Print this
The World According to Edward Altman

Edward Altman is one of the pioneers of modern credit analysis. The Max L. Heine Professor of Finance at New York University Stern School of Business and vice director of its Salomon Center, he is the author or editor of 20 books and more than 100 scholarly and professional articles on corporate bankruptcy and credit risk analysis. His latest book, Managing Credit Risk (coauthored with John Caouette and Paul Narayanan), has just been published by John Wiley & Sons.
Altman started out building statistical models—using traditional financial statement analysis and market-value data to project the health of companies. For this work, he received a Laureat 1984 award from the Hautes Etudes Commerciales foundation in Paris. He later focused on measuring and estimating default rates on high-yield bonds, and proposed a type of actuarial mortality risk analysis to estimate defaults. More recently, his research has taken on more general worldwide credit issues relevant to practitioners and bank regulators in the credit risk arena. Altman also serves as an adviser to the global corporate bond research group at Salomon Smith Barney, as a director of the Franklin Mutual Series Funds, and as a consultant to a number of international commercial banks. He spoke with editor Joe Kolman in October.

Derivatives Strategy: What's wrong with the way banks look at credit risk?

Edward Altman: Credit risk analysis has been around for as long as financial statements have been around—and even before that. There are a lot of good people who know how to practice traditional credit analysis. But although commercial and investment banks and other institutions have had some of the right tools, the tools weren't used in a credit culture that could have made their inputs meaningful.

In the United States in the late 1980s, there was a leveraged buyout and recap binge, which resulted in a tremendous increase in highly leveraged transactions. A good credit analyst would have said, “We shouldn't be doing all of these deals.” But what happened in these organizations, and what happened again recently in Korea and Japan, was that greed or growth—the two big G's that can get a bank in trouble—overwhelmed credit analysis.

DS: The credit analysts were overruled by other people in their institutions?

EA: Yes. The same thing is happening again today, to some extent. If there were a vote between three people about doing a deal, and one was a credit analyst and the other two were bankers, it would inevitably be two against one. That whole way of thinking came crashing down from late 1989 through 1991. The debacle helped put credit analysis back on its feet as a legitimate decision tool that should be used to accompany goals for growth and other non-credit-related objectives.

What all this teaches us is that banks can have the most sophisticated credit analysis techniques, but it doesn't matter if greed and other things get in the way. Senior people have to make it clear, by virtue of their decision-making processes, that credit risk has an important voice in the final decision. Then techniques such as Z-Score or mortality analysis—approaches I've worked on for more than three decades—can assist lending and portfolio managers in their difficult jobs.

DS: I think that will sound familiar to readers who have heard the same recommendations for risk management.

EA: I think risk managers should have all the influence they deserve, but there are many examples that show this is still not the case.

DS: You could find plenty of examples in the last few months.

EA: It's certainly clear that people were lending to Long-Term Capital Management without really understanding many of the principles of credit analysis. They were lending to the hedge fund because of a track record and the potential for big returns. You could argue about whether or not those were above-average risk-adjusted returns, but they were clearly large numbers even though the stock market in general was also doing very well. But I don't think credit risk dawned on people. Did anybody ask: “Are we over-concentrating in this one industry or this one firm? Are we making informed judgments based on due diligence, including credit, market and model risk?” I doubt it.

DS: Credit analysts may not have had the power, but did they even have the tools necessary to analyze what the exposures were?

EA: I think they had the tools to do concentration risk analysis. There is still no consensus, however, on the appropriate criteria for measuring the correlation between credit assets. I don't think credit analysts or anybody else really knew what was going on with respect to the risk of the strategies being followed by various hedge funds.

DS: They weren't told?

EA: That's right. So how could they assess what could go wrong? There are value-at-risk techniques for measuring credit asset portfolios. But that implies that there are probabilities, which implies that one knows the underlying process and products. In the credit asset area, for example, the inputs necessary to generate those confidence intervals are the initial rating of the obligation and the consequent expected and unexpected migration of the credit risk. The most negative of these migrations, of course, is default. Once you have default, you need to know the volatility around the expected value of recovery rates. Those are the basic inputs, along with the generation of correlations of risk between the various assets. But those fundamental inputs were not there in the type of hedge fund investments we're talking about.

DS: They didn't exist?

EA: It's impossible for them to exist if you don't know what the strategy is. These institutions were lending to a blind pool. The hedge funds essentially said, “You lend us money and we will decide what to do with the money.” To some extent, that's what happened when money was loaned to the leveraged buyout funds back in the 1980s, in which people invested but didn't know about the deals that were being done. All people knew was how they had done in the past, and that was what they based their decisions on. It is the same with these hedge funds. They had done extremely well in the past, and this implied that they were likely to do extremely well in the future. But perhaps this strategy will now be questioned.

"CreditMetrics and other similar programs were not developed to enhance the intellectual underpinning of financial analysis. they were built to generate trading in assets."

DS: If they can get the same kind of leverage.

EA: That's an interesting point. Will they ever get the same kind of leverage? Financial markets have short memories. But an organization with a credit culture would not have made those loans simply because the senior people would have had to rely on the credit risk manager for input into the decision.

Even though I've worked hard to develop models in this area, I believe they should be a guide, rather than the decision tool. There are some people who believe people should make decisions based strictly on the numbers that come out of these models, but I don't believe so. The models should be one input among other factors, and senior people should understand the quality of the inputs and the sensitivities of the models.

