Asset/Liability Management: Year-End Interest Rate Risk Review

6/21/2026

Asset/liability management is the discipline of ensuring that what a bank earns on its assets, and what it pays for its funding, do not move against each other badly enough to threaten earnings or capital.

For a community bank the whole exercise reduces to one structural fact: the institution funds long assets with short liabilities, gets paid for accepting that mismatch, and has to know how much of it it is carrying.

The Two Measurement Frameworks

Both are needed, and institutions that run only one have a blind spot.

Earnings at risk measures the effect of rate changes on net interest income over a near horizon, typically one to two years. It answers the question management asks: what happens to earnings next year if rates move? Its limitation is the horizon — it cannot see a repricing mismatch that lands in year four.

Economic value of equity measures the effect of rate changes on the present value of assets net of liabilities. It captures the long-dated exposure earnings simulation misses, and it is the framework that reveals the real risk in a portfolio of long fixed-rate assets funded with non-maturity deposits.

The two can disagree, and the disagreement is informative rather than a contradiction. A bank can show acceptable earnings at risk over two years and severe economic value exposure, which means the mismatch is deferred rather than absent.

Assumptions Drive Everything

Models produce whatever the assumptions imply, and in ALM the assumptions carry more weight than the model.

Non-maturity deposit behavior is the most consequential. Checking, savings, and money market balances have no contractual maturity, so the model requires assumptions about how long they stay and how their rates move with the market. These assumptions determine whether the institution appears asset-sensitive or liability-sensitive, and they are frequently inherited from a vendor default that describes a different bank.

Deposit beta — the proportion of a market rate change passed through to depositors — deserves specific scrutiny. A beta assumption drawn from a low-rate period will understate funding cost in a rising environment, and institutions discovered exactly this when rates rose after a long flat stretch.

Prepayment speeds on mortgages and mortgage-backed securities, which vary with rates in ways that are not symmetric.

Loan repricing behavior, including whether floors and caps are modeled correctly. Floors that were binding in a low-rate period stop binding as rates rise, and a model that ignores them misstates both directions.

New volume assumptions — what the bank will originate and at what rate — which quietly turn a risk measurement into a business plan.

The discipline that matters: back-test the assumptions against what actually happened. An institution that assumed a deposit beta of thirty percent and experienced sixty has learned something the model will keep getting wrong until the assumption changes.

What the Year-End Review Should Cover

Re-run the simulations across the full range of scenarios the policy requires, including parallel shocks in both directions and non-parallel scenarios — steepening, flattening, and inversion. Parallel shocks alone miss the risk that actually materialized for many institutions.

Compare results to policy limits and document any breach with a remediation plan rather than a note.

Back-test. Compare last year's projected net interest income to actual, and explain the variance. This is the single most useful ALM exercise and the one most often skipped.

Review and re-derive the key assumptions, with the institution's own deposit data rather than industry defaults, and document the derivation.

Independent model validation, at a frequency proportional to the model's significance. This is a model risk management obligation, not an ALM nicety.

Liquidity. Sources and uses, the contingency funding plan, borrowing capacity confirmed rather than assumed, and stress scenarios that include a deposit run and a loss of wholesale access simultaneously.

Concentration in the funding base — large depositor concentration, uninsured deposit levels, and reliance on brokered or reciprocal funding.

The securities portfolio — unrealized position, duration, and whether the classification still reflects intent and ability.

Liquidity Deserves Its Own Attention

Interest rate risk gets the modeling effort; liquidity is what actually closes institutions.

The questions worth answering explicitly at year end: How much of the deposit base is uninsured, and how concentrated is it? How quickly could a meaningful share leave, and through which channels? What borrowing capacity exists, and has it been tested by actually drawing rather than by reading a line agreement? What would be sold in a stress, at what discount, and how quickly? Does the contingency funding plan name people, thresholds, and actions, and has it been exercised?

Deposit behavior changed materially once moving money became instantaneous, and models calibrated on an era when leaving a bank required a visit understate the speed of an outflow.

ALCO That Functions

The asset/liability committee is where this either becomes management or becomes a meeting.

Composition should include senior management across lending, deposits, finance, and risk — not finance alone, because the balance sheet is built by the lending and deposit functions.

The packet should be read before the meeting, and the meeting should discuss what the numbers mean rather than walk through them.

Assumptions should be a standing agenda item, not an annual one, because they are what the results depend on.

Decisions should be recorded — pricing changes, duration targets, funding actions — along with the rationale, because the minutes are the evidence that the committee governs rather than observes.

Bad news should reach the board. An ALCO packet that shows only favorable scenarios is not oversight.

Structured coverage is available through the Asset Management Certificate Program, Financial Risk Management: Interest Rate Risk, and the Certificate in Risk Management.

Where Community Banks Get Caught

Four patterns recur, and each has produced real damage.

Vendor default assumptions never localized. The model describes a generic institution, and its deposit behavior assumptions have no relationship to this bank's actual customer base.

Asset duration extended for yield in a low-rate period, funded with deposits that repriced faster than expected when rates rose. This is the classic community bank mismatch and it is not hypothetical.

