Robert Mark is chief risk officer of the Canadian Imperial Bank of Commerce. Since joining the bank in 1992, he has taken on increasing responsibilities and is now in charge of monitoring risk throughout the bank's operations, including both its trading book and its banking book. Before joining CIBC, he served in risk management roles at both Coopers & Lybrand and Chemical bank. In 1998, he was named Risk Manager of the Year by the Global Association of Risk Professionals. He was interviewed by editor Joe Kolman in August.

Derivatives Strategy: How has risk management changed in the last few years?

Robert Mark: Our ability to analyze risk has become a lot more sophisticated, integrated and pinpointed. Take the risk of a corporate bond in the trading book, for example. This involves a lot more than simply considering the direction of interest rates. We not only look at the market risk of that bond, but we also look at the credit risk. The analysis involves examining credit-rating migration, default probabilities, liquidity and other factors. Then we try to aggregate the market and credit risk components to arrive at the price risk of the corporate bond.

DS: You are really looking at a holistic view of market and credit risk together. The old-fashioned way would be simply to calculate the market risk of that bond.

RM: Exactly. Suppose I have a single-A corporate bond. We ask ourselves, What's the chance that it will migrate down to a triple-B or up to a doubt-A? What's the chance that it will go into default? And how do I bring all that information together to calculate the price risk of the bond?

To do that, I need a database of accurate and pertinent credit information. As I've mentioned, we need to have probabilities of a credit migrating from its current grade as well as the probabilities of the credit going into default. We need to know the recovery rate of the corporate bond. We need to know how the bond is impacted by the rise and fall in risk-free rates. All this allows you to slice and dice the risk of the corporate bond and to gain a better understanding of the overall risk.

DS: Since 1998, banks have been able to use their own internal risk measurement models to determine the minimum amount of required regulatory capital in their trading book. What kind of analysis does your regulator ask before it approves your own internal models?

RM: Regulators analyze how you calculate risk in six different categories. The first two categories involve calculating foreign exchange and commodity risk. That's relatively straightforward. The third and fourth categories involve calculating two components of equity risk. Regulators ask you to split risk in your equity portfolio into general market and specific equity risk. If the price of your stock portfolio moves in perfect correlation with the general market, for example, all your risk is general market risk. But if the price of your stock portfolio moves independently of the general market, then you have specific risk. The last two categories involve interest rate risk. Regulators want you to calculate general market risk as well as specific credit risk for assets with interest rate risk.

DS: What are the benefits of BIS 1998?

RM: Actually, I see the benefits of in six dimensions. The first benefit is in what I call the carrot and stick dimension. If your regulator approves your internal models, you are able to obtain a regulatory capital advantage over and above a standardized approach. That's the carrot. But if you don't have sophisticated models, you don't get that advantage. That's the stick.

Another key dimension is to bring managerial or economic capital closer to regulatory capital. The BIS rules encourage this in a number of ways. The amount of capital they charge is a function of your own internally generated value-at-risk, for instance.

"If your regulator approves your internal models, you are able to obtain a regulatory capital advantage. That's the carrot. But if you don't have sophisticated models, you don't get that advantage. That's the stick.”

The third dimension is in the understanding of the risk dimension. As I noted earlier, you don't calculate just the market risk of a corporate bond. You calculate both the pure interest rate risk and the credit risk.

The fourth dimension is the risk transparency dimension.

The fifth dimension is in the internal discipline dimension. For example, BIS requires a back-testing discipline. VAR is calculated at a 1 percent loss tolerance level. That means that the amount of risk you declare in your VAR calculation can only be exceeded once in every 100 days. Statistically, you would expect your trading related revenue to exceed VAR two and a half times during the year. If you exceed it five times, for example, you get a capital penalty through the raising of your multiplier.

And finally, the sixth dimension is the risk analysis dimension. BIS expects that the regulatory capital numbers will be analyzed not just by risk management, but by the traders and senior managers.

