By the end of this page you should be able to take a bank's net interest margin and predict its return on assets — and, more useful, explain why two banks with the same margin post very different returns. The pivot is overhead, and its name is the efficiency ratio.
Lesson 1 ended with net interest margin: the spread a bank keeps on its earning assets. But NIM is a gross number, near the top of the income statement. Return on assets — net income divided by assets, the single most-quoted measure of bank profitability — sits at the bottom. Five things happen to the spread on the way down.
Everything in the waterfall is expressed as a percentage of assets, so the lines add and subtract cleanly down to ROA. That is a different denominator from NIM, which divides by earning assets. So the top line here — net interest income over total assets — runs a touch below the NIM from Lesson 1. Same spread, slightly larger denominator. Do not be surprised when a bank's 5.2% margin shows up as 4.9% at the top of its waterfall.
Four stages stand between the margin and the return. Learn to read each off the income statement and you can predict ROA from NIM — and, more importantly, know which stage explains a disappointing one.
Service charges, card interchange, wealth and trust fees, mortgage gain-on-sale. For most community banks fees are a modest topping, a median of about 0.4% of assets, roughly a tenth of revenue. For a few — mortgage shops, trust companies, card specialists — fees are the main event and dwarf the spread. Read the fee line to know whether a bank lives on its margin or on something else.
Salaries, occupancy, technology: the cost of running the bank. The median community bank spends about 2.5% of assets on itself. This is where the efficiency ratio lives — overhead divided by revenue (net interest income plus fees). A 40% ratio is a lean machine that keeps 60 cents of every revenue dollar; 75%+ means most of every dollar is eaten running the place, before a single loan goes bad. Overhead is the single biggest reason two banks with identical margins earn different returns.
Money set aside for loans expected to sour. In a benign year it is near zero (a median of 0.05% of assets); in a bad year, or the quarter after an acquisition, it can swallow a full year of pre-provision profit at once. It is the least structural, most cyclical line in the waterfall. Read it, but know it swings, and that one quarter tells you little.
A C-corporation bank pays corporate tax, roughly 0.3–0.7% of assets at these return levels. But many community banks are Subchapter-S corporations that pass income through to shareholders and pay no entity-level tax, so their ROA looks structurally higher. The universe median tax is only about 0.17% of assets, dragged down by all the Sub-S banks. When you compare two banks' ROA, first check whether you are comparing a taxpayer with a non-taxpayer.
To predict an ROA you need a feel for the size of each line. Below are the whole-universe figures for Q1 2026, across the roughly 3,500 banks above $100M in assets: the median for each stage and the middle-half range, all as a percentage of assets.
| Waterfall line | Median | Middle half |
|---|---|---|
| Net interest income | 3.6 | 3.1–4.0 |
| + Fees | 0.4 | 0.3–0.7 |
| − Overhead | 2.5 | 2.1–3.0 |
| = Pre-provision profit | 1.5 | 1.1–2.0 |
| − Provision | 0.05 | 0.0–0.1 |
| − Tax | 0.2 | 0.0–0.3 |
| = Return on assets | 1.2 | 0.8–1.6 |
| Efficiency ratio | Reading |
|---|---|
| under 50% | Exceptional. A lean cost base; most of revenue survives. |
| 50 – 60% | Strong. Better than the median bank. |
| 60 – 70% | Average. The middle of the pack (median ~62%). |
| 70 – 80% | Heavy. Overhead is claiming most of revenue. |
| over 80% | Struggling, or mid-build. Little reaches the bottom. |
Read the left table top to bottom and you have the shape of a typical community bank: a 3.6% spread, a small fee topping, 2.5% of overhead, almost no provision in a calm year, a light tax, and about 1.2% left as return on assets. The line that moves most between a good bank and a poor one is overhead. Fees help at the margin, provision and tax bite occasionally, but overhead is on every quarter and it is the widest-ranging line of the lot.
