Integrating Interest Rate Risk and Credit Risk in Asset and Liability Management
by Robert A. Jarrow of Cornell University, and
Donald R. van Deventer of Kamakura Corporation
December 28, 1998
Introduction: A recent study by the Federal Reserve [REF] came to the startling conclusion that no interest rate risk variables were statistically significant in predicting bank failures in the United States. The new FIMS monitoring system discussed in this Fed study predicts failure based on 11 key variables collected from the Report of Condition submitted to bank regulators in the United States. Five of the eleven variables are related to the riskiness of commercial lending, and none of the others are related to interest rate risk.
The demise of much of the savings and loan industry in the early and mid-1980s when interest rates were high and volatile certainly suggests that interest rates should be a significant risk factor. A cynic might argue that the regulators did not collect a meaningful interest rate risk measure from reporting banks until very recently. Nonetheless, the FIMS research is an indication that many market participants, including various software vendors, seem to believe that credit risk can be analyzed without the consideration of interest rate risk.
In stark contrast, the traditional approach to fixed income analysis (see Fabozzi and Fabozzi ) assumes that only interest rate risk, and not credit risk, is the important factor in pricing corporate debt. This approach utilizes the standard techniques of duration and convexity hedging to risk manage a portfolio of corporate debt. These techniques are in common use by the industry.
The purpose of this paper is to critically analyze these two contrasting approaches to pricing credit risk. Using a unique data set, we provide an empirical analysis - a case study - of which risk, interest rate or credit (or perhaps both), is most important in the pricing of risky debt. The data set is unique because it consists of weekly quotes on a bank's (primary) debt offerings, for various maturities, over an 8-year observation period. Two standard models are compared in terms of their hedging performance. One is Merton's risky debt model, which assumes that interest rate risk is non-existent. The second is the traditional fixed income duration/convexity approach, which assumes credit risk is non-existent.
The hedging results are quite intriguing. The traditional fixed income approach dominates Merton's model, indicating that interest rate risk is significantly more important than credit risk, in the pricing of corporate debt. The implication, of course, is that for pricing and hedging purposes, if one risk needs to be ignored, it should be credit risk. But, this is not the final conclusion.
The results also indicate that the traditional fixed income approach to valuation still leaves a significant component of the bank's debt unhedged. We attribute the remaining hedging error to the omission of credit risk. The punch line is that the newer models, those that include both interest rate (market) and credit risk, are needed for more accurate pricing and hedging (see Jarrow and Turnbull  and Jarrow, Lando, Turnbull ).
Books Referenced in this paper: (what is this?)
Download paper (143K PDF) 20 pages