Reading the Numbers · No. 1
Net Interest Margin
A bank earns a spread on money it borrows cheap and lends dear. Its net interest margin is that spread, and you can predict the rough size of it from the bank's structure alone, before you ever look at the margin itself.

By the end of this page you should be able to look at a bank's loan mix and deposit base, with the margin covered up, and call its net interest margin to within a point. Then we will mark the exact line where structure stops predicting and judgment takes over.

The machine, in one picture

A bank is a simple machine. Money comes in as funding, goes out as earning assets, and the bank keeps the spread between what the assets yield and what the funding costs.

FUNDING INEARNING ASSETS OUTSPREAD KEPT deposits + borrowings loans + securities interest earned (what you pay for) (what you earn on) − interest paid = net interest income

Net interest margin (NIM) is that net interest income expressed as a percentage of the earning assets that produced it, annualized. It is the single most mechanical number on a bank's income statement, because it is set almost entirely by two structural choices the balance sheet already tells you.

NIM = asset yield − funding cost
(interest income ÷ earning assets) − (interest expense ÷ earning assets)

Two levers move those two terms. Learn to read each one off the balance sheet and you can predict the margin.

Lever 1 · Earning-asset mix — sets the yield

Loans yield more than securities. Within loans, commercial and consumer credit yields more than residential mortgages. A bank that puts 80% of its assets into loans earns a high asset yield almost by construction. A bank that parks a third of its assets in bonds earns a low one. So: loan-heavy means high yield; securities-heavy means thin yield.

Lever 2 · Funding mix — sets the cost

Not all deposits cost the same. Non-interest-bearing deposits, ordinary checking balances, cost the bank essentially nothing. Interest-bearing savings and CDs cost real money, and borrowings cost the most. A bank funded heavily by non-interest-bearing checking pays almost nothing for its money. A bank funded by CDs and borrowings pays up. So: cheap deposits mean low cost; CD-funded means high cost.

The two levers compound. A loan-heavy bank funded by free checking earns a fat yield and pays almost nothing, so its margin is wide. A securities-heavy bank funded by CDs earns little and pays a lot, so its margin is thin. Most banks sit between. To predict a bank's NIM, read both levers off the balance sheet and net them.

A rough price list

To estimate a margin you need a feel for what each piece earns and costs. Below are typical ranges for the elevated-rate environment of early 2026. Treat them as orientation, not precision.

What assets earnTypical yield
Cash & fed funds sold4 – 5%
Treasury & agency securities2.5 – 4%
Residential mortgages4 – 5.5%
Commercial real estate5.5 – 7%
C&I (commercial)6.5 – 8%
Agricultural6.5 – 8%
Consumer & auto7 – 9%
Credit card (specialty)12%+
What funding costsTypical cost
Non-interest-bearing deposits~0%
Savings & interest checking0.1 – 1.5%
Money market1.5 – 3%
Retail CDs (time deposits)3.5 – 4.5%
FHLB advances / borrowings4 – 5%
Brokered deposits4.5 – 5%+

Read down those two columns and the levers turn into arithmetic. A bank with most of its assets in commercial real estate and C&I, funded by 40% free checking, is stacking a 6%-plus yield on a sub-1.5% cost. A thrift holding residential mortgages and low-coupon bonds, funded by CDs, is putting a 4%-ish yield against a 2%-plus cost. The structure sets the margin before management does anything.

The whole-universe figures confirm the anchors. Across the roughly 3,500 banks above $100M in this dataset for Q1 2026, the median bank's earning assets yielded 5.5% and cost 1.7% to fund, for a net interest margin of 3.8%, with the middle half running 3.3% to 4.3%. Split the universe and each lever is worth about 1.3 points on its own: loan-heavy banks yielded a median 6.0% against 4.7% for the securities-heavy, and cheap-deposit banks paid 1.1% against 2.3% for the CD-funded.

The cycle moves every number in this table. These are early-2026 rates, with the policy rate elevated. In 2021, when the Fed's rate sat near zero, the same banks yielded roughly 3.5% on assets and paid almost nothing, close to 0.3%, to fund. The ranges shift up and down with the rate cycle, and not in lockstep: a bank whose deposits reprice faster than its assets watches its margin compress as rates rise, while one with cheap, sticky deposits and floating-rate loans watches it widen. Use the price list to judge which bank earns the wider margin, not to nail an exact figure years from now. The ranking holds across the cycle. The absolute basis points do not.
Worked case — both levers pulling the same way

Here is the structure of a real bank, taken from its Q1 2026 FDIC call report. The margin is covered. Read the two levers and commit to a number before you open the reveal.

Grandview, TX · FDIC 3230
$703M assets · Q1 2026
Earning-asset mix (Lever 1)
Loans, % of assets77.7%
Securities, % of assets8.2%
Loan mix: real estate / C&I / other63 / 33 / 4
Funding mix (Lever 2)
Non-interest-bearing, % of deposits41.9%
Interest-bearing, % of deposits58.1%
Core deposits, % of deposits96.4%
Loan / deposit ratio87.5%
Predict. Lever 1: loans are 78% of assets, with a third of the book in higher-yielding C&I. High asset yield. Lever 2: 42% of deposits pay no interest, and the funding is almost entirely core, not brokered. Low cost. Both levers point the same way, up. Your NIM guess should be high. Write it down.
Reveal the margin
ComponentQ1 2026, annualized
Asset yield (interest income ÷ earning assets)6.61%
Funding cost (interest expense ÷ earning assets)−1.43%
Net interest margin5.18%
Reported NIM (FDIC NIMY)5.22%

5.22%. Top decile for a US community bank. Both levers delivered exactly as the structure promised. The 78% loan book, tilted toward commercial credit, produced a 6.6% asset yield. The 42% slug of free checking held the funding cost to 1.4%, even in a high-rate environment. The wide margin is not a surprise once you read the balance sheet. It is the mechanical consequence of it.

