- Shopping Bag ( 0 items )
Fully revised and updated
Here is the only comprehensive source that explains the variousinstruments in the market, their economic value, how to documenttrades, and more. This new edition includes enhanced treatment ofU.S. and worldwide regulatory issues, and new productstructures.
"If you want to know more about credit derivatives—and these daysan increasing number of people do—then you should read thisbook."
—Merton H. Miller, winner, Nobel Prize in Economics, 1990
"Tavakoli brings extraordinary insight and clarity to thisfascinating financial evolution . . ."—Carl V. Schuman, Manager,Credit Derivatives, West LB New York
Janet M. Tavakoli (Chicago, IL) is Vice President of the Chicagobranch of Bank of America, where she directs the company's overallmarketing of global derivatives and manages its CreditMetricsinitiative.
Credit Derivatives Market Overview.
Total Rate of Return Swaps—Synthetic Financing.
Credit Default Swaps and Options.
Sovereign Risk and Emerging Markets.
Synthetic Collateralized Loan Obligations.
Selected Documentation, Regulatory, Booking, and LegalIssues.
Future of the Global Market.
(Please note: Figures, Tables and any other illustrations cited in the following text refer to the print edition of this work, and are not reproduced here.)
Buyers and sellers enter into negotiated credit derivative contracts primarily for two reasons: to manage risk and to earn income. If there is a benefit to booking the transaction off balance sheet from a regulatory, tax, or accounting point of view, this is an added bonus. The buyers and sellers are called counterparties. The terminology can be confusing. A buyer can be a buyer of credit risk or a buyer of credit protection. A seller can be a seller of credit risk or a seller of credit protection. In this book I adopt a convention to avoid confusion. The protection buyer is the buyer of credit protection (a seller of credit risk). The protection seller sells protection (buys risk) and generally receives a fee for this protection.
The global size of this mainly privately negotiated market was estimated to be $100 billion to $200 billion at the end of 1996. The British Bankers Association (BBA) estimated the size of the London market to be $20 billion at the end of 1996. That didn't include transactions done by some Japanese securities firms, which included credit default puts imbedded in privately placed structures; one firm alone had done about $1 billion of this type of business. The BBA's estimate of year 2000 trading volume was $900 billion and the forecasted volume for 2001 is $1.6 trillion.
Because much of the credit derivatives market is off balance sheet and many of these negotiations are private, there is no way of knowing the size of the market unless participants volunteer that information.
The definition of what should be included in market size is also somewhat fluid. Some firms consider total rate of return swaps to be a form of financing. These may be handled in a department separate from the credit derivatives department, which handles credit default swaps. Credit spread puts may be in yet another department. Convertibility protection, if it is traded at all, may be part of yet another group within the same institution.
This is a very rapidly growing market. Many experts thought the market would double between 1996 and 1998. That projection appears much too low. The market probably doubled in the first six months of 1997 compared to the entire annual volume in 1996.
Banks typically increase volume from zero to more than $10 billion inside of one year and accelerate rapidly after that. Rabobank, a relative newcomer to the market, reported $5 billion notional in credit derivative transactions by July 1997. The notional principal amount, usually called the notional amount, is the amount against which fees, interest payments, price differential, and recovery values are based in a credit derivative contract. Union Bank of Switzerland (UBS) went from zero to $4.5 billion notional in transactions in their New York branch's loan group alone for the first half of 1997. UBS estimated total 1997 transaction volume at $15 billion. That figure did not include UBS's emerging market activities or London or Asian activities.
Collateralized loan obligations (CLOs) and collateralized bond obligations (CBOs) are queued in Moody's Investor Services's in box awaiting ratings. The number of deals has increased nearly 15-fold from 1996, although the number of viable deals has only tripled. Most of these deals have a credit derivative component, often equal in notional to the deal size. Synthetic securitization volume has increased 30-fold from the first deal in the fall of 1997.
Total rate of return swaps on loans have increased dramatically in volume as market participants have familiarized themselves with the documentation. Every month new entrants engage in their first transaction.
New participants have entered the credit default protection market in both Europe and the United States, and at least six brokers are now gearing up to service this new swell of business.
Credit derivatives are in high demand because they service an unfulfilled market need. At the moment, the products seem new and difficult to understand. But like the personal computer, it isn't a matter of whether the broader financial community will adopt the new product, it is merely a matter of how fast.
