to own such instruments on a buyand-hold basis.3 The spreads thus include a sizable liquidity premium. However, these spreads have diminished over time, partly because CDOs have proven to be an effective arbitrage. As soon as it is necessary to sell speculative-grade debt within a market-value structure, the liquidity premium is absorbed. The attempt to sell illiquid assets, particularly in size, will tend to drive down bid prices. Worse yet, if a fraction of a portfolio is sold in a troubled market, “normal” bid-ask spreads will widen considerably, and substantial losses may be incurred.

Still Attractive for Investors and Collateral Managers
Nonetheless, market-value transactions may be attractive to some investors and collateral managers. The structure makes particular sense where the collateral pool consists of instruments that do not produce predictable cash flows, such as distressed debt. Even for pools of conventional fixed-income instruments, the market-value structure may appeal to investors who are most comfortable in a mark-to-market environment, such as that provided by hedge and mutual funds. Collateral managers may prefer the greater trading flexibility that arises from the dynamic nature of the market-value tests. Market-value transactions also facilitate the purchase of assets that mature beyond the life of the transaction because the price volatility associated with the forced sale of such assets is explicitly considered. The capital structure of a market-value CDO resembles that of Figure 1 a cash-flow transaction. Figure 1 depicts a simple marketA Two-Tranche Market-Value Structure value CDO structure. The market-value concept is straightforward. A cushion of Collateral Pool CDO excess value protects investors from a sudden decline in the value of the portfolio; that is, unless the value of the assets $240 mm falls by an amount that exceeds the cushion, the noteholders Senior Notes can get out whole. Should the ratio of the value of the assets $300mm to the liabilities fall below some threshold – 1.25:1 in the case Assets illustrated in Figure 1 – then assets are sold and liabilities paid off until the threshold can again be satisfied. Thus when rating $60mm Equity such an instrument, the key consideration is determining the (downside) volatility of the market value of the portfolio, as well as any liquidity losses that might be incurred through a “forced” sale of assets.

FACTORS THAT AFFECT ASSET PRICE VOLATILITY
The returns on speculative-grade bonds may vary for several reasons, among them: • Changes in the rating of the instrument • Actual default • Changes in interest rates • Changes in investor preferences (and resulting changes in credit spreads)

Changes in Ratings
A speculative-grade bond is particularly vulnerable to changes in ratings. A rating change, which is reflective of a change in the likelihood of default, affects the discount rate that the market applies to the cash flows promised by the bonds. The discount rate must capture the likelihood of default, although it may also embody a liquidity premium and other factors. One can therefore think of the price impact of a change in a bond’s rating as the change in value implied by the change in the discount factor.4 The likelihood of any particular change in rating is given by a ratings transition matrix.5
2 Both cash-flow and market-value deals may also benefit from diversification, which reduces the likelihood that losses will exceed a structure’s credit enhancement. 3 See, for example, Bencivenga, Joseph C., “The High-Yield Corporate Bond Market” in The Handbook of Fixed Income Securities, Frank J. Fabozzi, Ed., Chapter 15. 4 This relationship is a focus of JP Morgan’s CreditMetrics. 5 See “Moody’s Rating Migration and Credit Quality Correlation, 1920-1996,” Moody’s Special Comment, July 1997.

2•

Moody’s Approach to Rating Market-Value CDOs

Defaults
Of course, an extreme case of a change in rating is the movement to actual default. The same rating transition matrix that suggests potential changes in rating level also gives the likelihood of default for a bond with any initial rating. However, the price impact of the change is not easily represented as a change in discount rate because any instrument in default can no longer be thought of as promising a well-defined set of cash flows. Instead, the price impact can be inferred from Moody’s studies of the recovery value of an instrument following default.6

Changes in Interest Rates
Although the returns on speculative-grade instruments are less closely associated with changes in interest rates than those of similar-maturity investment-grade bonds, they are nonetheless affected by interest-rate shifts through a corresponding change in discount rate. However, this price sensitivity may only be observed when the rating of the bond is held constant. Various studies have demonstrated that yield spreads and the general level of interest rates are negatively correlated because high rates tend to be associated with a strong economy, an environment in which firms are less likely to default.7

Changes in Investor Preferences/Credit Spreads
Even if a bond’s rating and the general level of interest rates remain unchanged, the returns on speculative-grade bonds may vary for a variety of reasons. For example, investors may grow concerned about the performance of the economy and choose to shun riskier investments. Or, regulatory actions may alter the relative appeal of speculative-grade bonds. Investors may also perceive that there are factors affecting the attractiveness of investments in the debt of a particular company that are not reflected in the firm’s debt rating. For these reasons and others, one cannot expect to fully explain the price volatility of speculative-grade bonds exclusively through changes in ratings and interest rates. These changes in preferences produce changes in credit spreads that have their greatest impacts on the most interest-sensitive bonds.

Special Considerations for Distressed Instrument Returns
Market-value structures are well suited to the inclusion of distressed instruments. These obligations of companies that are in bankruptcy proceedings do not produce a predictable stream of cash flows. Rather, investors may find distressed instruments attractive because they offer an equity-like potential for capital gains. The instruments include distressed debt (bonds and loans), equity that has been converted from debt through a reorganization, and trade claims. Trade claims typically represent obligations to suppliers of inputs to distressed firms. The return behavior of distressed instruments more closely resembles that of equity than debt. But unlike conventional equities – where dividend levels, growth rates, and interest rates are the key drivers of valuation – the prices of distressed instruments may be subject to sudden shocks as it becomes clear during the bankruptcy process that some claimants will fare better than others, or as the market attempts to determine the value of the distressed firm as a going concern in comparison to its liquidation value.

VOLATILITY IN A PORTFOLIO CONTEXT Portfolio Effects
The factors we have just discussed determine the volatility of the returns on a particular speculative-grade or distressed instrument. When considering the volatility of the returns on a portfolio of such instruments, one must also consider portfolio effects. The more diversified the portfolio, the less volatile will be the portfolio return, relative to the return volatility of its components. The degree to which a portfolio is well diversified depends not only on the portfolio shares associated with each component, but also on the correlations among the various sources of price volatility for each of the components. For example, there will be a tendency for the ratings of U.S. speculative-grade bonds to move together as the U.S. economy ebbs and flows. This suggests some co-movement in returns. To the extent that speculative-grade bonds are affected by changes in interest rates, variations in rates will induce some return correlation. If
6 See “Historical Default Rates of Corporate Bond Issuers, 1920-1997,” Moody’s Special Comment, Feb. 1998. 7 See, for example, Francis A. Longstaff and Eduardo S. Schwartz, “A Simple Approach to Valuing Risky Fixed and Floating Rate Debt,” The Journal of Finance, July 1995.