For example, when there's a question of risk concentration, there are a number of variables one could look at relating to the correlations of risk between various assets. One could look at equity prices, models that explain equity prices, the ratings themselves and the variation in ratings over time. But none of these are universally accepted. They can give you an indication of the correlation of migration of risk among various assets and the likelihood that those assets would default at the same time. But I haven't yet been convinced, even though I've worked with these models, that the correlations are stable and consistent—and therefore something one should have confidence in.

I believe this is the Achilles' heel of these models, if there is one. I am convinced, however, that the appropriate measure of risk in a credit asset portfolio is the concept of “unexpected loss.”

DS: What does all this say about the Federal Reserve and the other regulators charged with analyzing the credit and market risk of financial institutions? What kind of job have they done? What do the events of the last few months tell us about their intellectual capacity and the tools regulators have?

"Banks can have the most sophisticated credit analysis techniques, but it doesn't matter if greed and other things get in the way."

EA: Those are good questions. The Fed realizes that it isn't smart enough to dictate the kinds of models that banks should use. When it surveyed what banks were doing, it found a lack of consensus. The people at the Fed believe that models involving concentration risk and other exposures, however, are moving in the right direction. While they didn't dictate what to do, they suggested what banks should consider and added that they were comfortable with what the banks had in place.

DS: They must have known about Chase's $500 million credit line to LTCM, but somehow the risk of that loan didn't get registered.

EA: If Chase has half a billion dollars of exposure, that's Chase's problem. But did the Fed have the systems in place to know that LTCM had worked up the total leverage to all investors to the extent it had? What the Fed did in the last few months was to send a warning signal to banks. It said: “Just because we've had a benign credit cycle, don't let down your standards.” It didn't work. Those standards were lowered. It would have been difficult for the Fed, which is reactive rather than proactive, to have done much more, except perhaps to have better monitored total banking system exposure to individual borrowers and to have aggregated industry exposure.

DS: What you do think of the huge new interest in credit derivatives?

EA: There are two relatively new, separate developments: the securitization of credit assets, and derivatives built on credit events. The underlying goals of both are the same—to reduce or hedge credit exposures and reduce capital reserves.

Credit derivative products have been developed and pushed because they are moneymaking, profitable opportunities for firms that trade, package and build these structures. CreditMetrics and other, similar programs were not developed to enhance the intellectual underpinning of financial analysis. They were built to generate trading in assets, and dealers hoped the trading would be done on their trading desk rather than somebody else's. There was also reputation value in it for the institutions that developed them. Still, I believe credit derivatives serve a useful insurance purpose, in the same way that direct credit insurance does.

On the securitization side, you've had collateralized bond obligations being sold for several years now. These are essentially portfolios of risky and less risky bonds, packaged and resold to the public much the same way that mortgages were sold in the 1980s. From the point of view of the high-yield bond market, the goal has been to provide a continuing supply of purchasers for new as well as existing bonds. It's like a double sale: You sell the original bonds and then you buy them back from the buyers and sell them to the public a second time through the securitization.

These instruments have been attractive because they were already rated by the agencies. Ratings are important in this context not because the agencies always get it right—we know that nobody has the magic bullet in terms of always predicting credit problems. The ratings are important because we have a fairly long and well-documented history of performance of credit assets, classified by their ratings. So I can say, for example, that the probability of a single B-rated bond defaulting in the five years after it is issued is 22 percent, with a certain standard deviation around it.

We can say the same thing about probabilities of migration. Both of these are absolutely critical ingredients to the asset securitization and credit derivatives markets because these data provide the basis for deciding how much to charge as a premium.

DS: Why haven't CLOs been as popular as CBOs?

EA: I think the requisite kind of analysis has been much more difficult to perform for collateralized loan obligations—for all but the largest companies. There have been a whole lot of CLOs out there, but they've been mainly big banks packaging big loans and reselling them to the public. The rating agencies have only been rating loans since 1995, so there's no big database of loan performance—of defaults, migration and recoveries.

The CLO market, however, is potentially a much larger market than the CBO market. In addition, the middle market and the small-firm loans are still relatively untouched, and the main reason is that they are not rated assets.

I believe, however, that artificial intelligence techniques will soon be able to give a “shadow rating” to assets too small to be rated by the agencies. One technique I've worked with involves a neural network approach, which is a computerized method of simulating different models in order to explain human phenomena—which, in this case, are the ratings. I've worked with a firm that can replicate the rating process with 85 percent accuracy within one notch of the actual rating. The rating agencies or other vendors will, I believe, soon be coming out with these models, and this will make small and medium-size loans amenable to securitization. Perhaps it will soon be possible to get a rating over the Internet as well.

DS: You compared collateralized debt obligations to mortgage-backed securities, but it seems that this time Wall Street and the rating agencies have managed to come up with a product that's much more stable than I/Os and P/Os.

EA: I haven't thought of it that way. Mortgage-backed securities may be more unstable because there's so much systemic risk in the mortgage area. They are basically commodities that move together. Although the risks are well-understood, investors also have prepayment risk, which is extremely difficult to predict.

I don't know if I'd agree that these new instruments are less risky. What happens when all the counterparties get hit with a systemic problem? The issuers of these instruments could default, or lose their high ratings, which is almost as bad as a default for an insurance organization lending its balance sheet to an institution buying the credit insurance.

The irony is that credit derivatives were created at a time when credit risk was relatively small. Now they're being tested in a very different and difficult market. Let's see how they perform!