Net interest margin
What a lender earns on its assets minus what it pays for its funding, as a percentage of the assets. The spread, expressed as a rate.
A bank borrows at one price and lends at another. Net interest margin is the gap, stated against the earning assets rather than against equity, which makes it comparable between a small bank and a large one. It is the single most informative number about a lending business, because it tells you whether the model works before any question of scale.
SoFi ran a margin of 5.98% in the quarter we examined, producing $788,200,000 of net interest income against $45,500,000,000 of deposits. For context, a large diversified bank typically runs well under half that, because it funds itself with cheap institutional money and lends against safer collateral. A high margin is not automatically good news: it usually means the loans are riskier, and the question becomes whether the extra spread covers the extra losses.
Margin also moves with rates in ways that depend on the shape of the balance sheet. A lender whose loans reprice faster than its deposits gains when rates rise and gives it back when they fall. That asymmetry is a forecast embedded in the balance sheet, and it is why two lenders with the same margin today can have very different margins in two years.
Margin against credit cost, in practice: the SoFi report.
Net charge-offs
Loans the lender has given up on, written off against the balance sheet, net of anything later recovered. The realized cost of lending.
Every lender loses money on some loans. A charge-off is the moment it stops pretending otherwise and removes the loan from its assets. Net charge-offs subtract whatever is eventually collected from borrowers who were written off, so the figure is the true realized loss rather than the gross one.
Quoted as a rate against average loans, it is the number that has to be set against net interest margin. JPMorgan charged off $2.4B in the quarter we examined. A lender running a margin of six percent and losing four percent a year to charge-offs is a worse business than one running three percent and losing half of one.
The timing is the trap. Charge-offs are a lagging number: they describe loans made two or three years ago under credit standards that may no longer apply, which is why a lender growing its book quickly will always look better on this measure than it is. Growth dilutes the denominator before the losses arrive in the numerator.
Why a growing loan book flatters this number: SoFi and JPMorgan.
Provisions and the allowance
The charge a lender takes for losses it expects but has not yet suffered. An estimate, which makes it the most discretionary line in a bank's accounts.
The allowance is a reserve held against the loan book. The provision is the amount added to that reserve in a given quarter, and it runs through the income statement. So total credit cost has two parts: the losses actually realized, and the change in the estimate of future ones.
JPMorgan's credit costs of $2.5B in the quarter we examined were $2.4B of realized charge-offs plus $0.1B of reserve build. Separating those two is the first thing to do with any bank's results, because they mean opposite things. Realized losses are history. A reserve build is management telling you what it expects next, and a reserve release is management telling you the future looks better, which conveniently also raises reported earnings.
That discretion is why bank earnings are smoother than bank economics. A provision is a judgment call, it is made by people whose compensation depends on the result, and it is the most common place for a quarter to be quietly managed. Read the reserve ratio against the loan book rather than the earnings line.
What a one-time item does to reported profit generally: the one-time item that happens every year.
CET1 and capital ratios
The regulator's measure of how much real equity a bank holds against its risk. The constraint that decides what it can pay out.
Common equity tier one capital is roughly the bank's equity after removing the parts that would not absorb a loss, divided by risk-weighted assets rather than total ones. The weighting is the point: a Treasury bill and an unsecured consumer loan are not the same exposure, so they do not count the same in the denominator.
JPMorgan ran 14.1% in the quarter we examined, which is $303,000,000,000 of capital against $2,100,000,000,000 of risk-weighted assets. The minimum is set by regulation and the bank's own target sits above it; the gap between the two is the buyback capacity. That makes this ratio the most direct link between a bank's regulatory position and its shareholder returns.
For an equity holder it works in both directions. A bank comfortably above its requirement can return capital, which is most of the return in a mature bank. A bank near its requirement cannot, and a bank below it raises equity at the worst possible moment. The ratio is not a safety statistic; it is a forecast of whether you get paid.
What a forced equity raise costs the existing holders: leverage is a calendar.
Book value and tangible book value
What the balance sheet says the equity is worth, and the same figure with goodwill and intangibles removed. For a lender, the second one is the real one.
Book value per share is total equity divided by shares. It is a near-useless measure for most companies, because the assets of a software business or a brand are not on its balance sheet at anything like their value. For a bank it is the opposite: the assets are loans and securities carried at something close to a market figure, so book value means something.
Tangible book value strips out goodwill and other intangibles, which are the accounting residue of past acquisitions and would not absorb a loss. The gap can be large. JPMorgan's book value per share was $133 against tangible book of $113, so the stock traded at 2.66 times book and 3.12 times tangible book at the price we used. SoFi carried intangibles worth 14.4% of its equity.
Use the tangible figure. It is the denominator that pairs with return on tangible common equity, and pairing a return on tangible equity with a multiple of total book value is the most common arithmetic error in bank valuation.
The return that pairs with it: return on tangible common equity, and the JPMorgan report.
Justified price to tangible book
The multiple a lender deserves, derived from its returns rather than from what its peers trade at. The bank version of a discounted cash flow.
A bank's value comes from earning more on its equity than that equity costs. Put that in one line and you get the justified multiple: price to tangible book equals return on tangible common equity minus growth, divided by cost of equity minus growth. If the return equals the cost of equity, the fair multiple is exactly one times tangible book, whatever the growth rate. Everything above one is the spread between what the equity earns and what it costs.
Run it on SoFi. A guided return on tangible common equity of 8.7%, a cost of equity of 11% and long-run growth of 6% gives a justified multiple of 0.54 times tangible book. The shares traded at 2.28 times. That gap is the entire argument of the report, stated in one calculation, and it does not depend on a single peer multiple.
The formula is also a useful warning system. It is extremely sensitive near the bottom, because the denominator is a difference between two similar numbers, so a lender growing fast with a return only slightly above its cost of equity has a fair value that moves enormously on small changes in assumptions. JPMorgan earning 29.0% on tangible equity, or 23.0% excluding one-time items, sits in the comfortable part of the curve. A lender earning nine percent does not.
The method applied: SoFi, with the general version in the reverse DCF and the input in cost of equity.
DISCLAIMER ·
Education, not advice. Every example above is simplified to expose the mechanics. Real returns are volatile,
taxes and fees vary by country and account, and none of this considers your circumstances. 2 Comma Investor is not a
registered investment adviser or broker-dealer. Do your own research and consult a licensed professional before
acting.
Full disclosures →