DS: What goes on in other banks in terms of your fourth dimension, the transparency dimension?

RM: I'll tell you what goes on in our bank. Every morning for half an hour at 7:30 A.M., we review the risks in half an hour in the trading book through a global hookup. The heads of each of our trading businesses utilize the risk reports describing the risk taken the prior day to talk about the risks and their going-forward risk.

Every Tuesday morning, I review those same risks with our senior executive team, called the SET. I describe the risks for the week by walking through specific key positions and associated risks by asset class. We also have a capital and risk committee which meets weekly to review risk and capital related issues. We use our day-to-day risk measures to drive our economic capital. So from the top of the house to the bottom, we have extensive communication on the risk-and-return dynamics, which makes our trading and gap management strategies highly transparent.

DS: How is analyzing the risk of the trading book different from analyzing risk in the banking book?

RM: A transformational approach to managing risk is taking place in the banking book. We are seeing techniques borrowed from the trading world being applied to the banking book. VAR is now being applied to the banking book to measure credit risk, for example. The industry calls this approach credit VAR. We can use it to calculate how much money will be at risk with a 5-basis-point probability of loss broken down in a variety of ways. That's a new way of thinking.

Banks have traditionally relied heavily on their own internal risk rating systems. A "one” is the best rating corresponding to, say, a triple-A, a "two” is the second-best, corresponding to a double-A, and so on.

We're also more sophisticated about calculating portfolio effects across the banking book. We calculate the correlations between those loans. If we have a portfolio of loans in several industrial classes, we can calculate the correlation across these asset classes. That helps us understand the concentration risks in the banking book. We also formed a new credit portfolio management group, which actively looks for opportunities to sell down our loan exposure

DS: At the moment, however, the BIS hasn't quite caught up with these innovations.

RM: BIS 2000 plus is coming along. Last year, Basel distributed a series of consultative papers asking banks for their views on proposals to improve the calculation of regulatory capital.

We have relatively simple rules associated with BIS 1988. The current BIS rules fail to capture the actual risk. The rules do not differentiate by risk rating. For example, a $100 triple-A loan is charged the same capital as a $100 double-A loan. The rules do not capture portfolio effects. For instance, the current BIS 1988 rules say that if you make a corporate loan for $100 million, you'll be charged 8 percent or $8 of capital for that loan. If you make 100 loans of $1 each, you will also be charged $8. An unfounded revolver less than one year does not get charged any regulatory capital.

"The same analytic power that went into analyzing market risk on the trading floor is now being applied to the banking book for large corporate, mid-market and consumer loans.”

We can forecast where BIS is heading, even though the final BIS paper hasn't been published. It looks like the BIS will allow banks to use their own internal risk-rating systems. A bank with a well-designed internal risk rating system that allows it to effectively differentiate loan quality will be able to arrive more accurately at the amount of regulatory capital charged as a function of the credit quality of the loan.

As I understand it, the BIS is scheduled to publish its next consultative paper in the first quarter of 2001. Then it will most likely allow roughly six months for comments. Hopefully, we'll have something final by the end of 2001.

DS: What are you hoping for?

RM: I expect that the new BIS rules will be a leg up in sophistication. The new rules will result in a minimum required regulatory capital charge that will be more aligned with the actual risk, as opposed to some simple formula that does not appropriately differentiate risk. Sophisticated risk-literate banks will be implementing credit VAR measures that allow them to calculate the amount of capital as a function of risk. Sophisticated banks will be able to gain a regulatory capital advantage over less sophisticated banks and hopefully this will encourage further investment in building best-practice risk management systems.

DS: Does all this mean we've solved the basic problems associated with market risk?

