For informational & educational purposes only — not investment advice
Education · 05 · Sector

How to analyse a bank.

16 min read Last updated April 2026 Sector

A bank is not an ordinary business. It is a leveraged balance sheet with a branded front door — and it is only worth owning when it is exceptionally well run and exceptionally well priced.

Banks operate on fractional reserves. Liabilities typically run around ten times equity, compared to roughly three-to-one for a consumer staples business and less than one-to-one for a software company. That leverage is what makes banking profitable in good years and dangerous in bad ones. It is also what makes a bank run possible: depositors ask for their cash back simultaneously, the balance sheet cannot liquefy quickly enough, and the institution fails.

Bank equities carry three structural disadvantages: low growth, high cyclicality, and long-run underperformance versus broad equity benchmarks. A handful of high-quality franchises do outperform. The framework below is designed to identify those — and to insist on a deep discount before any capital is deployed.

Discipline

Banks are only bought when they are both high quality and significantly undervalued. Buying a bank under ordinary conditions is rarely worth the risk. The one legitimate exception is a conservative regional franchise held primarily for tax-efficient dividend income.

01Quality — ten years of numbers

Revenue and net profit

Start with a decade of financials. A quality bank shows a general uptrend in revenue and net profit across at least one full economic cycle. Profits will be cyclical — that is the nature of the business — but the trend line should be up. Inconsistent profits or declining revenue over five years or more is a red flag, and usually reflects either moat erosion, balance-sheet damage, or both.

Return on equity

ROE

Return on equity

ROE = Net Profit ÷ Shareholders' Equity
Target ≥ 10%, sustained

ROE measures how much profit the bank generates on each dollar of equity capital. Below 10% over a multi-year period usually means the bank is over-capitalised relative to its opportunity set, or that its underwriting and efficiency are structurally weak. Compare against regional peers, not global averages.

Dividend yield

Income profile

Yield = DPS ÷ Share Price
Target ≥ 5% at purchase

A 5% forward yield is a useful discipline — it forces the buyer to wait for a price at which the dividend stream alone is economically meaningful. Back out the share price at which yield would cross the threshold, and treat that as a working anchor. A high yield with declining earnings is a trap, not an opportunity.

Dividend growth

Five-year DPS growth

CAGR = (DPS₅ ÷ DPS₀)^(1/5) − 1
Ideally positive and aligned with revenue growth

Dividend growth should approximately track earnings and revenue growth. A rising payout against flat or declining revenue is unsustainable. A flat dividend against rising revenue usually signals either balance-sheet repair or management conservatism — read the filings to tell which.

Revenue / DPS correlation

Coherence check

Δ Revenue ≈ Δ DPS (over 5y)
Aligned over a full cycle

Over a full cycle, revenue and dividend growth should be directionally consistent. Divergence — particularly a dividend rising faster than revenue for multiple years — often precedes a cut. Treat alignment as a baseline requirement, not a positive signal.

02Safety — four bank-specific risks

A bank's balance sheet is more informative than its income statement. The four risks below — liquidity, market, credit, and reputational — are monitored through specific ratios. A failure on any one of them is disqualifying.

Risk A · Liquidity

Can the bank meet short-term demands for cash?

Liquidity risk is the risk that the bank cannot satisfy withdrawal or funding demands in the near term — the direct cause of most bank failures.

CET1 ratio

Core equity capital

CET1 = Common Equity Tier 1 ÷ Risk-Weighted Assets
Regulatory min 4.5% · Target > 10%

CET1 measures the highest-quality capital — common shares and retained earnings — against the bank's risk-weighted exposures. A ratio comfortably above 10% indicates a real cushion; below 10% indicates one the regulator has let pass but the market should not.

LCR

Liquidity coverage ratio

LCR = High-Quality Liquid Assets ÷ 30-Day Net Outflows
Target ≥ 100%

LCR measures whether the bank has enough high-quality liquid assets to fund 30 days of stressed outflows. Note: smaller regional US banks are not always required to report or pass this ratio — flag any such institution as inherently less safe than its large-bank peers.

Risk B · Market / Interest Rate

How exposed is the balance sheet to rates and markets?

Market risk is the risk that assets held on balance sheet — principally bonds — lose value when rates rise or when markets dislocate.

Securities / Assets

Securities concentration

Securities ÷ Total Assets
Target < 40%

A high securities-to-assets ratio means a large portion of the bank's earning assets are bonds and similar instruments. When long-dated bonds are marked down in a rate shock, the equity absorbs the loss. Ratios above 40% materially raise fragility; conservative franchises often operate in the 12–25% range.

Loans / Deposits

Core business ratio

Loans ÷ Deposits
Healthy range 50–90%

Below 50% suggests the bank is under-lending and diverting deposits into riskier securities. Above 90% suggests over-extension and a reliance on non-deposit funding. Both extremes matter — a recent large-bank failure sat well below the range; a well-known Swiss failure sat well above it.