Moody’s Approach to Rating Market-Value CDOs

•3

investors suddenly decide that speculative-grade bonds are undesirable, that too will force some correlation in the prices of these instruments. Sources of correlation need not be economy-wide. For instance, to the extent that firms within a particular industry exhibit similar economic performance, there may be a corresponding tendency for the prices of bonds issued by firms in the industry to be positively correlated. Investor attitudes toward an industry may have a similar impact. In some cases, price correlation may arise from regional, rather than industry-related factors. The relevant regions may be international: the recent sharp downturn in emerging market bond prices during the Asian crisis clearly cut across national borders.

THE ROLE OF LIQUIDITY
As we have already suggested, speculative-grade bonds are illiquid relative to investment-grade instruments. Distressed instruments are even less liquid. Hence, if one were to sell a speculative-grade or distressed bond in a “normal” market, the seller would suffer some dollar loss versus fair market value due to the presence of a bid-ask spread. But in assigning a rating to a market-value transaction, our concern is not what will happen in a more normal market. Rather, the concern is what will happen in a deteriorating market because it is only in such a market that investors in rated tranches stand any real chance of suffering a loss. Unfortunately, bid-ask spreads are almost certain to widen in this environment as investors try to bail out and business is no longer characterized by “two-way” flows. Indeed, during the recent Asian crisis, some market participants remarked that they “could not get a bid” for the affected bonds. Presumably, this isn’t literally true – the bonds could be sold at some price; nonetheless, it’s clear that liquidity had nearly dried up for certain issues. Going back even further, defaults by several Latin American countries in the 1980s produced sudden dislocations in the market that made it very difficult to dispose of assets without suffering sharp losses. To the degree that a particular portfolio manager is known to be engaged in a “fire sale” of assets, the tendency for spreads to widen will be exacerbated by an effective shift in the fair market value of the instruments that the manager is attempting to unload. This is relevant for the manager of a market-value CDO that must dispose of assets in order to ensure that the asset/liability value ratio remains above its threshold. The fact that the speculative-grade market has performed well over the past six years may exacerbate the problem in a market downturn, because market participants have little recent experience with a broad decline in prices.

MEASURING PORTFOLIO VOLATILITY
The price volatility of a portfolio can be modeled in a number of ways. The primary alternatives are analytic and simulation approaches. • An analytic approach would allow the calculation of the potential decline in value by applying some formula that captures each source of price risk. Unfortunately, that is quite difficult for a portfolio that may contain both fixed-income products and those that fail to generate predictable cash flows. • A simulation approach addresses changes in portfolio values directly, or simulates the factors that determine portfolio value. The direct approach is simply to simulate changes in the values of each of the assets that comprise the portfolio. The indirect approach entails a simulation of changes in ratings, interest rates and, perhaps, other factors that will lead to changes in market value. In combination with the characteristics of each component of the collateral pool – rating, duration, and the like – potential changes in portfolio value may be measured.8

Value of the Direct Simulation Approach
We have chosen to directly simulate asset price movements because we believe that the relationships between asset prices and the characteristics of each asset are unstable during the periods of greatest interest – market downturns. At such times, the “noise” associated with changing investor preferences will dominate. Moreover, the data requirements associated with the indirect approach, both for econometric estimation and for the collateral manager in running a “live” portfolio, are quite burdensome.
8 Note that CreditMetrics, for example, is only intended to addresses changes in value associated with credit events. Hence, neither the impact of interest-rate changes nor changes in investor preferences are considered. A full-blown analytic model would relate the characteristics of each asset – rating, duration, convexity – to the distributions that govern credit migration, interest rates and credit spreads.

4•

Moody’s Approach to Rating Market-Value CDOs

Even the choice of direct simulation of asset prices admits two possible approaches. The first – the parametric approach – requires assumptions about the relevant distributions and incorporates the parameters of those distributions within the simulation. This method has the advantage of allowing the generation of a very large number of scenarios, including those that have not been observed in the past. The second approach relies on a historical simulation. In this case, rather than generating changes in market variables from a theoretical distribution, one draws actual percentage changes in price (or in the underlying factors) from history. The advantage of the historical simulation approach is that no assumptions need be made about the distributions of market variables; the disadvantage is that only historically observed changes can be simulated.

A Blend of the Historical and the Parametric Approaches
Moody’s has chosen an approach that is primarily historical, but that provides flexibility to impose certain parametric choices where the price history is inadequate. The historical method is attractive because the parameters that affect the underlying distributions are highly unstable. Most notably, correlations can change dramatically over time, even changing sign on some occasions. Of particular concern is the fact that correlations often rise sharply during a period of crisis, having an enormous impact on the potential decline in portfolio value during a short period of time. The October 1987 stock market crash or the Asian crisis are stark reminders that seemingly independent markets often become unexpectedly linked when a sharp downturn occurs. Because of these high correlations during stressful periods, not only is the common assumption of normal returns at the individual asset level invalid, but the assumption is also inappropriate at the portfolio level.9 We will not, however, rely exclusively on history. The reason is simple: we don’t have a sufficient price history for each asset class to provide comfort that we have captured all the market scenarios that could reasonably occur. It is particularly difficult to stratify the data in order to select a meaningful sample comovements in prices during relatively rare, but critically important, market downturns. In these cases, we will impose volatility assumptions on asset classes, and correlation assumptions within and across asset classes, by reasoning that the price behavior should be similar to that of assets that are close substitutes, for which we have a more extensive price history. We will also, where necessary, impose higher volatilities and correlations than are embedded in the historical data. Such adjustments are appropriate offsets to incomplete data. The more limited our data, the greater will be the adjustment.

