CAPE has long been a cornerstone of long-horizon return forecasting. High valuations imply lower future returns. Low valuations imply higher future returns. Critics argue that its predictive power has faded in recent decades. This paper pushes back. It shows that the apparent decline is largely a measurement problem. When CAPE is constructed using aligned index constituents and market-cap weights, its out-of-sample predictive power exceeds 50 percent for ten-year returns. The result is a more precise and economically meaningful way to use valuation for long-term asset allocation.
CAPE ratios and Long Term Returns
- Ma, Marshall, Nguyen and Visaltanachoti
- Working paper, 2026
- A version of this paper can be found here
- Want to read our summaries of academic finance papers? Check out our Academic Research Insight category
Key Academic Insights
The traditional CAPE mixes different index constituents across time
The standard CAPE divides today’s index level by ten-year average earnings. However, firms enter and exit the S&P 500 over time. The numerator reflects current constituents. The denominator includes earnings from firms that may no longer be in the index. The authors construct a Component CAPE that aligns prices and earnings for the same firms. This simple correction materially improves predictive accuracy.
Out-of-sample predictive power exceeds 50 percent
Using a constant-slope out-of-sample framework, the Component 10-Year Earnings CAPE achieves an OOS R² of 0.5752, compared to 0.4667 for the Aggregate CAPE. The improvement is statistically significant and robust to bootstrap inference, Bonferroni correction, and false discovery rate adjustments.
The improvement is stronger in recent decades
The predictive gains are not confined to early sample periods. In fact, performance is stronger in the later subperiod when many commentators argued that CAPE had stopped working. The valuation-return relationship appears stable once measurement is corrected.
Weighting drives much of the difference
Mathematically, the traditional Aggregate CAPE is close to an earnings-weighted average of firm-level CAPEs. The Component CAPE instead uses market-cap weights. This weighting change explains a substantial portion of the difference in levels and predictive power. The market-cap weighted construction better reflects how capital is actually allocated in the index.
Economic value survives implementation tests
When used in a dynamic asset allocation framework, the Component CAPE delivers higher certainty equivalent returns than both the historical mean benchmark and a static 60/40 allocation.The gains are moderate but consistent across specifications.
Practical Applications for Investment Advisors
Align your valuation inputs
If you rely on CAPE for strategic allocation, use a version that aligns current constituents with their own historical earnings. Avoid mixing different firm sets across numerator and denominator.
Be mindful of weighting
Recognize that the traditional CAPE implicitly resembles an earnings-weighted measure. A market-cap weighted construction produces systematically different signals. Choose the weighting scheme deliberately.
Focus on long horizons
The predictive power documented here applies to ten-year returns. CAPE is not a tactical timing tool. It is a strategic allocation input.
Combine with complementary signals
Valuation works best as one pillar in a broader framework. Integrate it with trend, macro, or risk-based signals rather than relying on a single metric.
How to Explain This to Clients
“Valuations still matter. The apparent decline in CAPE’s usefulness largely reflects how it was measured. By aligning index constituents and using appropriate weights, we recover strong long-term predictive power. CAPE is not a short-term timing tool. It is a disciplined way to set long-horizon expectations and guide strategic asset allocation.”
The Most Important Chart from the Paper
The results are hypothetical results and are NOT an indicator of future results and do NOT represent returns that any investor actually attained. Indexes are unmanaged and do not reflect management or trading fees, and one cannot invest directly in an index.
Abstract
We demonstrate that 10-year equity market returns are considerably more predictable in relation to price–earnings ratios than previously thought. The traditional approach involves relating the current index price level, based on current index components, to the index earnings of previous years, calculated using those years’ components. When we estimate the cyclically adjusted price–earnings (CAPE) ratio, ensuring that index component prices and earnings are aligned, and apply a superior regression approach, out-of-sample R2 values are over 50%. The Component CAPE ratio weights individual stock CAPE ratios by their market capitalization, whereas the traditional CAPE ratio is more closely aligned with earnings weighting.
About the Author: Elisabetta Basilico, PhD, CFA
—
Important Disclosures
For informational and educational purposes only and should not be construed as specific investment, accounting, legal, or tax advice. Certain information is deemed to be reliable, but its accuracy and completeness cannot be guaranteed. Third party information may become outdated or otherwise superseded without notice. Neither the Securities and Exchange Commission (SEC) nor any other federal or state agency has approved, determined the accuracy, or confirmed the adequacy of this article.
The views and opinions expressed herein are those of the author and do not necessarily reflect the views of Alpha Architect, its affiliates or its employees. Our full disclosures are available here. Definitions of common statistics used in our analysis are available here (towards the bottom).
Join thousands of other readers and subscribe to our blog.
Facts Only
A working paper by Ma, Marshall, Nguyen, and Visaltanachoti (2026) examines the predictive power of the Cyclically Adjusted Price-to-Earnings (CAPE) ratio for long-term equity returns.
The traditional CAPE ratio divides the current index level by ten-year average earnings, but the index constituents change over time, creating a mismatch between current prices and historical earnings.
