CAPE vs forward P/E vs trailing P/E: three valuation measures, three horizons, three readings

Three equity valuation measures circulate continuously in financial commentary and are regularly conflated: the CAPE (P/E on 10-year real earnings), the trailing P/E (on GAAP earnings of the last 12 months) and the forward P/E (on consensus sell-side earnings projected over the next 12 months). Their levels diverge sharply and that very divergence carries distinct analytical information.

For the S&P 500 in June 2026, the CAPE stands at 40.2 per Robert Shiller’s data, the trailing P/E on reported earnings close to 26 (on a second quarter still estimated in June) and the forward P/E at 20.1 on 26 June per FactSet: a dispersion that does not reflect a technical disagreement but three distinct temporal optics on the same asset.

Three Constructions, Three Optics

Trailing P/E, abbreviated to P/E without qualifier in most analytical material, is the historically dominant measure. Its construction is mechanical: current S&P 500 price divided by the sum of GAAP earnings for the past four quarters (Trailing Twelve Months, TTM). It captures the immediate-past optic: how much the market currently pays for earnings actually realized over the last twelve months. Its interpretive simplicity explains its popularity, and its continuous publication by S&P Dow Jones Indices ensures the availability of a long historical series.

Forward P/E is built from sell-side projected earnings over the following twelve months. The figures are published weekly by FactSet from aggregated consensus estimates of analysts covering S&P 500 companies. The optic is anticipations: how much the market currently pays for earnings the consensus expects in the next twelve months. Forward P/E is less volatile than trailing P/E, anticipations move more slowly than realized earnings, but it embeds a documented bias I return to in section 3.

The CAPE is the third construction, formalized by John Campbell and Robert Shiller in 1988. Its optic differs from the other two: long-term retrospective. The current price is divided by the average of real (CPI-adjusted) earnings over the past ten years. The detailed CAPE calculation methodology covers the technical choices. What the ratio analytically measures is valuation against earnings normalized over a complete business cycle, capturing structural earning power rather than point-in-time earnings. the gap between CAPE and forward earnings multiples sets out the mechanism in detail. Adjacent reading: how real rates move equity multiples.

The three measures are not in competition for the same analytical use. They answer three distinct questions: how much am I paying for the immediate past (trailing), how much for the expected future (forward), how much for normalized earning power (CAPE). Conflating the three is the most frequent interpretive error in mainstream market commentary.

Historical Level Divergences

Over 1990-2025, the measures mostly kept the same order: CAPE exceeded trailing P/E in 79% of months per Shiller’s data, with the exceptions concentrated in 1991-1993 and 2001-2003, when a still-high ten-year average met depressed current earnings. This hierarchy does not come from inflation, which the CAPE neutralizes by expressing earnings in constant dollars. It comes from real earnings growth, which keeps the ten-year average below the current level, and, for forward P/E, from the growth the consensus expects, plus its optimistic bias.

On the S&P 500 in June 2026, the gap between CAPE and forward P/E is wide, 40.2 against 20.1 per Shiller and FactSet data. This gap is significant: the same asset class looks twice as expensive if the CAPE rather than forward P/E is used as reference. In late 2025 the gap was already very wide (CAPE at 39.6 in December). No reading of the equity market “level” can be indifferent to this divergence: depending on the chosen measure, one concludes that the S&P 500 sits in the extreme zone (CAPE at more than double its 16.6 median since 1881) or merely elevated (forward P/E barely above its ten-year average, per FactSet).

The divergences fluctuate over time. At the 2008-2009 crisis trough, trailing P/E exploded above 100 because TTM earnings had temporarily collapsed, while the CAPE bottomed at 13.3 in March 2009 per Shiller’s data, with trailing P/E peaking at 124 in May. This dispersion in crisis is diagnostically informative: an explosive trailing P/E at cycle trough signals the temporariness of the earnings collapse, not a structural overvaluation. Reading the trailing P/E alone in this configuration would have produced the opposite reading to the CAPE: one of the cases where measurement choice changes the economic interpretation. Related reading: The historical mapping of the equities-to-gold relationship.