SOURCES OF MARKET DATA
As a rule, the less liquid an instrument, the more difficult it is to obtain historical price data for the instrument. Therefore, it is somewhat more difficult to find price data for high-yield bonds than for investment-grade corporate bonds. It is still more difficult to obtain data for loans. Finally, it is particularly difficult to develop a price history for distressed instruments. Table 1 outlines the sources for a range of instrument types10 used in our analysis. Note that all instruments described in Table 1 and subsequent tables are US dollar denominated domestic securities; thus all results presented in this article do not apply to foreign securities.
Table 1

ASSET RETURN DATA SOURCES
Asset Type High-yield bonds High-yield loans Distressed bonds Distressed loans Distressed equity Data Source Period Covered Interactive Data Corporation 1982 – 1997 Loan Pricing Corporation 1991 – 1997 Moody’s Distressed Bond Database 1987 – 1997 Loan Pricing Corporation 1991 – 1997 PPM America, Inc. 1992 – 1996 Number of Assets 1500 213 470 106 58

9 The tails of the return distributions – precisely our focus in assigning a rating – are much too “fat” to be consistent with normality. This is true not only of individual assets, but even of entire portfolios because of the high degree of correlation that prevails during market breaks. Moreover, the distributions are skewed in the sense that sharp market downturns are more likely than sharp rallies. The normal assumption may be an adequate characterization of average portfolio behavior; however, our concern with the lower tail of the return distribution implies that the skewness and kurtosis (“fat tails”) associated with the returns for the relevant assets cannot be neglected. 10 It is extremely difficult to obtain historical price information on certain instruments, for example trade claims. For transactions involving trade claims, we stress the reorganized equity data to obtain advance rates for trade claims. As the price information for trade claims, or other instrument types, becomes available, we will incorporate them in our historical database.

Moody’s Approach to Rating Market-Value CDOs

•5

Although the number of assets representing each asset type is often large, the data are “sparse” in the sense that many of the assets remain in the samples for a relatively short period of time. High-yield debt may be upgraded to investment-grade or may default; in either case, the instrument will no longer be tracked in a high-yield database. Distressed instruments either disappear or lose their distressed status following the conclusion of the bankruptcy process.

VOLATILITY AND CORRELATION ADJUSTMENTS AND LIQUIDITY ASSUMPTIONS Volatility
We have adjusted the volatility of historical returns by multiplying each return by a factor that reflects both: 1. The length of the historical record 2. The desired rating for the CDO tranche. Because we have the most complete record for high-yield bonds, we apply a relatively small stress to the historical volatility for this instrument. At the other end of the spectrum, a relatively large stress factor is applied to distressed instruments, especially reorganized equities. The higher the desired rating of the tranche, the greater the stress factor we apply. The reasoning is straightforward: highly rated instruments rarely default. Therefore, an exceptionally long data series would be required to produce a valid test of a structure when the target rating dictates that a default should not occur more than, say, once every 1000 years. Lacking centuries or millennia of data (and an economic environment that remained stable over the entire period!), a degree of stressing is called for.11 There is one exception to our rule of applying the greatest stress to the shortest data series. Though we have a moderately large number of observations on performing loans, there is reason to believe that the data exclude some important events. Performing loans become distressed loans as soon as a borrower defaults. Unfortunately, there are very few cases in which we can observe the price impact of default, which may be quite large. Most of the loans that surface in the distressed data were not tracked when they were performing loans, so that the one-month price change cannot be observed. This lack of information warrants a harsher stress factor. The specific factors that are applied to the historical returns for each asset class and desired rating are outlined in Table 2.
Table 2

STRESS FACTORS FOR ASSET RETURNS BY ASSET TYPE AND RATING
Rating Asset Type Performing Bank Loans Performing High-yield Bonds Distressed Bank Loans Distressed Bonds Distressed Equities B 1.40 1.00 1.05 1.00 1.40 Ba 1.60 1.00 1.10 1.00 1.50 Baa 1.80 1.10 1.20 1.10 1.60 A 2.00 1.20 1.30 1.20 1.70 Aa 2.20 1.30 1.40 1.30 1.80 Aaa 2.50 1.40 1.60 1.40 2.00

11 Note that this mirrors our approach with respect to cash-flow transactions, where instead of stressing return volatilities, we stress default rates.

6•

Moody’s Approach to Rating Market-Value CDOs

In addition to boosting volatilities to reflect relatively limited data, we have also made an adjustment for data series where the prices do not appear to be true “trading” prices. For distressed instruments, the reported prices sometimes remain unchanged for a few consecutive months. We have removed these “zero-return” observations from the data, which has the impact of raising overall volatility. The stressing of asset returns directly increases the volatility of returns for the individual assets.12 By contrast, the stressing of correlations acts indirectly by raising portfolio volatility.

Correlation
Appropriate correlations are particularly difficult to infer from historical (or any other) data, because although correlations clearly rise during market downturns, there are few observations on such periods. We have imposed intra- and interindustry correlations by observing correlations for pairs of firms over a relatively long sample period and by comparing those longer term measures to correlations that prevailed in relatively stressful periods, such as 1987 and 1990-1991. We have chosen levels that are higher than those that prevail during “normal” periods to reflect our concerns about data imperfections; they are not, however, as high as those observed during the most stressful periods. The data are not sufficiently rich to provide a basis for distinguishing between correlations for particular pairs of industries. Rather, we use the assumptions presented in Table 3. Although correlations are embedded in the historical return Table 3 data, we impose these assumptions by sampling from the CORRELATION ASSUMPTIONS database in a nonrandom fashion. We achieve this by first ranking the returns for each asset from lowest to highest, and Pairs of firms: Assumed Correlation (%) by then selecting the individual asset returns in such a way that Within same industry 55 the correlation assumptions in Table 3 will be satisfied. 13 In different industries 40