The authors propose a "Component CAPE" that aligns prices and earnings for the same firms, improving predictive accuracy.
The Component CAPE achieves an out-of-sample R² of 0.5752 for ten-year returns, compared to 0.4667 for the traditional Aggregate CAPE.
The improvement is statistically significant and robust to various adjustments, including bootstrap inference and Bonferroni correction.
The predictive gains are stronger in recent decades, countering claims that CAPE has stopped working.
The traditional CAPE is closer to an earnings-weighted average, while the Component CAPE uses market-cap weights, better reflecting capital allocation in the index.
In dynamic asset allocation frameworks, the Component CAPE delivers higher certainty equivalent returns than historical mean benchmarks or static 60/40 allocations.
The paper recommends aligning valuation inputs, being mindful of weighting schemes, and focusing on long horizons for strategic allocation.
The research suggests that CAPE remains a valuable tool for long-term asset allocation when constructed with aligned constituents and appropriate weighting.
Executive Summary
The traditional Cyclically Adjusted Price-to-Earnings (CAPE) ratio has been a key tool for forecasting long-term equity returns, with high valuations suggesting lower future returns and vice versa. However, critics argue its predictive power has weakened in recent decades. A new working paper by Ma, Marshall, Nguyen, and Visaltanachoti challenges this view, demonstrating that the decline in CAPE's effectiveness is largely due to measurement issues. The authors introduce a "Component CAPE" that aligns index constituents and uses market-cap weights, significantly improving out-of-sample predictive power for ten-year returns. The Component CAPE achieves an out-of-sample R² of 0.5752, compared to 0.4667 for the traditional Aggregate CAPE, with stronger performance in recent decades. The improvement stems from correcting the mismatch between current index constituents and historical earnings, as well as shifting from earnings-weighted to market-cap-weighted averages. When applied to dynamic asset allocation, the Component CAPE delivers higher certainty equivalent returns than historical mean benchmarks or static 60/40 allocations. The findings suggest that CAPE remains a valuable tool for long-term strategic allocation, provided it is constructed with aligned constituents and appropriate weighting.
For investment advisors, the paper recommends using a CAPE version that aligns current constituents with their historical earnings, being mindful of weighting schemes, and focusing on long horizons rather than short-term timing. The authors emphasize that valuation should be one pillar in a broader framework, integrated with other signals like trend, macro, or risk-based indicators. The research underscores that while CAPE is not a tactical tool, it can provide disciplined long-horizon expectations for strategic asset allocation.
Full Take
The strongest version of this narrative is that CAPE's apparent decline in predictive power is not a failure of the metric itself but a failure of measurement. By correcting the mismatch between current index constituents and historical earnings—and by using market-cap weights instead of earnings weights—the authors demonstrate that CAPE's predictive power for ten-year returns remains robust, even in recent decades. This is a compelling argument, as it addresses a legitimate criticism of CAPE while preserving its utility for long-term investors. The paper's emphasis on alignment and weighting is a technical but important refinement, and the out-of-sample results are statistically significant.
However, the narrative assumes that the Component CAPE's improvements are solely due to measurement corrections, without fully exploring alternative explanations. For example, could the stronger performance in recent decades be influenced by other structural changes in markets, such as the rise of passive investing or shifts in corporate profitability? The paper also does not address whether the Component CAPE's predictive power holds across different market regimes or asset classes. Additionally, while the authors advocate for integrating CAPE with other signals, they do not specify how to weight or combine these signals in practice.
The root cause of this narrative is the tension between the desire for simple, reliable valuation metrics and the complexity of real-world market dynamics. The traditional CAPE's decline in predictive power has been a point of frustration for investors, and this paper offers a technical fix that restores confidence in the metric. However, the broader implication is that valuation metrics are not static; they require ongoing refinement to remain relevant. This raises questions about how other widely used financial metrics might be similarly improved—or whether they, too, are subject to measurement biases that distort their predictive power.
For human agency and dignity, this research empowers investors to make more informed long-term decisions, but it also highlights the need for vigilance in how financial metrics are constructed and interpreted. The benefits accrue to those who can implement the Component CAPE in their allocation strategies, while the costs may fall on those who rely on outdated or misaligned versions of CAPE.
Bridge questions: What other valuation metrics might suffer from similar measurement issues? How would the Component CAPE perform in non-U.S. markets or different asset classes? What are the practical challenges of implementing a market-cap-weighted CAPE in real-world portfolios?
Counterstrike scan: If this narrative were part of a coordinated influence campaign, the playbook might involve exaggerating the predictive power of the Component CAPE to sell proprietary investment products or services. However, the actual content does not match this pattern. The paper is transparent about its methodology, acknowledges limitations, and does not overpromise results. It appears to be a genuine academic contribution rather than a marketing tool.
Patterns detected: none
Sentinel — Human
This analysis suggests that the article is likely human-written. The text exhibits varied sentence lengths, a clear and coherent argument with a personal voice, and does not follow a specific template or present talking points verbatim.