RM: There is still a lot more to do in terms of upgrading the degree of sophistication in such areas as stress testing and dynamic VAR. Dynamic VAR seeks to capture liquidity effects over a prescribed time period. If the market becomes illiquid, then we can't roll out of positions as fast as we would otherwise be able to. Banks need to build certain trading rules into their risk systems to capture liquidity effects. The systems need to generate a lot of simulation paths and each simulation path would have a series of trading rules associated with it.

There are also problems with how you report VAR associated with the trading book. VAR is typically utilized as an overnight statement of the amount of risk you have in normal markets. Stress testing describes what happens in abnormal markets. A critical question to ask yourself is, How much can I lose over a month or over a quarter? If you're trying to articulate risks to senior management, rating agencies, equity analysts or your board, then you want to tell them not only how much is at risk from an overnight point of view but how much is at risk in a quarter in both normal and abnormal markets. In other words, we want to generate a term structure describing our risk profile. The amount of risk can look quite large over a quarter in abnormal markets, but relatively small on an overnight basis in normal markets.

DS: But for you anyway, the real challenge seems to be looking at the banking book.

RM: A new burst of energy is being applied across the financial industry to the credit side. The same analytic power that went into analyzing market risk on the trading floor is now being applied to the banking book for large corporate, mid-market and consumer loans.

"If you're looking at the default of, say, a double-A loan, you'll find the data are very sparse in the tails. How do you back-test it if you don't have significant data history?”

On the retail and small business side, for example, our goal has been to implement increasingly sophisticated and integrated credit scoring routines. The challenge is to obtain sufficient credit information on a timely basis. If a customer walks into a branch and wants to purchase several products, you want your credit scoring system to be based on a single integrated customer information file. In the new bank-in-a-box world, a bank acquires customers through remote electronic channels. You don't have the same familiarity with the customers as you once had in your local branch.

DS: Are there any problems associated with aggregating all these different types of risk?

RM: We now know how to calculate correlations between assets in the large corporate and commercial loan books. We can in practice do the same type of portfolio aggregation that we now do in the trading book. Yes, the liquidity is different, and time horizon is different, but from a formula perspective we can aggregate them in the same way. Of course, if I'm aggregating the risk of my consumer loan book, my small-business loan book and my large-corporate and commercial loan books, the aggregation issues become increasingly tougher.

DS: Simply put, it's apples, oranges and pears. What do you do?

RM: The common denominator is credit VAR. Prior to VAR, one would have an apples, oranges and pears approach because one would measure everything very differently. For example, risk in the banking book would be measured by the notional amount of the loan, risk in the trading book by VAR, risk in the consumer book by certain cut-off credit scores. Now, our energy is being placed on translating everything into credit VAR terms. The significant breakthrough is that we know how to calculate correlations across various asset classes and aggregate them in a way that gives us a portfolio VAR.

DS: What's the weak link in this whole methodology?

RM: Data. To do this well, you need to get the right amount of data.

DS: Why aren't we getting the right data?

RM: There is much less data available than there is in the market risk area. For example, you can't go back 20 years and get a robust, statistically significant history across the commercial loan book that will allow you to say, "Ahh, this is the data that I'm going to use to calculate the credit VAR in my commercial book.” You may have a 10-year history of data, but the sample size in terms of statistical significance in the tails would be less than desirable.

If you're looking at the default of, say, a double-A loan, you'll find the data are very sparse in the tails. So how do you describe the complete loss distribution of a double-A loan? How do you back-test it if you don't have significant data history?

There's another big challenge associated with credit VAR. The calculation of credit VAR requires that you deal with an asymmetric and long-tailed distribution, in contrast with market VAR, where it is assumed that the distribution is symmetric.

I believe we can break through the apples-and-oranges problem if we have access to sufficient data. I am quite optimistic about the data challenge with the arrival of BIS 2000+, the appearance of price discovery driven by the appearance of active credit portfolio management groups, as well as the increasing utilization of sophisticated analytics in credit decisioning. The evolution of an overall integrated risk management process will lead to the development of increasingly robust credit data and credit analytic measurement tools.