Gap, beta-adjusted gap, duration and even basic budgeting models only frustrate, confuse and even mislead the financial institution’s asset liability management committee (ALCO). Detailed gap analysis, fiddling with the distribution of savings balances and even calculating the duration of equity does not lead to better margins, nor do they mitigate rate risk.

Even more extensive full simulation models only perform three or four scenarios at a time, and some of these take hours to execute. Entrenched in traditional time-based methods, margin managers persist in the belief that more data is better, reports must always foot and graphs always have time as the X-axis.

My purpose here is to suggest a more useful tool with which to measure, control margin behavior, and mitigate your interest rate risk. Current hardware and software technology can provide a better analytical method that clearly communicates the extent of the financial institution’s rate risk and simultaneously suggest a successful strategy without the expense in time and energy of more primitive techniques.

Factors Affecting Margin Behavior

If interest rate risk is a behavioral problem, then the purpose of the ALCO is to control the behavior of the margin. The behavior of the financial institution’s margin, however, is itself a function of other factors, including the financial institution’s behavior in pricing loans and deposits and customer behavior.

These factors are dependent variables because they respond to another variable that is not controllable by the financial institution, i.e. market rates. Here we use the term market rates to mean the national rates such as prime, the Treasury yield curve and LIBOR. The challenge is to determine the behavior of the net interest margin as these rates changes.

Rate Shock: Rate Change, Not Timing

In order to standardize the measurement of margin behavior over a range of market rate changes, we must remove the timing issue from our analysis. Clearly, timing is important, but the rate changes up or down only clouds the issue.

Our purpose is to understand behavior, not predict margin. We remove the time factor by shocking the market rates, i.e. moving market rates immediately and holding them constant over the analysis period. Typically rate shock yields plotted against shock increments, can express margin behavior. However, irregular behavior is caused by customer behavior, prepayments, and the financial institution’s own pricing behavior as rates change.

Information Requirements

The data requirements for rate shocks are typical for all rate risk analysis, i.e. maturity and repricing data. However, we must also include both prepayment models, and a spread model of each account. These are not features found in all simulation models, so let’s review the methods for building them.

Prepayment Models

While prepayment models are a part of every mortgage-backed security and CMO analysis, many community bankers fail to apply the same logic to their in-house mortgage portfolios. These are, in fact customer behavior models and have market rates as their independent variable.

As market rates rise and fall, mortgage prepayments decrease and increase. Clearly, this has a significant impact on yield. The information needed to construct prepayment models is available by either examining the most recent prepayment information from the financial institution’s records, or from a securities broker who can use the prepayment characteristics of a MBS from your trade area.

Spread Models

Spread models reflect financial institution’s pricing as market rates change. It is common that community bankers widen spreads between prime and new loan rates as prime falls and the narrow them as prime rises. On the deposit side, it is common to lower rates quickly on savings and money market accounts as rates fall, but delay rate changes as rates rise.

Whether he or she is aware of it or not, the banker is expressing a pricing behavior model. Beta-adjusted gaps are a simplistic application of this concept translated into balance repricing, then carried to the gap report.

Shock Calculations

Putting these all together, we are ready to perform rate shocks on each category on the balance sheet and produce behavior graphs that describe our risk.

What happens within the rate shock model? First, the driver rate is changed by the shock amount. Next, the pricing model is used to determine the rate on new balances. Then cash flows, maturities, prepayments and interest payments, are computed for the next 12 (or 24) months. Finally, these cash flows are reinvested at the new rates and blended with the unchanging portion of the portfolio to determine a new blended yield.

Accounts that reprice immediately such as Fed Funds, of course, do not use maturities, and so the entire category changes. The average yield is computed for the entire time period analyzed.

Finally, these are combined into total asset yields and liability costs to produce the margin for the shock increment, the time required to perform these computations varies within the complexity of the balance sheet, but typically 1-2 minutes at most.

Once we have computed all of the information we can use EIS, Executive Information System, techniques such as drill-down to view information in different ways. One example would be in comparing the yield behavior of loans, investments, and deposits identifying potential strengths and weaknesses within your margin.

Spread Behavior Useful for Loan Pricing

Other uses for this information include loan pricing, where we can compare the yield behavior of a loan to that of its funding counterpart. Among other factors, as part of your price building function, the rate risk identified would be added to the loan rate determination.

In conclusion, current asset-liability management techniques are only extensions of the community bank budgeting process. Of the methods available, simulation provides the greatest benefit when it includes all of the factors that affect margin and income.

Rate shock, however, offers the benefits of simulation, with the added advantage of removing the timing issue, and performing many more scenarios automatically. This computational power, coupled with its descriptive graphic presentation is a strong analytical to that translates into a more productive ALCO process.