As instructed by the 1991 banking legislation, regulatory agencies are moving to require banks to calculate a measure of interest rate risk and are planning to incorporate this measure in the calculation of banks' risk-based capital requirements.
Bankers - especially those running smaller institutions - take a dim view of any regulatory change that increases their reporting burden, as this does.
The Federal Reserve System, Federal Deposit Insurance Corp., and Office of the Comptroller of the Currency, well aware of this attitude, have tried to devise a reporting procedure that strikes a balance between maximizing the accuracy of the risk measure and minimizing the reporting burden.
Nonetheless, the measure they have devised has enough built-in biases to raise the question of whether the improved monitoring it makes possible justifies the added reporting burden. Many bankers think not.
Capital requirements now are calculated on credit risk alone. Banks holding a high proportion of assets deemed subject to default risk must hold more capital per dollar of assets than banks holding mostly Treasury bonds, for example.
Avoiding Another Bailout
The idea, of course, is to protect taxpayers, who insure bank deposit liabilities through the FDIC, against massive shortfalls in the value of assets relative to liabilities. Such imbalances could necessitate a bailout of commercial banks like that in the thrift industry.
However, for some institutions, interest rate risk may be just as important as credit risk in determining the taxpayer's potential exposure.
The problems of the savings and loans became acute after sharp increases in interest rates in the early 1980s forced thrifts to pay increased interest to depositors, but provided no corresponding revenue increase since these institutions were holding mostly fixed-rate mortgages.
The regulatory agencies want to detect such problems before they become critical, and therefore propose to require that banks report a measure of interest rate risk.
Like a Gap Report
The suggested new reporting procedure is a glorified gap report: Banks classify their assets and liabilities into maturity categories ranging from 0-3 months to more than 20 years.
Entries in each category are then broken down into amortizing, non-amortizing, or deep discount assets or liabilities, since these have very different exposures to interest rate risk. (For a discussion of an early version of the proposed reporting procedure, see the August 1991 Federal Reserve Bulletin.)
Then, going beyond traditional gap reports, the procedure multiplies each entry by an estimate of the duration of items in that category. This duration measure represents an estimate of average interest rate risk of items in that class.
Banks would estimate the duration of net equity by subtracting duration of liabilities from duration of assets. This number is in turn divided by the magnitude of net equity to estimate the bank's exposure to interest rate risk per dollar of net equity.
This calculation means that if two banks have the same dollar exposure to interest rate changes, the bank with higher net equity will have the lower measure of interest rate risk. This is exactly what one wants since the equity provides the cushion protecting the three federal insurance funds against adverse interest rate changes.
An Accounting Hybrid
One problem with this procedure is that it is a hybrid of book-value and market-value accounting. Duration is a market-value concept in spirit: It is intended to measure the effect of a 1% change in interest rates on the market value of the cash flows generated by a security.
Yet in calculating interest rate risk, the procedure just summarized multiplies duration by the hook value of net equity.
Institutions currently placing big bets on hoped-for yield curve shifts are likely to have placed similar bets in the past. Depending on past yield curve movements, these institutions are likely to have equity with a market value significantly different from book value.
Therefore, in substituting book value for market value, the Fed distorts the estimated interest rate risk per dollar of equity, and the distortion is most severe for institutions with sizeable interest rate risk.
Duration vs. Rates
Another problem with the Fed's procedure is that the duration of most securities depends strongly on the level of interest rates. The procedure takes no account of this dependence, This would not be a serious problem except that the degree of dependence is different for various types of assets and liabilities, and there is no reason to believe that these effects typically cancel out.
For example, the duration of a noncallable bond rises as interest rates drop. since at low interest rates increased weight is given to cash flows in the distant future relative to those in the near future - a concept referred to as "positive convexity."
For mortgage-backed bonds the opposite is true, since low interest rates accelerate prepayments "negative convexity"). If noncallable bonds and mortgages enter a bank's balance sheet on opposite sides - as occurs when bond futures are used to hedge a position in mortgage pools, for example - these effects reinforce each other.
Banks, knowing this, alter hedge ratios as interest rates change - to ensure offsetting interest rate risk - or adjust otherwise to allow for convexity.
Big Errors Likely
By applying durations that do not change as interest rates change, the Fed's procedure errs in estimating interest rate risk: Even if the procedure gives the right estimate of risk for some particular level of interest rates, it is likely to make big errors when rates are lower or higher.
Third, there are major ambiguities in using duration analysis to estimate interest rate risk. Of these, the most important involves core deposits. These typically are rolled over from month to month and year to year, but depositors can withdraw them on short notice. What duration should they be assigned?
The Federal Reserve Bulletin article discusses this problem extensively. The Fed economists point out that, because core deposits are such a large fraction of bank liabilities, determining the net interest sensitivity of bank equity depends critically on how the duration of core deposits is estimated.
If core deposits are assigned a duration of 2.5 years, U.S. banks have approximately zero interest rate risk, on average. However, if their duration is 1.5 months, all but a few banks are adversely affected by interest rate rises, while if five years is the right figure, then a large majority of banks are adversely affected by interest rate drops.
What's the Point?
If the whole calculation is so strongly affected by an essentially arbitrary assignment of a duration number, bankers are entitled to ask whether the exercise has any point.
Also, durations are notoriously difficult to evaluate. They depend on interest rate volatility and hard-to-estimate deposit runoff rates, mortgage prepayment rates, and the like.
The Fed economists are well aware of these problems, but discuss them only in the context of dismissing "sophisticated computer models" as an alternative to the simpler procedure they recommend: They argue that "the chief benefit of simulation models resides, to a large degree. in revealing the sensitivity of results to the assumptions used."
However, if simulation models cannot be relied upon to provide good estimates of duration, where do the Fed's duration numbers come from, and how accurate are they?
There are replies to these criticisms. The regulatory agencies can deal with the dependence of duration on the level of interest rates by altering, as interest rates change, the weights used in computing average duration.
While it is true that there are various inaccuracies in estimating average duration, regulators say. their proposal appears to identify outliers successfully - in that it flags the same banks as do more sophisticated techniques for interest rate risk appraisal such as the ones used by the Office of Thrift Supervision.
Finally, while it is correct that the proposed procedure requires data beyond that found on call reports, even small banks should be monitoring their exposure to interest rate risk.
The regulators are correct that in as much as the purpose of the reporting procedure is to identify outliers, most of the biases and ambiguities in arriving at an absolute estimate of interest rate risk are immaterial to the extent that their effect on all banks is approximately equal.
However, a still simpler procedure might achieve that goal equally well. For example, banks might be asked to report the percentage of assets maturing more than five years in the future. The Fed is investigating how well such a screen correlates with its recommended procedure.
A major problem is that even if the screen works well on past data, the correlation may break down as banks alter their portfolio management practices in response to the altered monitoring of interest rate risk.
A Clear Tradeoff
The tradeoff is clear: A complex reporting procedure for interest rate risk imposes onerous reporting requirements, particularly on small banks. However, a simple reporting procedure is less accurate and easier for banks to circumvent.
There is a better solution. The, Fed's own data show that for all but a few banks, the increment to required capital necessitated by interest rate risk is less than 100 basis points. Therefore, there is virtually no chance that a bank with as much as 10% capital will be found capital-deficient because of the interest sensitivity of portfolio equity.
Why not exempt all banks with capital over 10%? Many small banks would be exempted. This would clear the way for in-depth examination of larger banks, which are generally more weakly capitalized, are more likely to hold interest-sensitive securities, and can better comply with a more rigorous reporting requirement.
Mr. LeRoy is Carlson professor of finance at the University of Minnesota's school of management.