Sector deep dive: banks

Definition

Banks are leveraged spread businesses: they borrow short (deposits), lend long (loans, securities), and earn the difference — which makes standard company analysis (fa-ratio-analysis) actively misleading. Revenue is interest arithmetic, the balance sheet IS the business, book value replaces enterprise value, and the equity is a thin slice atop 10x leverage whose value turns on credit losses and depositor confidence. Two KB episodes (2008, 2023) exist because this structure failed; this entry carries the analysis kit those failures built.

How it works / structure

• The earnings engine: NET INTEREST MARGIN (NIM — asset yield minus funding cost) × earning assets, plus fee income, minus credit costs (PROVISIONS) and operating expense (the EFFICIENCY RATIO — cost/income); rate sensitivity runs both ways (regime-rate-environments): rising rates lift asset yields but reprice deposits and mark securities down (the 2023 mechanism).
• The balance-sheet kit (engine-parameterizable): CET1 ratio (regulatory capital vs risk-weighted assets — the solvency headline, DFAST-tested); loan mix and NPL/charge-off rates (credit quality); deposit composition — the post-2023 checklist (episode-banking-stress-2023): uninsured share, concentration, rate-seeking beta; securities marks (AFS hits capital now, HTM hides until sold); loan-to-deposit ratio (funding self-sufficiency).
• Valuation grammar: price-to-tangible-book vs return on tangible equity (the sector’s multiple-vs-return regression is the standard frame — banks earning above their cost of equity trade above book, below-cost banks below); P/E on normalized provisions, never peak-cycle earnings.
• The public-data advantage: call reports and FDIC data (UBPR) publish standardized quarterly detail no other sector matches; DFAST results publish stressed loss rates per bank — risk-scenario-analysis inputs handed over publicly.

When it applies

Any financial-sector position (the generic toolkit misprices banks structurally); rate-regime theses expressed via banks (the NIM transmission is the trade — with the deposit-beta caveat); stress screening (2023-style fragility is measurable from filings); credit-cycle positioning (provisions lag the cycle — reserve builds and releases are the sector’s earnings-quality axis, fa-earnings-quality).

Risk profile & failure modes

• Confidence dependence: a solvent-looking bank can fail on funding in days (2023’s lesson — run speed is now digital); equity analysis without liability analysis is half the balance sheet.
• Opacity of credit: loan books are graded by the bank itself until losses emerge — provisions are management estimates (fa-earnings-quality applies with force); credit surprises cluster late-cycle.
• Rate-sensitivity ambidexterity: “banks benefit from rising rates” holds only while deposit betas lag and marks don’t force sales — 2023 falsified the slogan’s naive form.
• Regulatory regime shifts: capital rules rewrite the return math sector-wide — a standing exogenous variable other sectors don’t carry.

Evidence & limits

The metric kit is regulatory-standard (Fed/FDIC documented); DFAST methodology and results are public; the failure mechanics carry the 2008/2023 records. The price-to-book/ROTE regression is documented sector convention (labeled). Per-bank credit-book truth remains partially unobservable until stress — stated as the sector’s structural analytic limit.

Falsifiable-thesis examples

Illustrations only, not signals:

• “Bank X’s NIM expands 30bp+ over two quarters as its deposit beta stays below 40% in this hiking cycle (transmission thesis)” — falsified by the reported NIM and beta.
• “Banks in the cheapest price-to-tangible-book quintile with above-median CET1 and below-median uninsured deposits outperform the sector over 12 months (quality-value screen)” — falsified by the cohort return.

Cross-references

• The generic frame it replaces: fa-ratio-analysis, fa-sector-metrics
• The failure records: episode-gfc-2008, episode-banking-stress-2023
• The macro driver: regime-rate-environments
• The stress machinery: risk-scenario-analysis (DFAST)