A bank is not an ordinary company that happens to be in finance — it is a leveraged spread machine wrapped in regulation, and that makes the sector its own analytical dialect. A bank funds itself mostly with other people's money (deposits and borrowings) and earns the difference between what it pays for that funding and what it earns lending and investing it. Because the balance sheet is the business, bank analysis inverts the usual equity playbook: the balance sheet, not the income statement, is the primary anchor; debt is raw material rather than a risk flag; and the stock is typically valued on price-to-tangible-book (P/TBV) and return on tangible common equity (ROTCE) rather than on EV/EBITDA or revenue multiples. This section is the playbook for reading that dialect. Its core tension runs through every sub-topic: the same three forces that drive a bank's profit — leverage, maturity transformation, and credit extension — are the exact forces that destroy banks when a cycle turns.

What this section covers

The four levers below are not independent metrics; they are one interlocking system seen from four angles. A bank's earnings are a spread (net interest margin), bought with risk (loan growth and credit quality), exposed to rates (rate sensitivity), and backstopped by a cushion (capital). Analyze any one in isolation and you will misread the bank — the recurring lesson of the children below, and of every modern bank failure.

• Net Interest Margin (001) — the core profitability gauge: net interest income over average earning assets, typically ~3–4% for a traditional U.S. lender (FDIC industry NIM was 3.22% in FY2024). NIM is the spread the whole business turns on. The child doc's key correction: "higher rates widen NIM" is empirically weak in the short run — yield-curve slope and deposit beta dominate the level of rates, and a wide NIM can be uncompensated credit risk that hasn't defaulted yet.

• Loan Growth & Credit Quality (002) — the asset side, where growth and quality trade off across the cycle. Covers NPLs, net charge-offs, allowance coverage, and the CECL accounting plumbing. Carries the section's most robust academic finding: Fahlenbrach, Prilmeier & Stulz (1973–2014, U.S. banks) found top-quartile loan growers subsequently under-performed bottom-quartile growers by a benchmark-adjusted cumulative margin that exceeds 12 percentage points over three years — and that the effect is specific to loan growth (not asset growth) and not driven by mergers. Fast organic loan growth is a warning, not a virtue.

• Rate Sensitivity (003) — how a bank's earnings and economic value respond to rate moves, depending on which side of the balance sheet reprices faster (asset- vs. liability-sensitive). Distinguishes the earnings view (NII sensitivity) from the value view (EVE / duration). Its defining lesson — earnings sensitivity and value sensitivity are different risks — is the 2023 SVB failure: NII looked fine while a ~$15bn unrealized securities loss hollowed out economic equity.

• Capital Ratios & Regulation (004) — the loss-absorbing cushion and the Basel/CCAR rules that govern it (CET1, Tier 1, RWA, leverage/SLR, the Stress Capital Buffer, the unsettled Basel III "Endgame"). For an analyst this gates dividends, buybacks, and growth. Its hard caveat: a passing capital ratio is necessary but not sufficient — SVB was nominally "well-capitalized" shortly before it failed.

A note on scope: siblings reference a deposit franchise lens (deposit beta, deposit stickiness, the franchise's value as cheap, sticky funding) that the NIM and rate-sensitivity docs treat inline. If a standalone deposit-franchise node is added later it belongs here; for now its mechanics live within 001 and 003. The section also defers general fundamentals (leverage, book value, ROE) to the core fundamentals branch and the rate/yield-curve macro lens to the macro/regime branch — analyze them together rather than duplicating.

Why banks need their own playbook

Three structural features make generic equity analysis fail on banks:

1. Extreme, intentional leverage. Banks run with thin equity against a large balance sheet by design, so small percentage moves in asset value swing equity dramatically. This is why capital ratios and unrealized securities losses can matter more than the income statement. 2. Maturity transformation. Borrowing short and lending long is the source of both the margin and the fragility — it is what creates rate sensitivity and run risk simultaneously. 3. Heavy regulation and accounting specificity. Basel capital rules, CECL reserving, CAMELS supervision, and stress tests are not background noise — they directly set payout capacity and reported earnings. Bank earnings are also dominated by two discretionary, model-driven lines (the loan-loss provision and RWA), which management can use to flatter or smooth results.

Because of this, banks are valued differently: the standard FIG relative-valuation tool regresses P/TBV on forward ROTCE across a peer group, and as a matter of pricing logic a bank earning exactly its cost of equity tends toward ~1.0x tangible book. How well that regression fits is contested — FIG practitioner guides cite R² in the ~47–68% range for U.S. peer groups (roughly half to two-thirds of the cross-sectional variation), but other analyses of the same relationship report R² as low as ~0.11. Treat ROTCE as the dominant but not sole driver of bank multiples; franchise quality, deposit mix, and credit cycle position all move a name off the line.

When this section matters — and when it doesn't

This playbook is load-bearing whenever you analyze a bank, thrift, or regional/community lender, and partially relevant to diversified financials with large balance sheets. It matters most around scheduled catalysts: earnings (NIM guidance and deposit-beta assumptions), the annual CCAR/SCB stress-test results (payout capacity), and rate decisions / yield-curve shifts.

It is less applicable to fee-, trading-, or asset-light financials — insurers, asset managers, exchanges, payment networks, and most fintechs — which earn on float, fees, or transaction volume rather than a deposit spread and are valued on growth and margin, not P/TBV. Forcing the bank dialect onto those names is a category error.

Sources

• FDIC, Quarterly Banking Profile (industry NIM, noncurrent and charge-off data) — fdic.gov
• Fahlenbrach, Prilmeier & Stulz, Why Does Fast Loan Growth Predict Poor Performance for Banks? — NBER w22089
• Federal Reserve, Material Loss Review of Silicon Valley Bank (2023) — earnings vs. economic-value sensitivity
• Bank for International Settlements — Basel III capital framework; Federal Reserve — stress tests / SCB and Basel III Endgame materials
• Drechsler, Savov & Schnabl, How to Value the Deposit Franchise and Banking on Deposits (J. Finance, 2021); Drechsler, Savov, Schnabl & Wang, Deposit Franchise Runs (NBER w31138) — deposit-spread valuation, deposit beta
• FIG practitioner guides (ibinterviewquestions.com) — ROTCE/P-TBV valuation, ROE-to-P/TBV regression (cite R² ~47–68% for U.S. bank peer groups). DISPUTED: independent backtests (e.g. fructivore.com) report R² as low as ~0.11 on similar data — fit is unstable and sample-dependent; the ~47–68% figure is a practitioner claim, not an established academic result.

Section overview only — each claim is verified in depth in the four child docs (001–004); see them for full sourcing and flagged disputes.