Liquidity
Whenever an asset is sold, the seller must theoretically sacrifice half the bid-ask spread—the difference between a mid-market price and the bid. For most instruments in most markets, this is a trivial consideration in comparison to ordinary price volatility. For the instruments that typically find their way into market-value transactions, liquidity becomes a key consideration during a period of financial stress. We have made assumptions about the loss that a seller of a Table 4 high-yield or distressed asset would incur as a result of illiqLIQUIDITY “HAIRCUTS” BY ASSET CLASS uidity. Although we have tried to preserve the ordinal relationships among the bid-ask spreads for each instrument – the Liquidity assets with the largest bid-ask spreads receive the greatest Asset Type “Haircut” (%) liquidity “haircuts” – the discount for illiquidity greatly exceeds normal bid-ask spreads. Since no reliable time-series data Performing Bank Loans 7 exist on bid-ask spreads for the relevant assets, our assumpPerforming High-yield Bonds 5 tions follow from discussions with market participants and their Distressed Bank Loans 12.5 views as to how difficult it would be to sell the assets during a Distressed Bonds 10 sharp market downturn. Table 4 contains the liquidity assumptions for the respective asset classes. Reorganized Equities/ Trade Claims 20

12 If the random return for an individual asset has a standard deviation of σ, then stressing the return by a factor λ implies that the standard deviation of the stressed return will be λσ. 13 Specifically, let the historical return data be arranged in a matrix of m rows by n columns, where m is the number of issues, n is the number of periods, and the returns for each issue are ranked from the lowest to the highest. In a historical simulation, we need to select the row and column indices and obtain the return for each position in the portfolio. For any position in the portfolio, we select the issue (the row index) by sampling from a uniform distribution. When selecting the column indices for the positions, we first sample from a correlated normal distribution based on the correlation parameters in Table 3; the number of samples drawn for each simulation iteration is equal to the number of positions in the portfolio and thus each sample corresponds to a position in the portfolio. Each of these correlated normally distributed drawings is then mapped into a number ranging between 0 and 1 by using the inverse of the cumulative of the standard normal distribution. If we multiply this number between 0 and 1 by n (the total number of periods), the result gives the column index for a position in a portfolio. Because the returns are ordered for each issue, we wind up with correlations roughly consistent with the assumptions in Table 3. We say “roughly” because we have imposed “rank order” correlation, which may differ somewhat from the usual correlation measure.

Moody’s Approach to Rating Market-Value CDOs

•7

APPLYING THE SIMULATION APPROACH The Calculation of Advance Rates
As in cash-flow CDOs and other structured transactions, Moody’s rates market-value transactions on an expected loss basis. That is, the expected loss faced by the investor at a given rating level must be consistent with the experience of investors in like-rated conventional instruments. The expected loss measure revolves, in turn, around a particular set of advance rates that applies to the collateral pool. In a typical market-value transaction, an overcollateralization (OC) test ensures that the market value of assets discounted by the appropriate advance rates exceeds the liabilities. The advance rates thus represent “haircuts” that provide credit enhancement for the rated notes. We present here examples of the calculation of the advance rates in a manner that results in an expected loss that is consistent with the desired ratings for the debt tranches issued within the CDO. For the sake of concreteness, we assume here that the OC test is performed biweekly and the cure period, during which assets may be liquidated, is 10 business days. In practice, the advance rates will reflect the marking and cure periods proposed within each transaction. If the OC test is failed, assets have to be sold either to restore the balance or to unwind the deal. A loss occurs if the deal unwinds and the proceeds from liquidating the portfolio fall short of the obligations. Since the maximum period of time during which the portfolio is subject to market price movements is the mark-to-market period plus the cure period (a total of one month), the concern is whether the advance rates provide sufficient protection over a one-month period to offset the volatility and the illiquidity of the assets in the portfolio. We rely on our historical simulation approach to model the changes of the market value of the portfolio over the exposure period starting with the latest run of the OC test. We make the conservative assumption that in the last period, the market value of the portfolio (V), discounted by the portfolio advance rate AR,14 was exactly equal to the indebtedness (D). Hence, V·AR=D, or V=D/AR. The portfolio return over the subsequent month is calculated by randomly sampling returns from our database for each of the assets in the portfolio, which results in a return (and an end-of-month market value) for the overall portfolio. A loss to the investors will occur whenever the end-of-month value of the portfolio falls short of the indebtedness. If the monthly portfolio return is rp, a loss will occur if D-V(1+ rp)>0, or D(D/AR)(1+rp)>0, Put differently, the loss (L), relative to what is owed to investors in a particular return scenario, is

max(0,D –

(1) L=

D (1+rp)) AR

D

Notice that a loss occurs if the portfolio return is sufficiently negative that (1+rp)< AR. The expected loss, E(L), is the average of the losses across all the scenarios within the simulation:

(2) E (L)= ∫ -1 (1-

AR-1

1+rp AR

)ƒ(rp)drp

where f(rp) is the probability density function for rp. Although we don’t know what f(rp) looks like, the historical simulation will effectively map out the portfolio return distribution. From Equation (2), we see that the appropriate advance rates for a transaction seeking to achieve a certain expected loss target depend on the left tail of the distribution of the portfolio return, which in turn depends on the attributes of the portfolio itself, including diversification, asset types and asset compositions. The expected loss considered above applies to a particular one-month period. A CDO, however, normally has a life of several years. The calculation of expected loss thus requires a repetition of the one-month calculation over the life of the CDO – say, for 60 months in the case of a 5-year
14 The advance rate for the portfolio is the weighted average of the individual advance rates where the weights are portfolio shares for the different asset types.

8•

Moody’s Approach to Rating Market-Value CDOs

transaction. If, at any point over the five-year period, the entire portfolio must be liquidated and at least some of the investors take a loss, then the simulation stops at that point. In order to arrive at the appropriate rating for each tranche, one would compare the expected loss for the tranche over the entire 5-year period to the expected loss for a five-year conventional bond.