Start with the bank from Lesson 1 whose 5.22% margin we predicted from its structure. Now watch that margin fall to a return. Here is what you know before you open the reveal: the margin, the efficiency ratio, and the fee share of revenue. Predict the ROA.
| Net interest margin (Lesson 1) | 5.22% |
| Efficiency ratio | 38.2% |
| Fee share of revenue | 6.2% |
| Corporate form | C-corp (taxpayer) |
Reveal the return
| Waterfall (% of assets, annualized) | Q1 2026 |
|---|---|
| Net interest income | 4.92 |
| + Fees | 0.32 |
| − Overhead | 2.01 |
| = Pre-provision profit | 3.24 |
| − Provision | 0.17 |
| = Pre-tax income | 3.07 |
| − Tax | 0.60 |
| = Return on assets | 2.47 |
| Reported ROA (FDIC) | 2.49 |
2.49%, roughly double the community-bank median and inside the top few percent of the universe. The wide margin got Grandview to the start line; the 38% efficiency ratio carried it home. Overhead of just 2.0% of assets left three-quarters of the spread standing. Fees added almost nothing — this bank earns its living on spread. Provision took a sliver, and even a full 0.6% of assets in corporate tax could not pull the return below 2.5%.
The lesson to bank: nothing exotic produced this return. A wide margin met a lean cost base, and the waterfall did the rest. When you see a 2.5% ROA, look first at the efficiency ratio — it is almost always the reason.
1.2% – 2.0% — You landed near the median. You likely respected the margin but under-weighted how little overhead this bank carries; a 38% efficiency ratio is exceptional and most of the spread survives it.
over 3.0% — You forgot the subtractions. Even a lean, wide-margin bank still pays overhead, some credit cost, and full corporate tax. Those stages cost Grandview about 2.8 points of ROA.
Now the point of the whole lesson. Here is a different bank with almost exactly Grandview's margin — and, as it happens, more fee income. Watch what a single stage does to the outcome.
| Net interest margin | 5.15% |
| Efficiency ratio | 75.5% |
| Fee share of revenue | 17.2% |
| Corporate form | C-corp (taxpayer) |
Reveal the return + score yourself
| Waterfall (% of assets, annualized) | Q1 2026 |
|---|---|
| Net interest income | 4.61 |
| + Fees | 0.96 |
| − Overhead | 4.22 |
| = Pre-provision profit | 1.35 |
| − Provision | 0.06 |
| = Pre-tax income | 1.29 |
| − Tax | 0.44 |
| = Return on assets | 0.85 |
| Reported ROA (FDIC) | 0.86 |
0.86% — a third of Grandview's return, on the same margin. Nothing went wrong with the spread: at 5.15%, FirstBank's NIM is elite, and its fees are three times Grandview's. The entire gap is overhead. FirstBank spends 4.22% of assets running itself, against Grandview's 2.01%, and a 5%+ margin that would carry most banks to a fat return instead mostly pays FirstBank's own bills. That is a 76% efficiency ratio in one sentence.
Whatever the cause — a thin branch network spread across rural southeastern Oklahoma, heavy staffing, a build-out not yet earning its keep — the reading is the same from the outside: revenue is strong, but too little of it survives the cost base. This is the lesson of the whole page. NIM tells you the size of the pipe; the efficiency ratio tells you how much reaches the bottom. Provision and tax barely entered the story.
1.5%+ — You anchored on the 5.15% margin and the fat fee line and forgot to subtract the overhead. The wide margin is real; it just does not survive a 4.2%-of-assets cost base.
under 0.5% — You over-corrected. The overhead is heavy, but the margin and fees are genuinely strong, and provision and tax are light, so the bank still clears a positive return.
