Price-to-Earnings for Financials: Normalizing for Credit Cycles

Introduction

Price-to-earnings (P/E) is the most widely used valuation multiple in equity markets, but applying it to banks without adjustment produces misleading results. The problem is fundamental: bank earnings are dominated by two highly cyclical line items, net interest income (which fluctuates with the rate environment) and loan loss provisions (which swing dramatically across credit cycles), and are regularly distorted by large one-time charges. JPMorgan's trailing P/E of approximately 16x in early 2026 reflects 2025 earnings that include both normalized provisions and a favorable rate environment. But in Q4 2023, a $2.9 billion FDIC special assessment depressed quarterly earnings, while in 2021, billions in provision releases inflated reported EPS well above sustainable levels. Neither period's raw P/E captured the bank's true earning power. For FIG analysts, normalizing bank P/E for credit cycles and one-time items is not optional; it is the minimum standard for a credible valuation.

The core issue is provision cyclicality. Loan loss provisions are inherently procyclical: they fall during economic expansions (when credit quality is strong and few loans default) and spike during downturns (as defaults materialize and macroeconomic forecasts deteriorate). This means reported bank earnings systematically overstate true economic profitability during booms and understate it during busts.

During the pre-crisis period (2005-2007), major US banks reported peak earnings on minimal provisions. JPMorgan earned $15.4 billion in 2007, and the bank appeared "cheap" at 6-10x trailing P/E. But these low P/E multiples were a value trap: earnings were unsustainably inflated by near-zero credit losses during the housing bubble. When the crisis hit, JPMorgan's profits collapsed to $5.6 billion in 2008, and many banks reported outright losses, rendering P/E ratios either negative or astronomically high.

The pattern repeated during COVID. JPMorgan took approximately $20 billion in credit expense across just three quarters of 2020 (including $10.5 billion in Q2 2020 alone), crushing reported earnings. Then in 2021, as the feared credit losses failed to materialize, JPMorgan released over $9 billion in reserves, supercharging reported EPS well above sustainable levels. An analyst who valued JPMorgan on 2020 trailing P/E would have concluded the bank was severely overvalued; one who used 2021 trailing P/E would have concluded it was cheap. Neither was correct.

Mid-Cycle Earnings Normalization

Mid-cycle normalization replaces actual provisions with a "through-the-cycle" provision estimate that reflects average credit costs over a full economic cycle (typically 40-60 basis points of average loans for large diversified US banks, higher for consumer-heavy portfolios). The process: start with reported EPS, add back actual provision expense, subtract the normalized mid-cycle provision, strip out one-time items (legal settlements, restructuring charges, special assessments), tax-effect the adjustments, and arrive at "normalized" or "core" EPS. This normalized EPS represents the bank's sustainable, repeatable earnings power and forms the basis of credible P/E valuation. Pre-Provision Net Revenue (PPNR = net interest income + non-interest income - non-interest expense) is the related metric that isolates operating performance before credit costs entirely.

The transition from the incurred-loss model to CECL (Current Expected Credit Losses), effective January 2020 for large SEC filers, fundamentally altered how provisions flow through bank income statements. Under the old incurred-loss model, banks recognized credit losses only when it was "probable" a loss had been "incurred," creating systematic delays in loss recognition. Under CECL, banks must estimate lifetime expected credit losses at loan origination, incorporating forward-looking macroeconomic forecasts.

The day-one impact was substantial. Industry-wide CECL adoption reserves increased by an estimated $42 billion (44%). JPMorgan's allowance increased by $4.3 billion (30%), with credit card allowances nearly doubling (up $5.5 billion). Research has documented that allowance volatility increased 50% more for CECL adopters than non-adopters, and provision volatility increased 100% more, confirming that CECL amplified earnings swings relative to the old model.

For P/E normalization, CECL creates a specific challenge: when banks are growing their loan books (an expansion activity), provisions spike because every new loan carries a full lifetime reserve charge, even if credit quality is pristine. This can depress reported earnings during periods of strong growth, making fast-growing banks appear less profitable on a P/E basis than their underlying economics warrant. Conversely, when loan growth slows in a downturn, the day-one provisioning pressure eases even as actual credit losses increase, creating counterintuitive provision dynamics.

Beyond provision cyclicality, bank earnings are regularly affected by large, non-recurring items that further distort trailing P/E multiples:

The combined effect is that any single quarter's or year's reported EPS can deviate significantly from the bank's sustainable earnings power. FIG analysts systematically adjust for these items when constructing normalized P/E valuations.

The premium hierarchy in bank P/E multiples directly mirrors the ROE hierarchy that drives P/TBV multiples. JPMorgan commands the highest P/E (approximately 16x trailing) because its 17%+ ROE generates the strongest sustainable earnings per dollar of equity. Citigroup trades at the lowest P/E among the six largest US banks (approximately 12x trailing) because its lower ROTCE translates to lower sustainable EPS relative to tangible book value. Both multiples are telling the same story, just through different lenses: P/TBV prices ROE relative to cost of equity, while P/E prices earnings relative to the market's required return.

European banks trade at a stark discount on P/E, with forward multiples of approximately 6-8x versus 10-16x for US peers. This approximately 2x gap reflects the structural ROE difference between the two markets (European bank average ROE of approximately 10-12% versus 15-17% for top US banks), not a comparable valuation opportunity. European P/E multiples are compressed by lower profitability (a legacy of negative ECB interest rates from 2014-2022), fragmented markets limiting scale economies, and higher sovereign risk exposure. The ECB's IFRS 9 accounting framework introduces its own normalization challenges: the staged provisioning model (12-month expected losses for Stage 1 assets versus lifetime losses for Stages 2-3) creates different provision timing dynamics than CECL, making direct transatlantic P/E comparisons even more complex.

In practice, FIG analysts use P/E alongside P/TBV as complementary lenses on the same valuation question. P/TBV captures the ROE-to-cost-of-equity relationship that determines whether a bank creates or destroys value. Normalized P/E captures the earnings yield investors receive at the current price. When both multiples tell the same story (premium P/TBV and premium P/E for high-ROE banks), the valuation signal is strong. When they diverge (elevated P/TBV but depressed P/E, or vice versa), the divergence typically signals a one-time earnings distortion that normalization should resolve.

P/E remains an essential tool in the FIG valuation toolkit, but only when used with proper normalization. Together with P/TBV (which prices ROE relative to cost of equity) and the DDM (which values the present value of distributable cash flows), normalized P/E provides a third lens on bank valuation that is particularly useful for comparing banks with different capital structures, growth profiles, and accounting conventions.