Examples of the Advance Rate Calculation
The Simplest Case: One Rated Tranche and Up to 100% in a Single Asset Type

In the following discussion and in the accompanying exhibits, we provide examples of the calculation of the advance rates for specific structures. We emphasize that the particular calculation must be consistent with the structure actually proposed, which may be quite complex. The tables will, nonetheless, provide insights as to how various factors affect the advance rates. We begin with the simplest case of a one-tranche structure – one with subordination provided only by the advance rates. We make the following assumptions regarding portfolio diversification. 1. Maximum allowable investment in one issuer is 5%. 2. Maximum allowable investment in one industry is 20%. 3. Maximum allowable investment in one asset type is 100%. The least diversified portfolio consistent with these standards thus consists of 20 issuers and 5 industries. We obtain the empirical return distribution of this least diversified portfolio, with 100% invested in one asset type, by sampling the returns of individual positions from the historical price movements for that asset type. Assuming that the maturity of the transaction is five years, we then evaluate expected loss with different advance rates to obtain Table 5. Table 5 provides guidance for assigning advance rates for different asset types for the given diversification criteria of the portfolio. For specific transactions, certain structural provisions can result in more or less favorable advance rates for some or all of the asset types. We now consider the impacts of these provisions.
Table 5

Advance Rates for Different Asset Types and Rating Levels
(20 issuers, 5 industries, 100% investment in one asset type, 5 year maturity)

Asset Type Aaa Aa1 Aa2 Performing Bank Loans Valued $0.90 and Above 0.870 0.890 0.895 Distressed Bank Loans Valued $0.85 and Above 0.760 0.780 0.790 Performing High-Yield Bonds Rated Ba 0.76 0.79 0.80 Performing High-Yield Bonds Rated B 0.72 0.75 0.76 Distressed Bank Loans Valued Below $0.85 0.58 0.62 0.63 Performing High-Yield Bonds Rated Caa 0.45 0.49 0.50 Distressed Bonds 0.35 0.39 0.40 Reorganized equities 0.31 0.37 0.38

Target Rating Aa3 A1 A2

A3

Baa1 Baa2

Baa3 0.940 0.870 0.90 0.85 0.74 0.67 0.57 0.54

0.900 0.905 0.910 0.915 0.930 0.935 0.795 0.810 0.815 0.820 0.830 0.840 0.81 0.77 0.64 0.51 0.41 0.39 0.83 0.78 0.67 0.56 0.47 0.44 0.84 0.79 0.68 0.58 0.48 0.46 0.85 0.80 0.69 0.60 0.50 0.47 0.87 0.82 0.71 0.62 0.54 0.51 0.88 0.83 0.72 0.64 0.56 0.52

• Diversification Across Assets and Industries One of the most important determinants of the advance rates is the extent to which the portfolio is diversified, both by asset and by industry. Between the two, asset diversification is, perhaps, more important. The data suggest that in stressful times, correlations for high-yield and distressed assets within and across industries become somewhat similar. The assets behave as a class, rather than as members of an industry. Diversification across industries is of some value, but it is perhaps less critical than in the case of cash-flow transactions.
Moody’s Approach to Rating Market-Value CDOs •9

The effect of diversification on the advance rates is reported in Exhibit A. • Asset Type Limitations The advance rates for each asset type in Table 5 are obtained under the assumption that 100% of the portfolio investment is in that asset type. In some transactions, portfolio limitations are established with respect to investment in certain asset classes. For those transactions, we would need to test whether the advance rates shown in Table 5 are still appropriate for these asset classes given the portfolio limitations. Because of the interaction between various factors, there is no theoretical conclusion as to what the outcome should be. In some instances, more favorable advance rates might be achieved due to the restrictions. In other instances, the advance rates are unaffected. The outcome depends on the possible asset types in the portfolio, as well as any limitations on portfolio composition. Virtually any set of asset-class restrictions might apply. To focus on a particular case, suppose that the collateral manager wishes to invest mainly in performing high-yield bonds rated B3 and above, as well as in distressed bonds, and is willing to limit the investment in distressed bonds to be at most 30% of the portfolio. Given this limitation, our simulation produces advance rates for distressed bonds that are higher than those reported in Table 5. In fact, for a transaction in which an A2 rating is sought, the advance rate for distressed bonds rises from 0.48 to 0.54; for a transaction geared toward a Baa2 rating, the advance rate rises from 0.56 to 0.61. The advance rates associated with these two target ratings for the possible asset types in this transaction are presented in Table 6.
Table 6

ADVANCE RATES WITH A RESTRICTION ON THE PROPORTION OF DISTRESSED BONDS
(20 issuers, 5 industries, maximum 30% in Distressed Bonds)

Rating Asset Type Performing High-yield Bonds Rated Ba1-Ba3 Performing High-yield rated B1-B3 Distressed Bond A2 0.84 0.79 0.54 Baa2 0.88 0.83 0.61

To demonstrate that such limitations do not always result in higher advance rates, consider a case in which the collateral manager wishes to invest in bank loans and that up to 30% of the portfolio will be invested in distressed bank loans. Simulation results reveal that the advance rates for distressed bank loans shown in Table 5 are in fact not affected by the portfolio limitation in this case. Further examples of the impact of portfolio limitations are presented in Exhibit B. • Subordination If there are multiple debt tranches and there is a set of advance rates for each tranche, subordination might provide credit enhancement beyond that associated with the advance rates for senior tranches. This, however, will depend on the relative sizes of the debt tranches, the desired ratings and the possible portfolio composition limitations imposed by the structure. The impact of subordination is explored in Exhibit C. • A Minimum Net Worth Test May Permit Higher Advance Rates Some structures contain a minimum net worth test to ensure that a certain percentage of the initial equity level must be maintained; the test is often performed with a different frequency from the monthly period that is otherwise relevant. The impact of the test is to introduce a second relevant horizon – say, one quarter – over which changes in portfolio value must be evaluated. When the minimum net worth test is performed with a lower frequency than the OC test, we need to run simulations to evaluate the effect of this impact; indeed, it would be very difficult to incorporate such a test without running simulations. The introduction of a quarterly minimum net worth standard may serve to raise the advance rates by assuring that at the beginning of each quarter, equity will exceed some level. Even

10 •

Moody’s Approach to Rating Market-Value CDOs

though the level may fall during the quarter, a decline in value is by no means assured, potentially leaving a bit of extra cushion beyond what the advance rates provide. The interplay between a minimum net worth test and the advance rates is explored in Exhibit D.