Conversely, at the 1999 and 2021 peaks, the measures converge toward the high zone: in December 1999 the CAPE reached 44.2 and trailing P/E 29.7, in December 2021 38.3 and 23.6 respectively per Shiller’s data. Trailing P/E and forward P/E rise because TTM earnings inflate less rapidly than prices, and the CAPE is structurally in the extreme zone. This convergence of the three measures in the high zone is precisely what empirically distinguishes extreme-valuation phases from mere cyclical distortion phases, and the CAPE as the academic reference measure draws from this convergence its robustness as regime thermometer.

Documented Biases of Each Measure

Each measure carries a documented bias that limits what it can say on its own. Trailing P/E is cyclically unstable: its values at cycle troughs and peaks are not directly comparable to its values in expansion plateaus. This instability is not a conceptual defect (it reflects actual earnings mechanics) but it limits the use of trailing P/E as a long-term regime thermometer.

Forward P/E embeds the documented optimistic bias of sell-side consensus. McKinsey quantified it in 2010: from 1985 to 2009, analysts forecast earnings growth well above the realized pace, which was only 6% a year. This bias is not neutral: forward P/E published at time t tends to understate the P/E that will be observed retrospectively over the same period, all the more so when forecast growth is high.

The CAPE carries two documented biases. The first is its reduced reactivity to structural shifts: a change in underlying S&P 500 profitability, rotation toward high-margin services for example, takes ten years to fully incorporate into the decadal mean, which can keep the ratio in the “elevated” zone while the fundamental justification has structurally shifted. The second bias is accounting: according to Jeremy Siegel (Financial Analysts Journal, 2016), the 2001 FAS 142 rule on goodwill impairment has depressed reported GAAP earnings since, which mechanically inflates the CAPE; he proposes substituting national accounts (NIPA) profits. The methodological critique of the CAPE by Jeremy Siegel and others documents both biases in detail.

Acknowledging these biases does not disqualify any of the three measures: it is precisely what allows their correct use. A rigorous reading of equity valuation combines the three optics without privileging one, and reads divergences as analytical signals of their own.

The Signal of Divergences

The gaps between the three measures carry cyclical information of their own, independent of their absolute level. First signal: a CAPE − forward P/E gap widening beyond the historical mean signals that the consensus expects strong earnings growth. Forward P/E embeds twelve-month earnings anticipations; if anticipations are elevated, forward P/E falls, widening the gap with the CAPE which remains anchored on the decadal retrospective average. This configuration is observed in late-cycle phases, where consensus anticipates sustained earnings expansion whose intensity is not yet reflected in the decadal history.

Second signal: the gap between trailing and forward P/E measures the earnings growth the consensus expects over twelve months. On 26 June 2026, FactSet computed a forward P/E of 20.1, well below its trailing P/E on realized earnings, implying expected earnings well above those of the last twelve months. The wider this gap, the more valuation rests on growth still to come. A forward P/E above trailing P/E, a rarer configuration, would instead mean the consensus expects earnings to fall.

Third signal: simultaneous convergence of the three measures in the high zone (all above their respective historical medians) signals the robustness of an elevated valuation regime, as observed in 1999 and 2021. This convergence differs from a static reading of a single ratio in the elevated zone: it eliminates cyclical artifacts (explosive trailing P/E at troughs) and anticipational artifacts (optimistic forward P/E).

Fourth signal: the Excess CAPE Yield level adds a transversal information layer by integrating real rates. The ECY construction from the CAPE and real DGS10 enables the distinction between extreme-valuation phases with and without a risk-premium cushion, the analytical extension of the pure CAPE that is most useful over the long run.

Beyond these four direct signals, the practical use of the three measures together requires a discipline of horizon-matching. The trailing P/E is the natural reading for short-term valuation against realized fundamentals, and is most useful in interpreting one-year forward P/E expansion or compression. The forward P/E is the natural reading for evaluating the alignment between current valuation and consensus expectations, and is most useful at points of divergence between sell-side optimism and underlying earnings momentum. The CAPE is the natural reading for evaluating the regime against historical norms, and is most useful for setting expectations on the decadal return distribution. Mixing horizons (using the CAPE to time short-term moves, or the trailing P/E to project decadal returns) produces interpretive errors that have nothing to do with the quality of the underlying measures and everything to do with their misuse. Further reading: our guide to the all-time-high debate.

Last updated — 6 October 2026