For the past six years, financial institutions have beenlengthening the duration of their loans and investments to slow theshrinking of interest margins. As a result, the NCUA has becomemore concerned about interest rate risk (NCUA Letter to CreditUnions 14-CU-02) and is focusing more on each credit union’s methodof measuring and addressing IRR.

Other regulatory agencies have raised IRR concerns as well. Therecent run-up in interest rates has further stoked IRR concernsamong CEOs and regulators and has sharpened the debate among theseparties as to the efficacy of different approaches toasset/liability management modeling. That debate revolves aroundkey issues such as:

1. The assumptions (quantity and quality) that are being used as abasis for a financial institution’s ALM modeling process;
2. The depth of understanding of ALM modeling concepts on the partof managers, boards and examiners;
3. Using past experiences as a basis for assumptions fed into anALM model;
4. The efficacy of traditional ALM modeling processes(particularly net equity value);
5. The lack of proper consideration for operational expenses andnon-interest income in some ALM models;
6. The accuracy and fairness of examiners’ criticism of ALMmodeling outcomes used by managers who are trying to balanceprofitability with safety.

Clearly, in volatile times like now, a financial institution’sfuture could well depend on the ALM model it uses, the accuracy ofthat ALM model’s conclusions, and how management uses thoseconclusions in its planning and forecasting. Let’s compare twodistinct approaches to ALM modeling:

The Net Equity Value ALM Model

The traditional method for measuring IRR is an ALM modeling toolthat relies on estimating the net equity value of a financialinstitution’s balance sheet. Some students of ALM point out thatthere are inherent weaknesses in NEV that need to be taken intoconsideration by managers and regulators. These weaknessesinclude:

1. The level of dependency on assumptions to estimate the maturityof non-maturity deposits;
2. Using discounted cash flows to arrive at the present value of afinancial institution’s balance sheet (a financial institution’sliquidation value);
3. The assumption that discounted cash flows of a balance sheetcan be used to estimate changes in net worth and earnings resultingfrom changes in market rates;
4. The assumption that a financial institution’s loan portfolio isequivalent to bond portfolios that are widely traded;
5. The assumption that a financial institution’s deposit accountscan be treated like bonds which have contractual maturities and arewidely traded;
6. Small inaccuracies in the assumptions used that could lead toerroneous ALM conclusions that can in turn result in perilousmanagement decisions.

Next Page: Earnings at Risk

The Earnings at Risk ALM Model

Arguably, a more accurate and easier to understand ALM modeluses earnings at risk to measure a financial institution’s IRR. EARdoes not rely on assumptions to the extent that NEV does andtherefore has greater value for CEOs and CFOs who are trying toforecast the effects potential changes in interest rates will haveon profitability.

The best EAR models project cash flows and impacts onprofitability using actual payments coming from individual loansand investments in a financial institution’s balance sheet.Better-quality EAR models also take into account additional factorsthat affect profitability such as fee income, maturing CDs andoperating expenses.

EAR ALM models assure validity by holding constant the assetsand liabilities in a balance sheet so as to measure the actual IRRin the current balance sheet. Once the base IRR is established, anEAR model can also be used for multiple simulations wheremanagement can vary inputs and view the impacts each change orcombination of changes has on IRR, income and equity. Simulationsmay include (1) increasing or decreasing loans and/or investments;in combination with (2) increasing or decreasing deposits (specifictypes or general) including changing the mix of deposits; and (3)changes in operating expenses including provisions for loanlosses.

ALM models should include the effects of interest rate shocksthat measure the impact possible changes in interest rates willhave on balance sheets and profitability. Better models measure theimpact of two distinct possible rate shock scenarios: (1) animmediate, extreme (typically 300 to 600 basis points) increase ordecrease in rates; and (2) small but continued changes in interestrates (sometimes referred to as stepped shocks).

Many EAR models that are based on careful research use priceelasticity of demand applications to determine the percent of shockthat can be applied to each deposit type to maintain currentbalances of deposits. This price elasticity measure then employs aregression model to predict runoff at distinct interest rateincreases. This algorithm is used to determine the magnitude ofcost of funds increases required to fund loans and investments tomaturity (based on the current balance sheet configuration).

A financial institution’s ability to mitigate the effect ofinterest rate shocks generally rests in its ability to re-price itsloan/investment portfolios. Stochastically derived EAR modelsusually set reductions in loan/investment portfolio principalbalances in repayment schedules typically within a set range inmaturities and with even cash flows. Therefore, a carefullydesigned EAR model can use the weighted average maturity ofloans/investments to create an anticipated amortization schedulefor principal balances of loans/investments.

Loans/investments with common amortization patterns (asdetermined by WAM) are input into the model. Using the WAM, astraight-line calculation of amortization is derived. The amount ofamortization is then applied to each quarter and year as a baselinefor re-pricing. A statistically derived algorithm is then appliedto the amortization schedule in shock scenarios to allow foranticipated changes in prepayment speeds resulting from shocks ofincreasing magnitudes. This allows managers to choose differentshocks in simulations to test the limits of IRR in their balancesheets.

In summary, it can be argued that a stochastically derived ALMmodel utilizing EAR can provide a more meaningful IRR managementtool for financial institutions than NEV. Furthermore, argumentsare made that EAR provides more accurate planning applications – adefinite advantage over NEV models which often rely on manyunsubstantiated assumptions. Considering the advantages EAR appearsto have over the more traditional NEV, financial institutions wouldbe wise to give the EAR approach to ALM careful consideration.

DennisChild is a retired credit union CEO in Logan, Utah,now associated with Thompson Consulting andTraining in Boise, Idaho.