THE ROLE OF THE COLLATERAL MANAGER
Whatever the price history for high-yield and distressed assets may suggest, it is ultimately the responsibility of the collateral manager to make intelligent choices with regard to the buying and selling of assets. Thus, a rigorous analysis of portfolio price volatility must be supplemented by a subjective evaluation of the manager. Obviously, a manager that selects undervalued assets and sells prior to a deterioration in price is superior to one that lacks these instincts. However, Moody’s ratings are not based on a judgment that a particular manager will outperform the market. Rather, we seek comfort that a manager understands and can modify, where appropriate, the risks embedded in the asset portfolio. Hence, a track record of high absolute returns would be less reassuring than one of solid risk-adjusted returns. The credit analysts employed by the management firm should have an in-depth knowledge of the firms represented within the CDO portfolio. Credit analysis should be systematic and well documented. A process of seeking first-hand information, rather than relying exclusively on industry research, also gives reassurance. Of immediate concern in a market-value transaction is the ability to understand correlations in prices so that diversification will limit the downside risk in the portfolio. In evaluating a particular manager, Moody’s draws comfort from depth of experience in managing the types of assets that will constitute the CDO collateral pool. By “depth,” we refer to both the number of years of experience that the management team can claim and to the size of firm’s staff. The fate of the CDO should not be dependent on a single individual. Redundant management skills, credit analysis and systems expertise characterize better collateral managers. The back office environment is no less important than the activities of the front office. Though the trustee for the transaction will have a role in verifying transactions and the status of the portfolio, the collateral manager should be in a position to track the portfolio and to verify that any tests embodied in the transaction’s documents are met. The presence of independent auditors and control personnel, that do not report to the portfolio manager, give additional comfort. A particular concern in the market-value context is the marking to market of the collateral. Many of the instruments are illiquid, making it difficult to obtain meaningful mid-market price quotes. Despite the expertise of the manager in valuing the instruments, marks by competent, independent sources provide greater reassurance. For instruments whose prices are not typically quoted by dealers, it is often possible to obtain marks with some frequency from other sources, such as independent auditors or investment banks. The “haircuts” that we have applied to the more illiquid instruments reflect, in part, the concern that the market prices of such assets may be difficult to establish prior to an actual sale.

LEGAL CONCERNS
Though a thorough review of the documents governing a market value CDO is an absolute prerequisite to assigning a rating, the legal issues for these transactions are similar to those relevant to cash-flow transactions. In particular, the structure must mitigate concerns about (1) bankruptcy remoteness (so that the SPV will not face claimants other than the senior noteholders that might force a filing for bankruptcy), (2) true sale (i.e., the assets sold to the SPV are purchased in a legitimate fashion), and (3) the enforceability of the underlying agreements.15

SUMMARY CONCLUSION
We have mapped out a procedure for evaluating market-value transactions and provided various examples of its implementation. As with all of our rating approaches, we will reevaluate the procedure over time. In particular, we will update our database of asset returns at least annually, and more often should there appear to be a fundamental shift in the environment.16
15 For a detailed discussion of these issues, see “Rating Cash-Flow Transactions Backed by Corporate Debt: 1998 Update,” Moody’s Special Comment, April 1998. 16 At least on a temporary basis, such shifts can also be addressed by adjusting the appropriate stressing factors.

Moody’s Approach to Rating Market-Value CDOs

• 11

We will also extend the database to other asset classes, such as emerging market debt. In fact, there is little in the way of a theoretical limit on the type of instrument that can be incorporated into these structures, so long as the appropriate data are available. It should be apparent from our discussion that the interactions between the various components of market-value structures are complex, so that it is difficult to imagine that anything other than a simulation approach will do justice to these structures. Nonetheless, we hope that we have given some guidance as to the impact of various structural choices on the advance rates. The exhibits that follow should further clarify these impacts.

Exhibit A Impact of Industry/Issuer Diversification on Advance Rates
Table A1

Advance Rates For a Portfolio with 10 Issuers, 5 Industries
(100% investment in one asset type, 5 year maturity)

Asset Type Performing Bank Loans Valued $0.90 and Above Distressed Bank Loans Valued $0.85 and Above Performing High-Yield Bonds Rated Ba Performing High-Yield Bonds Rated B Distressed Bank Loans Valued Below $0.85 Performing High-Yield Bonds Rated Caa Distressed Bonds Reorganized equities

Aaa 0.845 0.730 0.70 0.65 0.52 0.35 0.31 0.22

Aa1 0.870 0.765 0.73 0.68 0.56 0.38 0.36 0.29

Aa2 0.875 0.770 0.74 0.70 0.58 0.39 0.37 0.30

Aa3 0.880 0.775 0.77 0.71 0.59 0.41 0.38 0.31

Target Rating A1 A2 0.890 0.790 0.79 0.73 0.62 0.49 0.43 0.38 0.895 0.795 0.80 0.74 0.64 0.52 0.45 0.41

A3 0.900 0.800 0.82 0.76 0.65 0.54 0.46 0.43

Baa1 0.910 0.815 0.84 0.79 0.68 0.58 0.51 0.46

Baa2 0.935 0.825 0.85 0.80 0.69 0.59 0.53 0.47

Baa3 0.940 0.860 0.87 0.82 0.71 0.61 0.54 0.49

Table A2

Advance Rates For a Portfolio with 20 Issuers, 5 Industries
(100% investment in one asset type, 5 year maturity)

Asset Type Performing Bank Loans Valued $0.90 and Above Distressed Bank Loans Valued $0.85 and Above Performing High-Yield Bonds Rated Ba Performing High-Yield Bonds Rated B Distressed Bank Loans Valued Below $0.85 Performing High-Yield Bonds Rated Caa Distressed Bonds Reorganized equities

Aaa 0.870 0.760 0.76 0.72 0.58 0.45 0.35 0.31

Aa1 0.890 0.780 0.79 0.75 0.62 0.49 0.39 0.37

Aa2 0.895 0.790 0.80 0.76 0.63 0.50 0.40 0.38

Aa3 0.900 0.795 0.81 0.77 0.64 0.51 0.41 0.39

Target Rating A1 A2 0.905 0.810 0.83 0.78 0.67 0.56 0.47 0.44 0.910 0.815 0.84 0.79 0.68 0.58 0.48 0.46

A3 0.915 0.820 0.85 0.80 0.69 0.60 0.50 0.47

Baa1 0.930 0.830 0.87 0.82 0.71 0.62 0.54 0.51

Baa2 0.935 0.840 0.88 0.83 0.72 0.64 0.56 0.52

Baa3 0.940 0.870 0.90 0.85 0.74 0.67 0.57 0.54

12 •

Moody’s Approach to Rating Market-Value CDOs

Table A3

Advance Rates For a Portfolio with 30 Issuers, 10 Industries
(100% investment in one asset type, 5 year maturity)