A third bank, the last of the Lesson 1 trio. Its margin is neither wide nor thin — it is average. So you cannot lean on the spread alone; you have to net all the stages. Commit to an ROA and a one-line reason before you reveal.
| Net interest margin (Lesson 1) | 3.80% |
| Efficiency ratio | 54.7% |
| Fee share of revenue | 13.5% |
| Corporate form | C-corp (taxpayer) |
Reveal the return + score yourself
| Waterfall (% of assets, annualized) | Q1 2026 |
|---|---|
| Net interest income | 3.57 |
| + Fees | 0.56 |
| − Overhead | 2.26 |
| = Pre-provision profit | 1.86 |
| − Provision | 0.10 |
| = Pre-tax income | 1.76 |
| − Tax | 0.35 |
| = Return on assets | 1.41 |
| Reported ROA (FDIC) | 1.42 |
1.42%, comfortably above the 1.21% median. The average margin never let Timberland reach Grandview's heights — there simply is not as much spread to work with. But a 55% efficiency ratio meant the bank kept a healthy share of the spread it had. Good cost control turned a middling margin into an above-median return. That is the whole trick of a well-run bank without a structural margin advantage.
over 1.8% — You gave the margin too much credit. At 3.8%, the spread is only average; even excellent cost control leaves less to work with than Grandview had.
under 1.0% — You under-weighted the efficiency ratio. A 55% cost base preserves most of even an average margin.
The three banks side by side
Same framework, same quarter, three outcomes. Read across and the story is one line: the margin sets the ceiling; overhead decides how close the bank gets to it.
| % of assets | Grandview | Timberland | FirstBank | Piscataqua |
|---|---|---|---|---|
| Net interest income | 4.92 | 3.57 | 4.61 | 1.50 |
| + Fees | 0.32 | 0.56 | 0.96 | 1.18 |
| − Overhead | 2.01 | 2.26 | 4.22 | 2.22 |
| = Pre-provision profit | 3.24 | 1.86 | 1.35 | 0.46 |
| − Provision | 0.17 | 0.10 | 0.06 | 0.03 |
| − Tax | 0.60 | 0.35 | 0.44 | 0.07 |
| = Return on assets | 2.49 | 1.42 | 0.86 | 0.36 |
| Memo: NIM | 5.22 | 3.80 | 5.15 | 1.62 |
| Memo: efficiency ratio | 38% | 55% | 76% | 83% |
Grandview and FirstBank have all but the same margin; the 1.6-point gap in their returns is almost entirely the 2.2-point gap in overhead. Timberland turns an average margin into an above-median return through cost control alone. And Piscataqua, the thin-margin thrift from Lesson 1, shows the floor: a 1.6% margin and a heavy fee reliance, and its return survives only because it takes almost no credit cost and pays almost no tax. Four banks, one waterfall, and overhead is the line doing most of the sorting.
The waterfall is the measurable part, and it explains most of the return. But it predicts a level, not a verdict, and the gap between the two is what the stages cannot see:
- Whether the overhead is bloat or investment. A 76% efficiency ratio can mean a flabby cost base or a bank mid-build — new branches, new bankers, a technology platform — whose revenue has not caught up yet. The ratio looks identical either way. One normalizes over the next few years; the other does not. Only the trend, and knowing what the bank is spending on, tells you which.
- Whether the fees are durable. Mortgage gain-on-sale and one-time gains inflate a quarter and vanish the next; recurring card and service fees are franchise. A fat fee line can be either. The single number does not distinguish them — the trend across quarters does.
- What the provision is hiding. A near-zero provision in a calm year says nothing about underwriting quality; the test only comes in the downturn. Provision is the most backward- and forward-looking line at once, and the least readable from a single quarter. That is the credit-quality lesson, later in the series.
- Whether the ROA is taxed. A Sub-S bank flatters its return by paying no entity-level tax. To compare it like-for-like with a C-corp you sometimes have to tax-adjust the return by hand. And ROA still says nothing about how much equity stands behind the assets — that is leverage, and it is the next lesson.
What you have learned: the margin is only the first line. Read the fee line, subtract overhead, allow for provision and tax, and you can predict return on assets from structure — and, more useful, name the stage that explains a weak one. Nine times in ten that stage is overhead, and the efficiency ratio is the number that reveals it.
Next: No. 3 — Return on equity & leverage, where this return on assets meets the balance sheet and we ask why a well-run bank can still post a mediocre ROE.
Is this useful? What would you add, remove, or change? Share feedback →