Risk C · Credit / Default

Are loans being repaid?

Credit risk is the risk that borrowers default on their loans — the slowest-moving of the four risks, and rarely the primary cause of modern bank failures, but still essential.

NPL ratio

Non-performing loans

NPL = Loans past due (typically 90d+) ÷ Total Loans
Lower is better; compare to peers

NPL measures the share of the loan book in trouble. What matters is the trend and the regional comparison, not the absolute number — some markets run structurally higher NPLs than others. A rising NPL ratio in a benign credit environment is a leading indicator of underwriting weakness.

Coverage ratio

Loan-loss provisions

Coverage = Loan-loss reserves ÷ NPLs
Target > 100%

Coverage measures whether the bank has already reserved against its problem loans. Above 100% means the losses, if they crystallise, are largely absorbed. Below 100% means future earnings will be consumed by provisions when the cycle turns.

Risk D · Reputational / Operational

Has trust been earned — and kept?

The hardest risk to quantify and the one that most often triggers failure. A loss of confidence can provoke a run irrespective of capital and liquidity ratios.

Read the ten-year history of each candidate. Scandals, regulatory actions, rate or market manipulation, money-laundering settlements, mis-selling, tax breaches, fraud — these compound into a reputational debt that no ratio captures. A bank with strong CET1 and LCR but a decade of serious misconduct can still fail when the market decides it will not extend credit any longer, as one large European institution discovered. Rate reputational risk qualitatively: low / moderate / high / critical. A "high" or "critical" rating should be disqualifying regardless of the rest of the scorecard.

03Valuation — price to book, with discipline

Bank earnings move with the cycle; book value moves much more slowly. That is why price-to-book, not price-to-earnings, is the primary valuation anchor for a bank. The method below forces both a fair-value reference and a strict buy threshold.

Step 1

Ten-year average P/B

Avg P/B = mean(P/B for each year, last 10y)

Establishes the bank's own valuation history across at least one full cycle. This is the reference, not a peer average.

Step 2

Fair value

Fair Value = Avg P/B × Current Book Value / Share

The price at which the bank trades at its own historical average multiple. Trading at fair value is not a reason to buy.

Step 3

One standard deviation

σ(P/B) = standard deviation of the 10-year P/B series

Measures how far from the average the multiple has historically drifted. Gives the dispersion of sentiment around this specific bank.

Step 4

Buy target

Buy Target = (Avg P/B − 1σ) × Current Book Value / Share

The price at which the bank is trading at the low end of its historical range — the disciplined entry point.

Current price vs. Reading Action
Above fair value Overvalued on its own history Do not buy
Near fair value Fairly valued — no margin of safety Do not buy
Between fair value & buy target Undervalued, but not yet cheap enough Add to watchlist
At or below buy target Trading at one standard deviation below its own ten-year average Strong buy candidate
Regional exception

Conservative regional franchises — particularly those with structurally low securities-to-assets ratios, strong deposit bases, and tax-efficient dividend regimes — may occasionally be bought modestly above fair value, but never above one standard deviation above the ten-year average P/B.

04Summary scorecard

Every bank analysis closes with the same verdict table. The standards are binary: a metric either clears the bar or it does not.

Metric Threshold Pass / Fail
10-year revenue trendGrowing
Return on equity≥ 10%
Dividend yield at purchase≥ 5%
CET1 ratio> 10%
Liquidity coverage ratio≥ 100%
Securities / assets< 40%
Loans / deposits50–90%
NPL ratio (trend & level)Low, stable
Reputational riskLow – Moderate
Current P/B vs. buy targetAt or below

Four verdicts follow:

05Rules when in doubt

Takeaways

Seven rules for buying a bank

  • Banks are leveraged balance sheets. Only the strongest are worth owning, and only at a discount.
  • Insist on ten years of growing revenue and net profit and a sustained ROE above 10%.
  • Require a forward yield of at least 5% at purchase, with dividend growth aligned to revenue growth.
  • Demand a CET1 ratio above 10% and an LCR at or above 100%. Flag banks exempted from LCR reporting.
  • Watch the securities-to-assets ratio (< 40%) and the loan-to-deposit ratio (50–90%). Both extremes are warnings.
  • Reputational damage is disqualifying. A decade of scandals is a balance sheet the regulator has not yet recognised.
  • Buy only when price is at or below the ten-year average P/B minus one standard deviation. Otherwise wait.
Next in the series How to analyse a REIT →
This article is published for informational and educational purposes only. It does not constitute investment advice, a recommendation, or an offer to buy or sell any security. Regulatory thresholds referenced are illustrative and vary by jurisdiction. Examples are historical. Past performance is not indicative of future results. All investors should consider their own circumstances and seek qualified professional advice before acting on any information contained here.