Asset Type Performing Bank Loans Valued $0.90 and Above Distressed Bank Loans Valued $0.85 and Above Performing High-Yield Bonds Rated Ba Performing High-Yield Bonds Rated B Distressed Bank Loans Valued Below $0.85 Performing High-Yield Bonds Rated Caa Distressed Bonds Reorganized equities

Aaa 0.875 0.780 0.78 0.73 0.62 0.50 0.42 0.38

Aa1 0.895 0.795 0.82 0.77 0.66 0.56 0.47 0.43

Aa2 0.900 0.800 0.83 0.78 0.67 0.57 0.48 0.44

Aa3 0.905 0.805 0.84 0.79 0.68 0.58 0.49 0.45

Target Rating A1 A2 0.910 0.815 0.86 0.80 0.70 0.61 0.52 0.47 0.915 0.820 0.87 0.81 0.71 0.62 0.53 0.48

A3 0.920 0.825 0.88 0.82 0.72 0.64 0.54 0.49

Baa1 0.930 0.835 0.89 0.83 0.73 0.67 0.58 0.53

Baa2 0.935 0.850 0.90 0.84 0.74 0.68 0.59 0.54

Baa3 0.940 0.870 0.91 0.86 0.76 0.70 0.61 0.55

Table A4

Advance Rates For a Portfolio with 40 Issuers, 10 Industries
(100% investment in one asset type, 5 year maturity)

Asset Type Performing Bank Loans Valued $0.90 and Above Distressed Bank Loans Valued $0.85 and Above Performing High-Yield Bonds Rated Ba Performing High-Yield Bonds Rated B Distressed Bank Loans Valued Below $0.85 Performing High-Yield Bonds Rated Caa Distressed Bonds Reorganized equities

Aaa 0.880 0.790 0.81 0.74 0.62 0.50 0.43 0.41

Aa1 0.895 0.805 0.84 0.78 0.66 0.56 0.47 0.45

Aa2 0.900 0.810 0.85 0.79 0.67 0.57 0.48 0.46

Aa3 0.905 0.815 0.86 0.80 0.68 0.58 0.49 0.47

Target Rating A1 A2 0.910 0.820 0.87 0.81 0.69 0.61 0.52 0.49 0.915 0.825 0.88 0.82 0.70 0.62 0.53 0.50

A3 0.920 0.830 0.89 0.83 0.71 0.64 0.54 0.51

Baa1 0.930 0.840 0.90 0.84 0.73 0.67 0.58 0.54

Baa2 0.935 0.850 0.91 0.85 0.74 0.68 0.59 0.55

Baa3 0.940 0.870 0.92 0.87 0.76 0.70 0.61 0.56

Moody’s Approach to Rating Market-Value CDOs

• 13

Exhibit B Impact of Portfolio Limitations on Advance Rates (20 issuers, 5 industries)
Table B1 Table B2

Advance Rates With Portfolio Limitations
(Investment Limited to Performing High-Yield Bonds (rated Ba3 and above) and Distressed Bonds)

Advance Rates With Portfolio Limitations
(Investment Limited to Performing High-yield Bonds (rated B3 and above) and Distressed Bonds)

Target Rating A2 A2 A2 Baa2 Baa2 Baa2

Limitation on Distressed Bonds No Limit Up to 50% Up to 30 % No Limit Up to 50% Up to 30%

Advance Rate For Distressed Bonds 0.48 0.54 0.56 0.56 0.61 0.62

Target Rating A2 A2 A2 Baa2 Baa2 Baa2

Limitation on Distressed Bonds No Limit Up to 50% Up to 30 % No Limit Up to 50% Up to 30%

Advance Rate For Distressed Bonds 0.48 0.52 0.54 0.56 0.60 0.61

Exhibit C Impact of Subordination on the Advance Rates
Suppose, for example, that the capital structure consists of two debt tranches: x% of senior debt and y% of subordinated debt, plus the unrated equity. Two sets of advance rates are normally used to provide protection for (i) the senior debt and (ii) the senior and the subordinated debt, collectively. In essence, two OC tests ensure that (i) the market value of assets discounted by senior portfolio advance rate (AR1) is at least as great as the senior debt, and (ii) the market value of assets discounted by the subordinated portfolio advance rate (AR2) is at least as great as the senior debt plus the subordinated debt. As a result of these two OC tests, the market value (V%, expressed as the percentage of the total capitalization) of the portfolio in the last period in which the OC tests were met will be: x+y x (3) V=max ( AR , ) 1 AR 2 The effective senior portfolio advance rate (ARE1) is then given by,

(4) ARE 1 = max (

x x+y x , AR AR 1 2

)

The intuition behind formula (4) is that, when

(5)

x+y AR 1

>

x , AR 2

the OC test for the subordinated debt can provide additional protection for the senior debt. If equation (5) holds for all possible portfolio compositions, the rating of the senior debt implied by a) the rating of the subordinated debt and b) the relative size of the senior and subordinated tranches, is higher than the rating associated with AR1, the “nominal” senior advance rate. Notice that AR1 and AR2 are portfolio advance rates (see footnote 14) and the above result is highly sensitive to the portfolio composition, as well as the relative sizes of the debt tranches. Given the dynamic nature of the portfolio composition, the positive effect of the subordination has to be viewed in the context of the overall structure. Consider a case where the collateral manager wishes to invest in high-yield bonds rated B3 and better and that up to 50% of the investment is in distressed bonds, if the capital structure

14 •

Moody’s Approach to Rating Market-Value CDOs

dictates that the senior and subordinated debt account for 75% and 5% of liabilities, respectively, and if the ratings sought for the senior and subordinated debt are A2 and Baa2, respectively, then the effective senior advance rate for performing high-yield bonds rated B3 or above (based on Equation (4)) is 0.78, which corresponds to a higher rating (A1) than the desired rating (A2). This does not imply that we should assign a rating of A1 to the senior debt, because this rating applies only if 100% of the portfolio investment is in performing high-yield bonds rated B3 or above. Simulation shows that for other possible portfolio compositions, e.g., 70% investment in performing high-yield bonds and 50% investment in distressed bonds, A2 and Baa2 are indeed the appropriate ratings for the senior and subordinated debt, respectively. If, however, senior and subordinated debt account for 70% and 10% of liabilities and the portfolio limitation assumptions still hold, then the senior rating consistent with the rating of the subordinated debt (Baa2) and the relative sizes of the tranches (70% senior, 10% subordinated debt) is in fact A1, one notch higher than the initially desired rating A2.
Table C1