The concepts of portfolio theory and risk versus reward gained popularity early in the twentieth century, and various models were developed in that century to measure risk. The most important concept in investment theory is illustrated in Figure 1.1. This concept, developed by Jack Treynor, William Sharpe, and John Lintner, is known as the capital asset pricing model. It states that the expected risk premium varies in direct proportion to beta, the sensitivity of an investment's return to market movements. This concept can be and has been generally applied since human beings started thinking in terms of money and increasing their amount of money.
The simple graph in Figure 1.1 is the key to all financial management. I deliberately made it simple because, for all of the simplicity of this concept, it is violated time and again in the financial markets. Promoters of Ponzi schemes, "respectable" investment managers, and "sane" banking executives try to promote the idea that one can earn a high risk-free return without risking capital. That's nonsense, and I give limited examples of some of this nonsense and the undoing of a few of the participants in this nonsense in this decade in Chapter 2.
The key to investment management is to minimize risk while maximizing return. In theory, for every risk appetite there is an "efficient frontier" of returns. This is sort of the demilitarized zone (DMZ) of investment management. Below the DMZ one is safe--too safe to win the war against inflation. At the efficient frontier, one is taking the reasonable amount of risk for the desired return. Above the DMZ, the likelihood increases of having your investment shot down and taking a casualty to your principal.
Credit derivatives are a tool to help move the DMZ farther into risky territory without taking more casualties. Specifically, credit derivatives can help diversify the credit risk of a portfolio to dampen the volatility of potential returns. Credit derivatives can help portfolio managers diversify their portfolios. One of the key concepts in risk management developed in recent decades is the concept of diversification of assets. This includes classes of assets such as equities, bonds, real estate, and cash. It also includes diversification of the market risk of various assets through management of the sensitivity to interest rate or currency fluctuations. Recently, the focus of risk management has included management of credit risk in investment portfolios as well. Usually this means diversification through increasing the size of a portfolio, but it can also mean reducing exposure to an obligor while adding exposure to another obligor and keeping the size of the portfolio constant. In any case, the number of obligors or the number of assets increases, although the size of the portfolio itself need not increase.
When diversifying a credit portfolio, it is important to keep the correlation between assets as low as possible. This means that, to the extent possible, credit events should not affect assets in the same way. Credit-quality changes of the two assets should not be related. To the extent that they are related, the positive impact of diversification is diluted. The best assets of all would be assets that have negative correlation. If one asset is adversely affected, the other asset is positively affected by the same event. The covariance--a measure of the behavior of two random variables in relation to each other--should be as low as possible. As the number of assets in a portfolio increases, the portfolio variance approaches the value of the average covariance. Covariance matrices are used in many financial models. Sharpe's capital asset pricing model may be the most famous.
Rating agencies attempt to measure the ability of obligors to meet their obligations. The long-term ratings are based on ability to pay principal and interest and on the likelihood of not defaulting on an obligation. A chart of long-term credit ratings from highest to lowest is shown in Table 1.1. Ratings shown are from Standard & Poor's (S & P), Moody's Investors Service (Moody's), Fitch, Duff & Phelps (D & P), Best's, and Dun & Bradstreet (D & B).
A credit rating below BBB is considered noninvestment grade. In addition to the ratings in the table, the rating agencies sometimes give "+" and "-" or add numbers to indicate additional gradations within classes. For instance, Bank of America's long-term debt is rated Aa2 by Moody's (other choices are Aa1, the highest in this class, and Aa3, the lowest in this class) and AA--by S& P. Obligors rated D are in legal or technical default or have filed a bankruptcy petition. There are other rating services besides the ones listed in Table 1.1, but these are the most common in the United States and are usually a benchmark in the global markets. Japan, for instance, has its own rating agencies, which tend to rate Japanese institutions higher than Moody's or S & P does.
Rating of single securities is not enough to define credit risk. Opinions of ratings may vary widely. Banks, sensitive to credit after the foreign loan debacle of the 1980s, often have their own internal rating system for their clients. The internal rating systems of banks are often much different than those for Moody's and S & P. Furthermore, the rating of a single security doesn't tell us much about the risk of a portfolio of securities and doesn't tell us anything about the correlation of securities in a portfolio. As we shall see in the next section, there are ways to get around some of these deficiencies.
Credit derivatives are not new. Only the hype about them and their classification as a stand-alone product are new. The new crop of credit derivative specialists is quite young, some in their late twenties, which is perhaps why these products are often hyped as brand new. The risk/reward profile of the act may have changed recently, but the fundamentals are the same.
Humans have evaluated credit risk ever since they started bartering goods. Insurance companies have always looked at event and individual credit risks. Banks have evaluated credit risks for as long as they have been in business.
Even the "new" products aren't particularly new. Options on corporate bonds have been traded quite actively since the 1970s, even before there was an option methodology employed by the major Wall Street firms.