The small gap between the 5.18% we derived and the 5.22% the FDIC reports is just averaging: the FDIC divides by average earning assets over the quarter, we used the period-end figure. The decomposition reconciles to within a rounding error, which is the point. The margin is not mysterious. It is the yield minus the cost, and you read both off the structure.

The mirror image

To prove the framework rather than the example, run it backwards on a bank built the opposite way: a conservative New Hampshire mutual savings bank.

Portsmouth, NH · FDIC 17443 · Mutual
$381M assets · Q1 2026
Earning-asset mix (Lever 1)
Loans, % of assets53.3%
Securities, % of assets33.0%
Loan mix: real estate / other99 / 1
Funding mix (Lever 2)
Non-interest-bearing, % of deposits2.6%
Interest-bearing, % of deposits97.4%
Loan / deposit ratio67.6%
Reveal the margin
ComponentQ1 2026, annualized
Asset yield (interest income ÷ earning assets)4.01%
Funding cost (interest expense ÷ earning assets)−2.38%
Net interest margin1.63%
Reported NIM (FDIC NIMY)1.62%

1.62%, against Grandview's 5.22%. The same framework explains a 3.6-point gap with no appeal to management quality or luck. Lever 1 worked against Piscataqua: only half its assets are loans, and those loans are 99% residential real estate, the lowest-yielding kind, so the asset yield is just 4.0%. Lever 2 worked against it too: almost none of its deposits are free checking, so it pays 2.4% for its money. Low yield minus high cost equals a thin margin. Both levers pointed down, and the margin sat where they put it.

Hold the two banks side by side and the decomposition does all the work. The yield gap (6.6% vs 4.0%) comes from the asset mix. The cost gap (1.4% vs 2.4%) comes from the funding mix. Add them and you have the whole 3.6 points. Nothing else is needed to explain the difference in margin.

Your turn — predict, then reveal

A third bank, structure only, Q1 2026. This one does not have both levers pulling the same way, so you will have to net them. Commit to a number and a one-line reason before you reveal.

Hoquiam, WA · FDIC 28453
$2.0B assets · Q1 2026
Earning-asset mix (Lever 1)
Loans, % of assets71.9%
Securities, % of assets10.3%
Loan mix: real estate / C&I / other92 / 9 / 0
Funding mix (Lever 2)
Non-interest-bearing, % of deposits23.4%
Interest-bearing, % of deposits76.6%
Core deposits, % of deposits89.5%
Loan / deposit ratio83.3%
Work it. Lever 1: loan-heavy at 72%, which argues for a good yield, but the book is 92% real estate with little commercial credit, which pulls the yield back toward the middle. Lever 2: 23% free checking is solid, better than a thrift, short of Grandview's 42%, so funding is cheap but not free. Good asset side, good funding side, neither exceptional. Net them. What is your NIM?
Reveal the margin + score yourself
ComponentQ1 2026, annualized
Asset yield (interest income ÷ earning assets)5.35%
Funding cost (interest expense ÷ earning assets)−1.59%
Net interest margin3.76%
Reported NIM (FDIC NIMY)3.80%

3.80%. Squarely between the other two, exactly where the structure says it should be. The loan-heavy balance sheet gave a 5.4% yield, below Grandview's 6.6% because the book is residential real estate rather than commercial. The 23% slug of free checking gave a 1.6% funding cost, above Grandview's 1.4% but well below the thrift's 2.4%. Good yield, good cost, neither extreme. A margin in the high 3s.

3.4% – 4.2% — You read both levers and netted them. This is the win.
4.5%+ — You over-weighted the loan-heavy asset side and forgot the book is residential, not commercial, and that funding is not free.
under 3.0% — You under-weighted a genuinely loan-heavy bank with solid core deposits.

The exact 3.80% was never the target. A range in the high 3s was. That gap between the range you can predict and the basis point you cannot is the whole subject of the next section.

Where the numbers go silent

Everything above is the measurable part, and it is most of the margin. Structure sets the range. Read the two levers and you can rank any bank's structural margin potential and predict its NIM to within a point. That is real, and most people never learn to do it.

But the framework predicts a range, not a number, and the gap between the two is exactly what structure cannot see:

What NIM does not tell you

So the honest statement of what you have learned: the numbers tell you the structural earning power of a bank's spread machine, and they tell you with enough precision to predict the margin to within a point. They do not tell you whether management is pricing well, whether the yield hides risk, or whether the margin will last. Those take the later lessons, and in the end, judgment. Knowing exactly where that line sits is most of what separates reading a bank from guessing at one.

Next: No. 2 — The income cascade, where this margin becomes the first line of the return on assets.