Print this


Till Guldimann
vice chairman, Infinity, a SunGard company
Joe Kolman
editor, Derivatives Strategy
Kelly Kremm
vice president of emerging markets,
HSBC Markets
William Margrabe
president, William Margrabe Group
Diane Petan
sovereign fixed-income analyst,
Goldman Sachs
Philip Prince
vice president of global credit derivatives and head of synthetic securitization,
Chase Securities
Leslie Rahl
principal and cofounder,
Capital Market Risk Advisors
Bob Strong
executive vice president and chief credit officer, Chase Manhattan Corp. and
Chase Manhattan Bank
Jennifer White
director, Aon Financial Products
Gregg Whittaker
managing director and global head of credit derivatives, Chase Securities

Joe Kolman: The derivatives world is based on statistics, not on fundamental or qualitative assessments of how a particular country is doing. Yet we've heard many people say that the credit models used in Asia didn't prepare us for the huge economic dislocations that occurred. How can we improve these credit models to help us predict future crises or to assess the creditworthiness of different countries? Do we have to reconstruct them from the ground up?

Bob Strong: I think you'll find one of the most compelling issues in credit models when you look at foreign exchange and some of the estimates of exposure on some of the basic derivative products. It's clear that there was a big difference between the VAR analysis and the stress-test analysis. Far too many people predicated everything they did on VAR analysis. When you have the rupiah at 2,400 with creeping devaluation every year, no one stressed it to 16,000 to the dollar.

If you were to redo your models to the stress periods of the last year or so, you really wouldn't be doing any business for the next couple of years. How do you position yourself between the extreme stress event and the traditional VAR event? That's a real challenge for everybody in the market—balancing being active in the market vs. the real view of risk. Some banks ran into situations where their losses with some clients on foreign exchange were a multiple of what the approved exposure had been. The basic lesson we've learned at Chase, and that we're applying throughout our risk management disciplines, is that stress-testing is a much more critical element. We put it on an equal plane with VAR analysis today, elevating it from second place.

Leslie Rahl: I agree completely with Bob's comments on the importance of stress-testing, but I guess I would go one step further. I think people are really making a mistake in not also stress-testing their VAR models. Many people found out that their exposures with many counterparties were multiples of what the credit department thought they should be. Unfortunately, most stress-testing has been done in the market risk arena, rather than applied to the potential credit exposure. I would contend that the models aren't nearly as bad as they might be portrayed, but that the assumptions being put into the models as well as the stress-testing of those assumptions have not received enough attention.

Till Guldimann: I think we are in a new world. We're selling assets to a new class of investors. They look at current market values constantly and believe they can always get out. When too many of such investors hold the same assets and all of them want to get out the door at the same time, huge price distortions occur. These price distortions are extremely hard to model.

“If you were to redo your models to the stress periods of the last year or so, you really wouldn't be doing any business for the next couple of years.”

You can always talk about stress-testing, but new price distortions hardly ever look like the stress scenarios you imagined. Still, you can't simply say, “Let's forget about stress-testing.” Liquidity, or the potential lack of it, is of inordinate importance in a world where more and more investors believe they can always get out. But the effect of disappearing liquidity is not reflected in most modeling. I think we have to look at the increasing dependence on liquidity and figure out how we can model that in estimating price risks.

Jennifer White: I was in Singapore until last April, and one thing that we've seen since then is a recognition of the divergence between a sovereign credit and a corporate. The typical thought, a year or so back, was that a sovereign agent would be able to support many of these corporates. Not just Korea supporting Korean Development Bank, which is an obvious one, but the next tier down. Consequently, a lot of people were hedging corporate exposures with sovereign credit derivatives.

We're now seeing those spreads widen. The spread between Korea sovereign and KDB default protection rose from about 20 basis points to close to 100 basis points a couple of weeks ago. You see the same thing happening in China between protection on the sovereign and protection on the Bank of China. It seems to me that the sovereign spreads are probably still too high. Other than Russia, we haven't seen a real default at the sovereign level, and I question the accuracy of the prices of sovereign default protection. Will a sovereign actually default on external public debt? I don't know that we've really answered that question.

Kelly Kremm: During the crisis, there was a continual flow of money through the emerging markets. I think part of that flow could be viewed as a play among investors to construct a “virtual” credit derivative, based on some assumed sovereign indemnity. The assumption was that a government would either be bailed out by some supranational or the G-10, or that certain corporates would be supported by their sovereign governments. You continued to see that play until recently in the case of GITIC in China.