Salomon and others have offered total rate of return swaps on mortgage-backed securities since the mid-1980s. Merrill, Lehman, Salomon, and others have offered debt warrants on corporate debt since the mid-1980s. Merrill and others stripped U.S. government risk from Latin American sovereign debt in the mid-1980s.
Even the idea of structuring credit risk is not a particularly new concept. Banks have been doing this since the late 1970s. Some of the most savvy credit analysts went on to make fortunes for their banks with structured credit products in the late 1980s. In 1988 Stephen Partridge-Hicks and Nicholas Sossidas approached Citibank's clients with a unique proposal and launched the Alpha fund, soon followed by the Beta fund. Their management of credit risk was so successful that they left Citibank in the mid-1990s and started Sigma Finance Corporation. Gordian Knot Limited investment management, headed by the two gentlemen from Citibank, is responsible for day-to-day management. Sigma Finance is rated AAA by Moody's, S& P, and Fitch. The commercial paper program is rated A+.
The structure is deceptively simple. Sigma Finance is a corporation owned by institutional investors. The equity is privately placed with well-known institutional investors whose names are kept private. Sigma offers investors either equity (which pays dividends) or capital (which pays interest). These have equal rank. Investors can choose either a 7- or a 10-year maturity. The returns are paid out to investors every six months, and therefore investors hold an instrument that is identical to a floating rate note with a return over the London interbank offering rate (LIBOR). The return gradually builds up over a two-to three-year period and levels out. The equity is similar to subordinated debt. Returns are not volatile, and the equity has a finite maturity. The corporation itself does not have a finite maturity because it can continue issuing equity and/or capital. Sigma can expand to $5 billion in equity capital, if it so chooses.
The fund managers purchase a portfolio of AA assets (lower-rated in-vestment-grade assets are allowed with more limited leverage and more constraints) and issue Aaa/AAA notes in the market. Through this method, Sigma earns a credit spread and takes no market risk. Sigma also uses leverage (8X for AA-rated assets) to accumulate more AA-rated (on average) assets. The investments and leverage are strictly controlled with an optimization program, which equally considers credit quality, maturity, portfolio diversification, asset-to-liability maturity gap, and liquidity eligibility.
Sigma invests in sovereigns, supranationals, general finance companies, high-credit-quality corporates, asset-backed securities, and other high-credit-quality assets. Sigma also buys assets with a monoline insurance wrap from Financial Security Assurance (FSA) and Capital Markets Assurance Corporation (Cap Mac), among others. The average credit quality is high AA.
Gordian Knot does not take market risk, currency risk, or credit risk outside of the fund preestablished parameters. They promise their clients they will do their best to preserve principal and to give them the highest returns possible. They do not promise magical returns either or no risk and high rewards. In fact, they aren't guaranteeing anything. They simply promise to do the best job of managing a portfolio according to well-defined risk parameters. They made good on that promise. The resulting returns allow their clients to sleep well at night while they continually grow wealthier.
Gordian Knot employs a structured credit play. Some of the "new" funds promise flashier returns. There is nothing wrong with higher returns. But they come with higher risks. There is nothing wrong with higher risk, if that is what you want and if you know that is what you are getting. Unfortunately, that isn't always what is represented. We examine some of these structures later in this book. The fundamental principles of credit structures apply. After all the models have had their say, after all the mechanics of the structure have been analyzed, after all graphs have been produced, after all the "guaranteed" returns have been published, the fundamentals still apply. In an efficient market, if you are reaching for higher return, you will take more risk. No matter how much evolution we see in credit derivatives, we have not evolved beyond this fundamental tenet.
There are a few exceptions, mainly in the cross-border credit and cross-border tax markets; but that is due to market or regulatory inefficiencies, and my caveat does not apply to those cases. Later, we examine these exceptions as well.
Models proliferate for looking at credit risk measurement on a portfolio basis. It is a daunting task to try to model credit risk on a global scale.
One of the earliest models for looking at credit risk in the United States was the Altman Z score. A high Z score implied a low probability of default on the part of a potential borrower. The formula was very simple, and a conventional balance sheet held most of the information required. The following is the Altman formula:
Much of this information is not available to foreign borrowers, particularly Asian borrowers, who may use different accounting methods and who do not disclose financial information as U.S. obligors do. For sovereigns, much of the information is not applicable. The Z score is mainly useful only for U.S. corporations.
