Guide

Bank net interest margin (NIM) explained

Harbor Credit Union's $4.8B balance sheet reported record loan originations in 2023 while net interest margin fell 38 bp year over year. The board deck showed higher average mortgage and commercial yields on the asset side — so why was the core spread shrinking? The ALM team had been tracking headline loan rates against a single blended deposit cost, ignoring repricing lags on deposit beta, securities book duration, and a shift from low-cost checking into money-market accounts. NIM looked like a mystery when it was really three moving parts: asset yield, funding cost, and balance-sheet mix.

The refactor built a quarterly NIM bridge that decomposed changes into rate (repricing), mix (product and tenor shifts), and volume (balance growth on different-margin buckets). Forward NIM forecast error fell from 22 bp to 6 bp over four quarters. This guide explains what NIM measures, how it differs from net interest income in dollars, why NIM expands early in hiking cycles and compresses later, the Harbor Credit Union ALM workflow, a technique decision table versus return on assets and the efficiency ratio, pitfalls equity analysts and treasury desks hit, and a production checklist for forecasting and stress testing bank spreads across yield-curve regimes.

What net interest margin is

Net interest margin (NIM) is the spread a bank earns between what its interest-earning assets yield and what its interest-bearing liabilities cost, expressed as a percentage of average earning assets. The standard formula:

NIM = net_interest_income / average_earning_assets

Net interest income (NII) is interest revenue minus interest expense — the dollar engine behind most traditional bank earnings. NIM normalizes NII by asset size so you can compare a community bank to a money-center franchise or track one institution through growth without conflating balance-sheet expansion with spread improvement.

Earning assets typically include loans, leases, and interest-bearing securities (Treasuries, agencies, MBS). Non-earning assets — cash in reserves, premises, goodwill — sit outside the denominator. Funding includes deposits, wholesale borrowings, and long-term debt. Fee income (interchange, advisory, servicing) is excluded; NIM is pure spread economics.

Reported NIM is usually presented on a tax-equivalent basis for banks with large municipal bond portfolios. Always check whether you are comparing GAAP, FTE-adjusted, or peer-group normalized figures before ranking franchises.

The NIM bridge: rate, mix, and volume

Quarter-over-quarter NIM changes decompose into three intuitive buckets:

  • Rate effect — assets and liabilities reprice at different speeds. Floating-rate loans may reset monthly while fixed-rate mortgages and securities lag by years. Deposit costs follow deposit beta curves, not the policy rate instantly.
  • Mix effect — growth in high-margin commercial loans versus low-margin residential mortgages, or migration from non-interest-bearing checking into interest-bearing savings, changes blended yields and costs even if every product's rate is unchanged.
  • Volume effect — balance growth on wide-spread products lifts dollar NII but may dilute or concentrate NIM depending on marginal asset yield versus marginal funding cost.

A disciplined ALM team publishes a bridge table each quarter: prior NIM, plus rate, plus mix, plus volume, plus other (day-count, prepayment, securities gains embedded in yield), equals current NIM. Without the bridge, management narrates “we grew loans” while NIM falls — a story that confuses investors until mix is explicit.

Asset sensitivity and liability sensitivity (often summarized in a gap or duration-of-equity report) predict which rate direction helps NIM. Asset-sensitive banks (more floating-rate loans, more non-interest-bearing deposits) tend to benefit when the Fed hikes; liability-sensitive banks (long fixed-rate assets, high-beta deposit franchises) feel compression sooner.

NIM through the rate cycle

Banks rarely move in lockstep with the policy rate. A typical hiking cycle unfolds in phases:

  1. Expansion phase — short rates rise first. Floating loans and new origination reprice up while legacy deposit costs stay sticky. NIM widens. This is the “deposit lag" windfall that peaked for many U.S. regionals in 2022.
  2. Compression phase — deposit competition intensifies, betas rise, and wholesale funding costs catch up. Fixed-rate asset yields lag on the way down but deposits reprice down more slowly in cutting cycles, reversing the pattern.
  3. Curve shape matters — a flat or inverted yield curve squeezes maturity transformation: banks borrow short and lend long, so when long yields fall below short funding costs, securities portfolios and mortgage pipelines earn less than incremental deposits cost.