Floors modeled as permanent. Loan floors that supported yield in a low-rate environment stop mattering as rates rise, and a model carrying them forward overstates asset sensitivity.

Liquidity assumed rather than tested. Borrowing lines that were never drawn, collateral that was never pledged, and securities assumed saleable at book value.

The common thread is that each is an assumption problem rather than a modeling problem — which is why assumption review, not model sophistication, is where the year-end effort belongs.

Reading an ALM Report Without Being an Economist

Most directors and many senior managers receive ALM reporting they cannot meaningfully challenge, which converts oversight into acceptance. Four questions make the packet interrogable by someone who is not a modeler.

"What assumption is driving this result?" Every headline number rests on a small number of assumptions, usually deposit behavior. A presenter who cannot name the two or three assumptions the result is most sensitive to has not examined the model, only run it.

"How did last year's projection compare to what happened?" This is the back-test, and it is the only evidence that the model has predictive value. A committee that has never seen this comparison is trusting a forecast with no track record.

"What would have to be true for us to breach a policy limit?" Reframing the output as a threshold rather than a point estimate makes it concrete. "Rates up 200 with a deposit beta above fifty percent" is something a director can reason about; a table of percentages is not.

"What are we doing about the exposure we are choosing to keep?" Some mismatch is the business. The question is whether it was chosen and priced, or accumulated by default because lending originated long fixed-rate assets and nobody connected that to the funding side.

For management, the corresponding discipline is to write the packet so those questions are already answered — a page stating the key assumptions and their basis, the back-test, the distance to each policy limit, and the actions taken since the last meeting. A packet that leads with fifty pages of scenario output and buries the assumptions in an appendix has communicated precision rather than understanding.

The underlying point is that ALM failures are rarely caused by bad models. They are caused by nobody in the room having enough grasp of the assumptions to notice that the model describes a bank other than this one.

Deriving Deposit Assumptions From Your Own Data

Since assumptions decide the answer, it is worth being concrete about how to build them from the institution's own history rather than accepting a vendor's.

Decay and average life. Pull five to ten years of balances by product and by account, and measure how balances actually behaved — how much of a given cohort remained after one year, three, five. Segment the analysis: a long-tenured retail checking relationship behaves nothing like a rate-sensitive money market balance opened during a promotion, and blending them produces an average that describes neither.

Beta, measured against actual repricing. For each product, compare the change in the rate the institution paid against the change in a market benchmark over the same period, through both a rising and a falling cycle if the data reaches back that far. Betas are frequently asymmetric — faster up than down, or the reverse — and a single symmetric figure hides that.

Surge balances. Deposits that arrived during an unusual period, whether from stimulus, a flight to safety, or a rate promotion, may not behave like core funding. Identifying and treating them separately prevents a model from assuming stability that was never earned.

Concentration and channel. Balances held by a small number of large depositors, and balances acquired through digital channels or rate-comparison sites, behave differently from a branch-acquired retail base. Modeling the whole deposit book as one homogeneous pool is the most common way an institution understates how quickly funding can leave.

Document the derivation, including the period examined and the segmentation. The documentation matters for two reasons beyond good practice: model validation will ask for it, and next year's review needs to know what was assumed and why in order to back-test it meaningfully.

An institution that does this once has something no vendor default can provide — a description of how its own depositors actually behave.

Frequently Asked Questions

What is the difference between earnings at risk and economic value of equity?

Earnings at risk measures the effect of rate changes on net interest income over a one- to two-year horizon, answering what happens to next year's earnings. Economic value of equity measures the effect on the present value of assets net of liabilities, capturing long-dated exposure the earnings horizon cannot see. Both are needed, and when they disagree the mismatch is deferred rather than absent.

Which ALM assumption matters most?

Non-maturity deposit behavior — how long checking, savings, and money market balances stay and how their rates move with the market. These accounts have no contractual maturity, so the assumption determines whether the institution appears asset- or liability-sensitive. Deposit beta assumptions drawn from a low-rate period systematically understate funding cost when rates rise.

What is deposit beta?

The proportion of a market rate change that is passed through to depositors. A bank assuming a thirty percent beta that experiences sixty percent will find its funding cost rising twice as fast as modeled. Back-testing the assumption against what actually happened is the correction, and it is frequently skipped.

Why is back-testing important in ALM?

Because it is the only check on whether the assumptions describe this institution. Comparing last year's projected net interest income to actual, and explaining the variance, identifies assumptions that are systematically wrong. Without it, a model can produce confident results that have never once matched reality.

What should a liquidity review cover?

The proportion and concentration of uninsured deposits, how quickly funds could leave and through which channels, borrowing capacity confirmed by actually drawing rather than by reading an agreement, what would be sold in a stress and at what discount, and whether the contingency funding plan names people, thresholds, and actions and has been exercised.

Who should sit on ALCO?

Senior management across lending, deposits, finance, and risk — not finance alone, because the balance sheet is created by the lending and deposit functions rather than measured into existence. The committee should discuss what the numbers mean rather than walk through them, treat assumptions as a standing item, and record decisions with their rationale.