JP Morgan hopes to define a new industry standard with its CreditManager software, which applies the CreditMetrics methodology developed to examine credit risk. The methodology is to compute the exposure profile of each asset in a portfolio. The model also computes the volatility of value caused by upgrades or downgrades in credit quality or volatility due to defaults. Long-term migration likelihoods are factored into the model. A key component of the model is to also compute the correlation between assets in the portfolio. These results are then boiled down to a one-year time horizon assessment of the value-at-risk due to credit in a portfolio using a mark-to-market framework.
The data are mainly taken from publicly available information. S & P and Moody's have calculated how credit quality is likely to move over time. Table 1.2 shows an S& P transition matrix for corporate debt for a one-year time horizon. The numbers in the table do not necessarily add up to 100 percent because they exclude entities whose ratings were withdrawn or changed to "not rated" due to obligation payoffs or that had insufficient information after a merger or restructuring. Probability of default and recovery rates are similarly estimated. The probability of default and assumed recovery values associated with various ratings are discussed further in the sections on pricing in Chapters 2 and 3.
Correlations between assets are estimated using equity price correlations, where available. Fundamental data analysis or the model user's own data are alternative sources of data for correlation calculations. Data, particularly data used for calculating correlations, are sparse. Data for Asian instruments, emerging markets, and nonpublicly traded data are difficult to obtain.
In the end, the CreditManager software can help portfolio managers identify credit risks according to key parameters. Three of the most useful are absolute size, percentage level of credit risk, and absolute amount of risk. CreditManager has powerful report tools to enable portfolio managers to categorize risk. Figure 1.2 shows a typical CreditManager report.
CreditManager is able to determine the size of credit risk by maturity and by country. The power of this tool is that it also allows portfolio managers to change parameters and view by obligor industry, for instance. One of the more useful graphs shows exposure size versus risk (Figure 1.3). Risky assets and large exposures are quickly highlighted. Large-exposure--low-risk and small-exposure--high-risk assets are also highlighted, as these may be problem areas. Another feature of this graph is that it shows the portfolio data; and if one were to draw a curve that captures most of the data points (other than the high-risk and large-size exposures), one could define the boundary condition of credit risk for this portfolio. If that boundary is consistent with the risk philosophy of the portfolio manager, that is fine. If the boundary suggests that the portfolio manager assumes too little or too much risk relative to the investment philosophy, changes can be made to shift the portfolio risk boundaries.
CreditMetrics and CreditManager are tools, but they are not the answer. As a means of checking up on the status of a portfolio, they can be very useful tools indeed. But there are caveats.
Franklin Lee, head of the corporate bond proprietary-trading desk at Merrill Lynch, has this to say about CreditMetrics and statistical models in general:
At first glance, it looks like they use the statistical method for risk control, which I don't feel very confident about. Statistical methods that have been used in the past have never worked. However, the section on transitional probabilities is relevant in valuing credit-spread levels.
Statistical models do have their shortcomings. We apply a model to describe data; and if it looks nice on the graph, we feel we have done our jobs. But when someone shows me a probability distribution of defaults or credit spreads, my first question is: "Says who?"
The key to CreditMetrics, as in most statistical models, is dealing with the probability distribution. What is the best way to deal with skewed (asymmetry of data) distributions and kurtotic (bunching or dispersion of data relative to a normal distribution) irregularities? Simply stated: How much is the credit and value of the security/securities really likely to change? Gaussian (normal) and even lognormal distributions are merely mathematical representations designed to attempt to answer this question. There is no magic about either of them. If we know the exact probability for certain, we don't need to use a Gaussian or lognormal or other mathematical distribution. We could simply use the raw data as is. If we know the exact distribution for certain, we don't need a model. Everything else in the model is mere mathematical manipulation of the result of the initial probability assumptions, and the model rises and falls on these initial assumptions.
When prepayment-probability distribution assumptions failed to reflect reality in the mortgage portfolio, Wall Street firms with sophisticated models took massive hits to their profit-and-loss statements (P & L). The memory needed to store historical data and programming code put the Pentagon's Cray computer to shame. The models were impressive and sophisticated, but the underlying probability distribution assumptions were fatally flawed. Mountains of historical data were precisely manipulated to give inaccurate results, on which people relied as if the models were a religion. Firms took massive losses to their mortgage-backed securities (MBS) trading portfolio as a result of inadequately hedged positions--a classic case of more precision than accuracy.
Nonetheless, historical data provide a good place to start; and with some bonds or loans, there will be more history than others. But what about the future? Historical data are not always good at predicting the future, historical data do not incorporate news of current events, and credit risk does not lend itself to a Monte Carlo--type analysis in the way that interest risk does.