Credit quality is adjacent but distinct. Rising charge-offs hit provision expense and capital, not NIM directly — unless non-performing loans stop accruing interest, which reduces reported asset yields. Analysts should separate spread economics from credit costs when judging earnings power.

Harbor Credit Union: rebuilding the NIM forecast

Harbor's original model projected NIM from a single “loan yield minus deposit cost” spread applied to total earning assets. It worked in stable rate environments and failed in 2022–2023 when:

  • Commercial loan share rose from 28% to 34% of earning assets (higher yield, higher risk weight).
  • Non-interest-bearing deposits fell from 41% to 33% of funding as customers chased money-market yields.
  • The securities portfolio duration extended when reinvestment rates were still low, dragging average asset yield down with a lag.

The rebuilt model segmented six asset buckets and five liability buckets, each with its own repricing schedule and beta assumption. Quarterly output included: dollar NII forecast, NIM level, full bridge versus prior quarter, and a sensitivity grid (+/- 100 bp parallel shock, steepener, flattener). Board materials replaced a single NIM line with the bridge chart — investors stopped asking why loan yields rose while NIM fell because mix was visible. Forecast error on forward NIM dropped from 22 bp to 6 bp.

Technique decision table

Metric or tool Best for Weak when
NIM level and bridge Spread economics, ALM storytelling, rate-cycle positioning Fee-heavy banks, trading revenue dominates
Net interest income ($) Absolute earnings power, capital planning Comparing banks of very different sizes without normalization
Return on average assets (ROA) Holistic profitability including fees and expenses Isolating spread drivers from cost discipline
Efficiency ratio Non-interest expense control Explaining why spreads widened or compressed
Deposit beta alone Liability repricing forecast Full NIM without asset-side and mix context
Duration of equity / gap report Directional rate-risk sign and magnitude Precise NIM dollar forecast without cash-flow reprice models
Peer NIM percentile Franchise quality vs similar-size banks Different business mixes (CRE-heavy vs mortgage warehouse)

Common pitfalls

  • Confusing NIM with ROA — a bank can have wide NIM but poor ROA if expenses or credit costs are high.
  • Ignoring FTE adjustment — tax-exempt muni yields inflate NIM for some peers only.
  • Single-spread shortcut — one loan yield minus one deposit cost misses product mix and repricing lags.
  • Annualizing one noisy quarter — securities prepayments and day-count quirks distort single-period NIM.
  • Excluding securities book — for many banks, 15–30% of earning assets are bonds; their yield drag or lift matters.
  • Assuming symmetric hike/cut dynamics — deposit betas are often asymmetric: slow up, sticky down (or the reverse for brokered funding).
  • Peer comparison without mix — mortgage-heavy banks run lower NIM than card lenders by business model, not by incompetence.

Production checklist

  • Define NIM = NII / average earning assets; document FTE convention.
  • Segment assets and liabilities into repricing buckets with contractual reset dates.
  • Estimate deposit beta by product tier; link to deposit beta guide assumptions.
  • Publish quarterly NIM bridge: rate, mix, volume, other.
  • Run parallel +/- 100 bp shocks and curve twists on NII and NIM.
  • Reconcile model NII to general ledger interest income/expense monthly.
  • Track marginal NIM on new originations versus portfolio average.
  • Separate non-accrual impact on asset yield from spread compression.
  • Benchmark against peer group with similar business mix, not just asset size.
  • Board and investor materials: show bridge, not just headline NIM delta.

Key takeaways

  • NIM is net interest income divided by earning assets — the normalized spread a bank earns on its lending book.
  • Decompose changes into rate, mix, and volume before narrating a quarter.
  • NIM often expands early in hiking cycles and compresses when deposit costs catch up.
  • Deposit beta, securities duration, and loan mix are as important as headline loan yields.
  • Harbor Credit Union cut NIM forecast error from 22 bp to 6 bp with a segmented bridge model.

Related reading