Another thing to realize is that when you are pricing credit derivatives in some markets, certain technical factors come into play that are not easily accounted for in a model. If you look at sovereign China risk right now, it's trading maybe 250 basis points over Treasuries. To many players, it has always seemed tight. Part of the reason it trades so tight, and why it stayed relatively tight even during periods of high stress in Asia, is that it is extremely hard to borrow this bond, because the repo market for many Asian bonds is less developed than others. Therefore, natural shorts are forced out or discouraged, based on technical factors. These sorts of situations are going to create distortions in risk models.

Kolman: It's beginning to seem to me that it's going to be difficult or impossible to create a model to factor in all the intangibles as well as deal with all the limited statistical data. Am I a pessimist or should we just give up on this whole project?

Strong: I think we should not ignore the fact that models are models and that judgment needs to be brought to bear in using these models. I think we can improve the models, and we're beginning to do just that.

We're getting increasingly comfortable as we go through these periods. For example, we did a stress-test a year ago on our mark-to-market portfolio, based on a scenario that played out quite accurately over the last couple of months. Our results were good for the firm. And we felt pretty comfortable.

The one thing we missed in a flight-to-quality scenario is that we assumed there would be a strengthening of the dollar. We didn't see the dollar down against the yen in that kind of scenario. So now we've learned from that, and we've added that as another element.

At the end of the day, models provide parameters around which we try to manage the business. We don't use them to manage business on a stand-alone basis. I think a lot of progress has been made and we're a lot closer now than we've ever been, and these models will continue to be refined and improved.

Philip Prince: There's another fundamental assumption when you talk about this problem in terms of models. The assumption is that the outcome in a credit is 100 percent the result of exogenous variables. I think there are an awful lot of people in the credit business who would disagree with that statement quite strongly. When you're talking about a company, the behavior of the creditors has an impact on what happens in terms of recovery and default rates. That's the fundamental question that still needs to be answered in this product set. We talk about it elliptically. For example, earlier we acknowledged that all the passive mutual fund investors in Asia had an impact. Would it have been much different if there had been a manageable group of people who could have worked out the situation? That's something I'm not sure one can ever model.

Petan: Although I am the analyst responsible for sovereign credits, I have always thought it necessary to incorporate a credit view on the banks and corporates as much as possible. Presumably, since I am from an investment bank—and considering the strength of our relationships—one might assume that I have access to reliable information. But I can point to many instances in which information presented to me by banks and governments proved to be incorrect. I have since learned that many of these entities did not have the proper information themselves. They did not have the infrastructure in place to collect reliable data from their banking system on the degree of risks and nonperforming loans, and were thus plagued by structural weaknesses. In many cases, officials were giving me data they may have believed to be true, but that came from the limited sources they had.

“You can always talk about stress-testing, but new price distortions hardly ever look like the stress scenarios you imagined. The effect of disappearing liquidity is not reflected in most modeling.”

I'd like to add another point: I turned negative on Korea quite early, but it was not based on my typical approach to analyzing sovereign fundamentals, or on any models I had. I simply looked at the Korean equity market in 1996 and was alarmed by the steady decline. This prompted me to look at the equity analysts' reports that discussed Asian bank fragility. They revealed that the banking systems of Korea and Thailand were among the weakest in the region. I thought to myself, “How can you have a strong country if your banking systems are this fragile?”

I couldn't support my negative view with Korea's sovereign variables, since they seemed to be strong—as was evidenced by Korea's entry into the OECD in December 1996. As a result, many people didn't want to believe me. So the moral of the story is: look at as many variables as possible. Even if you are a fixed-income investor, look at equity research on the corporates and the banks, and always review a sovereign credit with a healthy degree of skepticism.

Gregg Whittaker: Joe, you've made the suggestion that maybe the models weren't any good. I would suggest that maybe it's the interpretation that we apply to the models. As Jim mentioned, you can make the case that in certain circumstances things happen that we never envisioned. But in other circumstances, including certain models that cover this example, I'm sure that the models could very well have been right—and on an expected return basis everything was appropriate. The real problem was in the interpretation—or the misinterpretation. While expected returns are fine, what's going on in the tails? What is the unexpected return? The information may have been there and may have been apparent, but we didn't put as much stock in the tail risk as we should have.

William Margrabe: I think we ought to keep in mind one expression: In the land of the blind, the one-eyed man is king. I think it's important to have models because the people with pretty good models will be in really good shape, compared with the other people.

The second point is, it's a very Darwinian world. Institutions will actually die from their mistakes. It may take some time for a shake-out to occur, but I think we're now seeing some institutions suffering and maybe even going through death throes. Ultimately, the market will decide who's got the best models.

But there's another element: When the models are inadequate, it could be that they're missing a random factor or it could be that there's a moral hazard element going on. And I would say, generally speaking, that the people who trust in models too much miss those moral hazard elements. They're the ones who are really going to get nailed—and consistently. The people who miss elements that are simply market factors will last a little longer. And the people with the best models are going to last the longest.