Excess CAPE Yield: Shiller’s equity risk premium mechanics and the 10-year forward returns forecast

Formalized by Shiller, Black and Jivraj in their 2020 paper “CAPE and the COVID-19 Pandemic Effect,” the Excess CAPE Yield (ECY) corrects a weakness of the pure CAPE by integrating the 10-year Treasury real yield — an operational metric of the current equity risk premium.

By mid-2026, the U.S. ECY stood at 1.4% in May 2026 on Robert Shiller’s data, its lowest level since 2007 — an analytical observation on the expected risk premium, not a timing signal.

The Mechanical ECY Construction

The Excess CAPE Yield is built in two simple steps that must be treated separately to avoid conceptual confusion. Step 1: invert the CAPE to obtain the CAPE-based earnings yield, representing the implicit real equity return under the Shiller measure. At a CAPE of 40.1 (May 2026), this earnings yield equals 1/40.1 = 2.49%. This inversion produces an intuitive measure of expected real return if decadal-normalized earnings remain constant in real terms — a strong assumption but useful as analytical reference. The same logic extends in the link between the stock index and gold as a real-wealth benchmark.

Step 2: subtract the 10-year Treasury real yield to obtain the spread between the two real yields. Shiller builds that real yield by subtracting average CPI inflation over the previous ten years from the nominal 10-year yield. In May 2026, with a nominal 10-year yield of 4.48% and average inflation of 3.39% a year over ten years, the real yield comes out at 1.09% and the ECY at 2.49% − 1.09% ≈ 1.40%. Measured with the market real yield of 10-year TIPS (2.04% on average in May 2026), the gap falls to 0.45%: the choice of real leg moves the result by a full point. The reference grid of the CAPE as academic measure provides the context for both variables.

The ECY conceptually measures the expected risk premium of equities over real Treasuries, read through the Shiller CAPE filter. Our excess-CAPE data sets out how the indicator is built. A positive ECY means equities offer an expected real return above Treasuries — the historically normal configuration. An ECY close to zero or negative means the expected risk premium is weak or null — a historically rare configuration associated with bubble peaks.

The metric was formalized by Shiller, Black and Jivraj in a paper published in November 2020, when Treasury yields were at historically low post-COVID levels and the CAPE was rising rapidly after the March 2020 drop. The authors’ goal was to address a recurring critique of the pure CAPE: what can a high CAPE mean if Treasury yields are also very low, that is, if the bond alternative offers no significant opportunity cost? The ECY mechanically answers this by integrating opportunity cost into the measure.

ECY History 1881-2025

Across 145 years of data, the U.S. ECY has oscillated within a broad range but with a clearly centered distribution. The 1881-2025 median is 3.3% on Shiller’s data (3.4% over 1881-2023), which can be interpreted as the “normal” expected long-term equity risk premium in U.S. markets. This median value is consistent with academic estimates of realized equity risk premium over long periods, which run around 3-5% across methodologies. Related question: how the equity risk premium is measured.

Extreme low zones (ECY below 1%) have been visited only a few times in the series. The late 1920s before the 1929 crash saw ECY drop below 1% in November 1928 and below zero in 1929 (−0.6% at the low), a configuration combining a high CAPE (32.56 at peak) and moderate but positive real Treasury yields. The late 1990s before the dot-com burst saw ECY turn negative in 1999 and reach −1.5% in 2000, a configuration (also seen in the 1880s-1890s and in 1929) where the implicit real return on equities became inferior to the real return on Treasuries — an anomaly that directly preceded the 2000-2009 lost decade.

The 2021 peak, CAPE 38-40, coexisted with a strongly negative real DGS10 (around −0.8% in early 2021 per FRED), which paradoxically kept the ECY around 3-3.5%, close to the historical median. This configuration distinguishes the 2021 peak from the 1929 and 1999 peaks: the extreme equity valuation was partially justified by the post-COVID real rate compression. At that point, reading the CAPE alone produced an extreme signal; reading the ECY produced a near-normal signal — a divergence that is precisely the diagnostic contribution of the metric. On the same theme: valuation seen through real rates and profits.

The May 2026 configuration (CAPE of 40, positive real rates, ECY at 1.4%) brings the ECY to its lowest level since 2007, without falling below 1% as it did from 1994 to 2002. This configuration combines extreme equity valuation and positive real rates, unlike 2021 where real rates were negative. Statistically, it resembles 1929 and 1999 more than 2021 — an observation that structures part of the current debate on the 2024-2026 valuation regime developed in the satellite on the current CAPE level.

