Skip to main content

Loan Growth & Credit Quality

Updated Jun 24, 2026 at 8:22pm

Research Draft Medium 1,242 words

For a bank, the loan book is the core earning asset, and the central tension of bank analysis is that loan growth and loan quality trade off against each other over the cycle. Any bank can grow loans quickly — simply by loosening underwriting, cutting price, or chasing weaker borrowers — but those loans tend to default later, after the growth has already flattered earnings and the stock. The analyst's job is therefore not to celebrate fast growth but to ask whether it is being bought with deteriorating credit standards, and to track the lagging indicators (delinquencies, non-performing loans, charge-offs, reserves) that reveal the answer one to several quarters after the loans were booked.

The metrics and how they're calculated

Credit quality is read through a stock-vs-flow pair plus reserve-adequacy measures:

  • NPL / noncurrent ratio (stock) = non-performing loans ÷ total loans. A loan is "non-performing" / "noncurrent" when it is 90+ days past due or on non-accrual status (the bank no longer expects full repayment on schedule). Delinquencies are usually disclosed in buckets — 30–89 days past due (early warning) and 90+ days / non-accrual (problem).
  • Net charge-off (NCO) rate (flow) = (gross charge-offs − recoveries) ÷ average total loans, annualized. This is realized loss actually written off.
  • ACL coverage = allowance for credit losses ÷ total loans (reserve as a cushion against the whole book).
  • NPL coverage = allowance ÷ non-performing loans (reserve against identified problem loans; above 100% means reserves exceed flagged bad loans).
  • Texas ratio = (NPLs + OREO/foreclosed real estate) ÷ (tangible common equity + ACL) — a solvency-stress gauge.

The accounting plumbing matters. The provision for credit losses is the income-statement charge that builds the allowance: Beginning ACL + Provision − Net Charge-offs = Ending ACL. Since 2023 (2020 for large filers) U.S. banks reserve under CECL — they book lifetime expected losses at origination rather than waiting for losses to become probable, which front-loads reserves on new loans. This is why CECL banks typically carry higher allowance-to-loans ratios than IFRS-9 European peers, a key adjustment in cross-border comparisons (OCC; PwC).

How it's used in practice

Analysts read these as a chain that runs in sequence: early-stage delinquencies (30–89 days) → non-performing loans → net charge-offs → provisions/reserve build. As a matter of accounting sequence a loan is generally flagged noncurrent (90+ days / non-accrual) before it is actually charged off, so net charge-offs lag NPL formation by several quarters and a rising NPL ratio today is a forward signal of higher charge-offs ahead. The most informative read is the trajectory and mix, not the level: which portfolios (credit card, C&I, construction, commercial real estate) are deteriorating, whether reserve build is keeping pace with NPL formation, and whether management is releasing reserves to manufacture EPS late in a benign cycle.

Loan growth is assessed for quality, not just quantity. Useful questions: Is growth concentrated in higher-risk segments (construction, leveraged/subprime, non-owner-occupied CRE)? Is the bank winning share by cutting price or covenants? Is reserve build (provision) tracking loan growth, or is the bank "growing into" a thinning allowance? Because the provision is the single most volatile line on a bank's income statement, it is also the primary lever for normalized-earnings analysis — analysts often anchor a "through-cycle" normalized NCO/provision assumption (for a diversified bank in stable conditions this is frequently cited in the rough vicinity of half a percent of average loans, but the right number is heavily mix-dependent — see the base-rate section) rather than trusting a depressed current provision.

Adoption, debate & evidence

These metrics are universal, standardized in U.S. regulatory filings (Call Reports, FFIEC), and aggregated by the FDIC. They are not contested as measurements. What carries real, peer-reviewed weight is the fast-loan-growth-predicts-poor-performance finding. Fahlenbrach, Prilmeier & Stulz (Why Does Fast Loan Growth Predict Poor Performance for Banks?, NBER w22089 / Review of Financial Studies, 1973–2014 sample) found that banks in the top quartile of three-year loan growth subsequently underperformed bottom-quartile banks by a benchmark-adjusted cumulative >12 percentage points over the next three years. The mechanism: fast growers under-reserve while growing, then suffer lower ROA and have to build reserves later. Strikingly, the banks, analysts, and investors did not appear to recognize the lending was riskier — the loans were systematically mispriced, not knowingly risky (NBER; Harvard CG summary). Asset growth other than loan growth, and growth via mergers, did not show the effect — the signal is specific to organic lending.

This aligns with the historical pattern that most major U.S. bank failures (Continental Illinois 1984, Washington Mutual 2008) followed reckless loan expansion, and with Fitch's recurring warnings that rapid lending growth raises later credit-issue risk (Investment Executive / Fitch).

Honest base rates / "folklore vs measured": Aggregate U.S. asset quality has been benign recently — FDIC Q3 2025 reported ACL-to-noncurrent-loans of 178.4% (down from 179.4% the prior quarter) and an overall noncurrent loan ratio of 1.49%, below the stated pre-pandemic average of 1.94% (FDIC Q3 2025 QBP). But the average masks pockets: large-bank (>$250B assets) non-owner-occupied CRE PDNA (past-due-and-nonaccrual) ran 4.18% in Q3 2025 — down from a 4.99% peak in Q3 2024 but well above the ~0.59% pre-pandemic norm — and the industry quarterly net charge-off rate was 0.61%. For historical scale, the industry total noncurrent rate peaked around 5.5% in early 2010 during the financial crisis (per FDIC QBP series). Card net charge-offs run materially higher than mortgage/C&I in every cycle (the all-bank credit-card NCO rate hit a record ~10.5% in late 2009 and a record-low ~1.6% in 2021, per FRED CORCCACBS) — so all benchmarks must be read relative to portfolio mix, never absolutely.

Strengths & limitations

These are among the highest-signal fundamentals in all of equity analysis: the loan-growth/under-reserving relationship is one of the more robust predictive findings about bank stocks. Limitations: every credit metric is backward-looking and lagged — by the time NPLs spike, the bad underwriting happened years earlier, and the stock may have already de-rated. CECL makes reserves more forward-looking but also more judgmental and model-driven, giving management discretion to flatter or smooth earnings via reserve releases. Cross-cycle and cross-bank comparisons are corrupted by portfolio mix and accounting regime. The #1 misuse: treating strong loan growth as an unambiguous positive. The evidence says the opposite — rapid organic growth with lagging reserve build is a warning, not a virtue, and clean current credit metrics late in a benign cycle tell you almost nothing about the next downturn.

Sources

Flagged disputes: "Normal" benchmark levels are cycle- and mix-dependent — there is no single correct NPL or NCO threshold, only ranges relative to portfolio mix. The through-cycle normalized NCO figure (~0.5% area) is an industry rule-of-thumb, not a measured constant. Aggregate "benign" credit metrics carry low predictive value for the next downturn.