Implied Senior Rating For Various Subordination Levels
(20 issuers, 5 industries; Investment Limited to Performing High-Yield Bonds rated Ba3 and above; up to 50% in Distressed Bonds)

Senior Debt 75% 70% 65%

Subordinated Debt 5% 10% 15%

Rating of Subordinated Debt Baa2 Baa2 Baa2
Table C2

Implied Senior Debt Ratings A2 A1 Aa3

Implied Senior Rating For Various Subordination Levels
(20 issuers, 5 industries; Investment Limited to Performing Bank Loans and Up to 30% in Distressed Bank Loans Valued Below $0.85)

Senior Debt 85% 80% 80%

Subordinated Debt 5% 5% 7%

Rating of Subordinated Debt Baa2 Baa2 Baa2

Implied Senior Debt Ratings A2 A1 Aa3

Exhibit D Impact of a Minimum Net Worth Test on the Advance Rates
To quantify the effect of a minimum net worth test on the expected loss and advance rate calculation, let us first assume that the test is performed with the same frequency as the OC test. Under this assumption, the value of the assets the last time the OC test was run, relative to the initial capitalization, must be at least x+y+p*(1-x-y); hence, the effective portfolio advance rates for senior and subordinated debt are:

(6a) ARE 1=min (AR 1 , (6b) ARE 2=min (AR 2 ,

x ) , and x+y+p*(1-x-y) x ) x+y+p*(1-x-y)

Let us consider a case where investments are limited to performing high-yield bonds rated Ba3 and above. Assume that we have a structure with 80% senior debt, 10% subordinated debt, and 10% equity and that the desired ratings for the senior and the subordinated debt are A2 and Baa2, respectively. If the minimum net worth test dictates that 50% of the initial equity level must be maintained, then the effective advance rates for senior and subordinated debt are 84% and 88% respectively. Since these effective advance rates are the same as the advance rates
Moody’s Approach to Rating Market-Value CDOs • 15

presented in Table 5, the minimum net worth test does not have a direct impact on the advance rates in this structure. If, by contrast, the structure provides 60% senior debt, 10% subordinated debt, and 30% equity, then the effective advance rates implied by the minimum net worth test are 71% and 82%, respectively, and are therefore lower than the advance rates for performing high-yield bonds presented in Table 5. For this example, a simulation reveals that due to the additional protection provided by the equity and the minimum net worth test, we can increase the subordinated advance rate for performing high-yield bonds rated Ba3 from 0.88 to 0.89.

Table D1

Impact of Minimum Net Worth Test (20 issuers, 5 industries)
(Investment Limited to Performing High-Yield Bonds rated Ba3 and above)

Senior Debt 70% 60% 50%

Subordinated Debt 10% 10% 10%

Equity 20% 30% 40%

Target Rating Baa2 Baa2 Baa2

Advance Rates High-Yield Bonds 0.88 0.89 0.90

© Copyright 1998 by Moody’s Investors Service, Inc., 99 Church Street, New York, New York 10007. All rights reserved. ALL INFORMATION CONTAINED HEREIN IS COPYRIGHTED IN THE NAME OF MOODY’S INVESTORS SERVICE, INC. (“MOODY’S”), AND NONE OF SUCH INFORMATION MAY BE COPIED OR OTHERWISE REPRODUCED, REPACKAGED, FURTHER TRANSMITTED, TRANSFERRED, DISSEMINATED, REDISTRIBUTED OR RESOLD, OR STORED FOR SUBSEQUENT USE FOR ANY SUCH PURPOSE, IN WHOLE OR IN PART, IN ANY FORM OR MANNER OR BY ANY MEANS WHATSOEVER, BY ANY PERSON WITHOUT MOODY’S PRIOR WRITTEN CONSENT. All information contained herein is obtained by MOODY’S from sources believed by it to be accurate and reliable. Because of the possibility of human or mechanical error as well as other factors, however, such information is provided “as is” without warranty of any kind and MOODY’S, in particular, makes no representation or warranty, express or implied, as to the accuracy, timeliness, completeness, merchantability or fitness for any particular purpose of any such information. Under no circumstances shall MOODY’S have any liability to any person or entity for (a) any loss or damage in whole or in part caused by, resulting from, or relating to, any error (negligent or otherwise) or other circumstance or contingency within or outside the control of MOODY’S or any of its directors, officers, employees or agents in connection with the procurement, collection, compilation, analysis, interpretation, communication, publication or delivery of any such information, or (b) any direct, indirect, special, consequential, compensatory or incidental damages whatsoever (including without limitation, lost profits), even if MOODY’S is advised in advance of the possibility of such damages, resulting from the use of or inability to use, any such information. The credit ratings, if any, constituting part of the information contained herein are, and must be construed solely as, statements of opinion and not statements of fact or recommendations to purchase, sell or hold any securities. NO WARRANTY, EXPRESS OR IMPLIED, AS TO THE ACCURACY, TIMELINESS, COMPLETENESS, MERCHANTABILITY OR FITNESS FOR ANY PARTICULAR PURPOSE OF ANY SUCH RATING OR OTHER OPINION OR INFORMATION IS GIVEN OR MADE BY MOODY’S IN ANY FORM OR MANNER WHATSOEVER. Each rating or other opinion must be weighed solely as one factor in any investment decision made by or on behalf of any user of the information contained herein, and each such user must accordingly make its own study and evaluation of each security and of each issuer and guarantor of, and each provider of credit support for, each security that it may consider purchasing, holding or selling. Pursuant to Section 17(b) of the Securities Act of 1933, MOODY’S hereby discloses that most issuers of debt securities (including corporate and municipal bonds, debentures, notes and commercial paper) and preferred stock rated by MOODY’S have, prior to assignment of any rating, agreed to pay to MOODY’S for appraisal and rating services rendered by it fees ranging from $1,000 to $550,000.

16 •

Moody’s Approach to Rating Market-Value CDOs