Documented Predictive Power

The fundamental analytical property of the ECY is its documented predictive power on 10-year forward equity returns. Computed on Shiller’s data, the correlation between the month’s ECY and the annualised real return of US equities over the following ten years is +0.58 over 1881-2016 (last complete window): a high ECY has preceded high returns. Over the same window, the pure CAPE’s correlation is −0.48. A related question: how Shiller CAPE differs from forward P/E.

This predictive power is higher in absolute value than the pure CAPE’s, and the ECY offers a qualitative advantage: robustness to rate regimes. The pure CAPE predicts less well when Treasury yields diverge sharply from their historical average — for example in 2021 when real rates were negative, the CAPE alone would have projected very low forward 10-year returns while the risk premium cushion remained normal. The ECY, by integrating this variable, produces a more stable projection across rate regimes. In the same vein: the CAPE calculation, from real earnings to the Shiller ratio.

Mechanically, the Shiller-Black-Jivraj 2020 regression takes a similar form to the pure CAPE regression but with ECY as the explanatory variable and a positive sign: Real Return 10Y ≈ constant + coefficient × ECY. On Shiller’s data, the fitted line reads 3.6% + 0.66 × ECY: at 1.40%, as in May 2026, it gives about 4.5% real annual return over ten years, an average around which historical dispersion is wide.

The predictive power is not a timing power. The ECY can remain in the low zone for several years (it did between 1996 and 2000) without signaling immediate reversal. It is precisely this persistence that distinguishes the ECY from a tactical indicator: it informs about mean expected returns over ten years, not about one-year turns. The same methodological discipline as for the pure CAPE applies to its ECY cousin.

As of 6 October 2026, the ECY has kept falling, to 1.01% in September 2026 on Robert Shiller’s data (CAPE of 40.6), its lowest level since May 2002. Measured with the 10-year TIPS real yield (2.64% on average in September 2026, Federal Reserve), the gap turns negative, at −0.17%.

Limits and Distinctions with the Fed Model

The ECY carries several limits that must be named explicitly to avoid over-interpretation. First limit: its sensitivity to the choice of real DGS10 measure. Shiller subtracts past ten-year inflation; the market alternative uses the 10-year TIPS (Treasury Inflation-Protected Securities) yield, which is exactly the nominal DGS10 minus the T10YIE breakeven, since FRED computes that breakeven as the difference between the two. The gap between the two approaches reached a full point in May 2026 (ECY of 1.40% against 0.45%), which changes how an ECY near 1% reads. Changing the maturity of the rate leg moves far more than the measure of it: the choice between the 2-year and the 10-year in the equity risk premium is not neutral on the sign obtained.

Second limit: the ECY inherits the underlying CAPE biases, limited reactivity to structural shifts, sensitivity to the 2001 FASB 142 break. If the CAPE is mechanically inflated by 2-4 points per the Siegel critique, the ECY is mechanically depressed by the inverse — roughly 0.2-0.3% lower. The methodological critique of the CAPE and its derivatives covers these inherited biases.

Third limit: the ECY is conceptually close but mechanically distinct from the Fed model, a valuation tool popular in the 1990s-2000s that compares trailing earnings yield (1/trailing P/E) to nominal 10-year Treasury yield. The Fed model uses current earnings and nominal rates; the ECY uses normalized CAPE earnings and real rates. This difference is not cosmetic: the Fed model is sensitive to cyclical earnings distortions and nominal inflation, while the ECY is smoothed by real decadal earnings. Over long periods, the two measures can give opposite signals — the ECY is analytically more rigorous but the Fed model has a longer practitioner usage history. The debate on the empirical relationship between real rates and CAPE documents the complete empirical audit of this articulation.

Fourth limit: as with any expected risk premium measure, the ECY is not directly observable, it is an analytical construction based on explicit methodological assumptions. No “real” measurable ECY exists in the market: what is measurable is the ex-post ECY (computed after the fact on realized returns), not the ex-ante ECY (which anticipates those returns). This distinction is conceptually important and limits ECY use to an analytical framing instrument rather than a deterministic predictive indicator.

Last updated — 6 October 2026