There is a market instrument, however, that incorporates current information: options. Puts and calls on underlying securities incorporate views of the future credit--influenced price moves of securities. For instance, if we want to construct a probability distribution of the terminal value of a security due to interest rate and credit moves, we could canvas the market for a menu of puts and calls in, at, and out of the money for the relevant time horizon. From this we can not only back out implied volatilities, but also check the market view of the implied probability distribution of the payoff outcomes of the security and, from that, back out the market-implied probability of credit migration for the relevant time horizon. This method is a good "reality check" for the results coming out of a model that uses only historical data. Price discrepancies should be due to market views of the influence of current events on the credit migration of the security above and beyond that predicted by the historical data. In an efficient market, the market should price all of this information into the option premiums. Of course, the market is not really efficient, due to information and interpretation dislocations. So even the combined method won't be perfect. But that is true of any security we price, including U.S. Treasury bonds, albeit with one less factor--the credit factor. If there were enough options available, modelers could spend happy weeks using this sort of methodology.
More simply, knowing the spread of a "pure" floater would also be very valuable. If a floater resets instantaneously, the element of price fluctuations due to interest-rate moves drops from the equation. Longer resets incorporate larger interest-rate-influenced price effects. The price move of a floater incorporates a lot of credit information, but even this market has glaring inefficiencies and inconsistencies, especially with emerging market debt. In any case, it is much easier to examine the history and the current pricing in this market than it is to try to model spreads.
I am not suggesting here that portfolio managers should abandon models. In fact, I am a proponent of CreditMetrics and CreditManager. Although portfolio managers must use judgment inputting the data into CreditManager, JP Morgan has come up with an effective snapshot of portfolio credit risk in this model. JP Morgan doesn't claim or represent that the snapshot is a moving picture. Models are best when they are sorting, organizing, and manipulating data. Models cannot exercise judgment. JP Morgan's model does a good job of massaging the data so that portfolio managers can exercise better-informed judgment.
In 1989, Stephen Kealhofer, John McQuown, and Oldrich Vasicek founded KMV Corporation to focus on corporate credit-risk measurement and management. KMV Corporation's credit risk model is another valuable tool. KMV has done considerable fundamental analysis on credits and on correlations between credits. KMV defines the benefits of "cutting off" weaker credits (as defined by KMV) of same-rated securities. An investor can dramatically reduce the risk of default in a portfolio of same-rated securities. The strength of the KMV model is the enormous potential benefit to improvement in portfolio quality at no additional cost.
This doesn't violate the efficient market statement made earlier, however. KMV is successful because the market does not have access to this kind of detailed information because of the barriers to obtaining it. The methodology can be cumbersome, and skill and judgment are needed for some of the assessments of a corporation's balance sheet strength. The model also does not lend itself to the analysis of sovereign risk without the necessity for the modeler to make some crude assumptions. KMV is less accessible to portfolio managers other than large banks because it is very expensive. Despite these weaknesses, it is an impressive portfolio management tool for modeling credit risk and possibly the best technology and data currently available in the credit markets.
The ideas put forth by JP Morgan are not new. The concept of portfolio diversification has been around since the 1970s, with some of the best work being done by Roger Ibbotson and R. Sinquefield. The fundamental concepts remain exactly as they applied them. Diversification works because prices of different credit securities do not move exactly together. A decline in one credit may be partially or completely canceled out by a rise in another. The risk that can be potentially eliminated by diversification is called unique risk, or specific risk. This name stems from the fact that many of the risks that pertain to a specific credit are peculiar to that credit. There is risk that cannot be mitigated by diversification. That is market risk. No matter how many credits one holds, there are global economic risks that can threaten all credits or clusters of credits, and investors will always be exposed to market uncertainties.
Few models can combine credit risk and market risk. Neither CreditManager nor KMV attempts to do so. Algorithmics has developed a model, Regret, that provides a framework for risk management at the enterprise level. Algorithmics is the only company in the business to get three stars from Meridian Research. The model links market and credit exposure and has complex netting hierarchy. One can build an infrastructure to simulate a forward moment in time for the value of a portfolio. Regret prices the cost of a put option on the net portfolio exposure. It minimizes the price to hedge the net exposure.
KMV's models can be used within the Algorithmic framework. Algorithmics provides the infrastructure to stress test, create synthetics, and use the Regret option.
Regret and the models of a few other new firms, which plan to compete with Regret, are new on the scene. No one actually uses them as trading models yet. They are time-consuming and difficult to implement properly and not close to a near-term industry standard. As we shall see in Chapters 2 and 3, market professionals presently retreat to the fundamentals of relative value.