CreditMetrics-DOs and DON'Ts
When RiskMetrics came out in 1994, many people seized on the idea as
a general panacea for all their risk problems. Managers rushed to make sure
they could run RiskMetrics-style Value-at-Risk-sometimes before they identified
whether Morgan's covariance methodology was really appropriate for their
portfolios. Software vendors rushed to have "RiskMetrics compatible"
Now that CreditMetrics is on the scene, two questions arise: How can
I implement CreditMetrics effectively? And, What can-and can't-CreditMetrics
and CreditManager, Morgan's CreditMetrics software program, realistically
do for me? (For a description of the CreditMetrics methodology, see "JP
Morgan Tackles Credit Risk," May.)
- Carefully review the CreditMetrics methodology.
CreditMetrics is much more complex than RiskMetrics, and therefore
demands a careful review. Indeed, within the CreditMetrics framework, users
are faced with a variety of choices. For example, CreditMetrics allows
users to adopt one of four different approaches to estimating correlation
among various credit types-historical data, bond spreads, equity correlations
or uniform constants.
There are pros and cons to each method. The method using historical
rating and default data represents real events, but, because credit events
are rare, this data is sparse and uneven. Bond spreads are readily available
for some securities, but they are not a pure reflection of the market's
opinion on the underlying credit. Equity prices are liquid and provide
market information, but they may not be as useful in illiquid and private
markets. The uniform constant is extremely easy to implement-and is reasonably
accurate in the case of a well-diversified portfolio-but it certainly does
not capture the entire picture. Choose carefully because your decision
could have a big effect on the outcome.
- Decide whether CreditMetrics is right for you.
According to Mac McQuown, a vice president at San Francisco-based KMV,
CreditMetrics is a quality credit-risk-measurement tool, but it does include
certain simplifications. For example, he explains that using commercially
available credit ratings, such as Moody's and S&P ratings, to evaluate
credit quality can be imprecise. Instead, he explains, some financial institutions
may wish to create more customized credit risk weights-as well as their
own proprietary "transition matrices" for determining the probability
that these credit weights may fluctuate. Although CreditMetrics is flexible
enough to handle proprietary credit ratings, a highly intricate setup might
be better handled through a more complex methodology.
- Use CreditMetrics as an education tool.
According to Blythe Masters, one of the principal architects of CreditMetrics,
one of the goals behind the product's release is to raise the level of
dialogue on the topic of credit risk management. Even if CreditMetrics
is not a perfect fit with your operations, the well-documented methodology
is an excellent guided tour of quantitative credit risk management and
thus an appropriate training tool.
- Consider getting CreditManager.
Credit risk measurement is widely recognized to be much more difficult
than market risk measurement. At $10,000 to install and $25,000 for annual
maintenance, most software experts we contacted rate the CreditManager-assuming
its functionality even comes close to the hype-as a steal.
- Don't add CreditMetrics and RiskMetrics numbers together.
CreditMetrics generates a Value-at-Risk for credit risk and RiskMetrics
generates a Value-at-Risk for market risk. So, you might think, wouldn't
adding the numbers together give you a total risk measurement for both
your market and credit risk? Not necessarily. Because there are correlations
between market movements and changes in credit exposure, the sum of a purely
market-based Value-at-Risk and a pure credit Value-at-Risk could substantially
overstate-or understate-your risk, depending upon whether your credit and
market exposures are positively or negatively correlated.
- Don't confuse the Credit Exposure Calculator in FourFifteen with the CreditManager.
In order to help clients calculate the counterparty exposure associated
with "market"-based instruments such as interest rate swaps,
foreign exchange forwards and so on, JP Morgan has built a credit exposure
calculator into its FourFifteen product, which allows users to retrieve
this credit exposure from their market-based instruments. This credit exposure
can then be lugged into a CreditMetrics calculation. But the two are not
- Don't wait for JP Morgan to integrate market and credit risk.
For some institutions, it is critical to begin integrating credit and
market risk management as quickly as possible. While CreditMetrics-as a
companion to RiskMetrics-is a considerable step toward creating an integrated
framework for both credit and market risk management, it may be quite some
time before total integration is complete. According to Ron Dembo, founder
of Algorithmics, a risk management software and research and development
firm, a risk management methodology must have grounding in solid financial
theory if it is to be easily extensible. He suggests that the reason why
JP Morgan-and other purveyors of risk measurement methodologies-may have
some difficulty creating a combined credit and market risk measure is that
they have addressed both credit and market in an ad hoc fashion, rather
than as components of a larger, more holistic theory.