Can Valuation Predict Returns? Using CAPE and Earnings Yield as a Long-Horizon Compass

Photo by Bradley W. on Unsplash

Every few years the question resurfaces: if the market looks expensive on long-run valuation measures, can that tell us anything useful about what comes next? The honest answer is a qualified yes — but the qualifications matter more than the yes. This piece looks at what the cyclically-adjusted price-to-earnings ratio (Shiller CAPE) and its cousin, the earnings yield, can and cannot forecast, and then examines the one U.S.-listed ETF that carries the CAPE name to see whether it actually implements the idea people think it does.

Context: what CAPE and earnings yield actually measure

The Shiller CAPE ratio divides today's real price by the average of ten years of inflation-adjusted earnings. The ten-year smoothing is the whole point: it strips out the boom-and-bust noise in a single year's profits, so a high reading is harder to dismiss as a temporary earnings dip. Earnings yield is simply the inverse — trailing or cyclically-adjusted earnings divided by price — expressed as a percentage you can hold next to a bond yield.

The empirical claim, developed across decades of work by Robert Shiller and later formalized in the academic literature on long-horizon return predictability, is narrow and specific: starting valuation is inversely related to subsequent long-run returns. Buy the U.S. market at a CAPE of 10 and history has been kind over the following decade; buy at 35 and the base rate of forward real returns compresses. The relationship is statistical, noisy, and — critically — operates on a horizon most investors are not actually watching.

The CAPE ETF is not the CAPE trade

Here is the point most coverage gets wrong. The DoubleLine Shiller CAPE U.S. Equities ETF does not go to cash when the aggregate market looks expensive. It is always fully invested in U.S. equities. What it does is rank the eleven-or-so GICS sectors by each sector's own CAPE relative to that sector's history, buy the cheapest cohort, apply a momentum screen to avoid value traps, and rotate. It is a relative-value sector strategy wearing a famous name.

That distinction is the non-obvious insight worth sitting with. The aggregate-CAPE "market timing" application — sell when the whole market is dear — has a weak live track record and a brutal opportunity cost, because expensive markets can stay expensive for a decade. The relative-sector application sidesteps that problem entirely: there is always a cheapest sector, so the strategy never has to make the all-or-nothing call about whether U.S. equities as a class are overvalued. Whether the sector version works net of a 0.65% fee is a separate question, but it is at least a coherent use of the metric rather than the discredited one.

Valuation is a compass for expectations across a cycle, not a switch you flip. The metric that tells you where the market probably sits in a decade tells you almost nothing about where it sits next quarter.

The fund by the numbers

DoubleLine launched this vehicle on 31 March 2022, so the live record spans roughly four years — and, importantly, an unusually specific four years: a rate-hiking shock, a concentrated large-cap growth rally, and an AI-driven regime that rewarded exactly the sectors a value-tilted rotation model tends to underweight. That is single-regime risk in its purest form. Any performance figure over this window is a sample of one macro environment, not evidence about the strategy's behavior across cycles.

Source: yfinance price and distribution data pulled 2026-07-16; expense ratio, AUM, inception, and strategy mechanics from the DoubleLine issuer fact sheet. The null CAGR and drawdown fields are not omissions — the fund simply has not existed long enough to populate them, which is itself the most important fact on the page.

Reading valuation against today's rates

The earnings-yield framing is most useful when you place it next to the risk-free alternative. As of mid-2026 the 10-year Treasury sits at 4.58% and the fed funds rate at 3.63% (FRED, asof 2026-07-14 and 2026-06-01). With CPI running near 3.7% year-over-year (FRED, asof 2026-06-01), the real 10-year yield is positive but modest. When cash and bonds pay something close to their long-run average, the equity earnings yield has to work harder to justify the risk premium — which is precisely the condition under which a high market CAPE (a low earnings yield) becomes more uncomfortable, because the cushion over the risk-free rate thins.

This is where valuation earns its keep for a long-horizon investor: not as a sell signal, but as an input to the expected return you pencil into a plan. If starting valuation is rich and real yields are positive, prudent planning trims the forward equity return assumption rather than the equity allocation. That is a subtle but real difference, and it connects directly to how rebalancing discipline adds to buy-and-hold returns — bands force you to sell what has become expensive and buy what has become cheap without ever needing a macro forecast.

What the predictive record honestly looks like

The regression that made CAPE famous — starting valuation on the x-axis, subsequent 10-year real return on the y-axis — has an R² that is respectable at the decade horizon and collapses toward zero at the one-year horizon. Two caveats are non-negotiable when you read that chart. First, look-ahead and data-mining risk: the ten-year smoothing window and the specific functional form were chosen with the benefit of hindsight on U.S. data, and the U.S. is the single best-performing large equity market of the twentieth century — survivorship at the country level. Second, the overlapping-windows problem: decade-long return periods sampled monthly are heavily autocorrelated, which inflates the apparent statistical significance far beyond what the number of truly independent observations supports.

Initially I treated the CAPE-versus-forward-return relationship as close to a law. Then I looked at how few genuinely independent decade windows exist in the usable data — a century of history contains only a handful of non-overlapping ten-year periods — and the confidence intervals widened considerably. The signal is real. It is also far weaker than the tidy scatterplots suggest, and it says nothing actionable about the next twelve months. For investors thinking about how sectors move relative to one another over a cycle, the related idea of market rotation is a useful companion frame — and the final-decade asymmetry of compounding is a reminder that where you start on valuation matters far less than whether you stay invested through the middle.

Implementation friction worth naming

Three practical points. The 0.65% expense ratio is roughly twenty times a plain S&P 500 index fund, and that gap compounds relentlessly against the strategy's realized edge — the sector-rotation alpha has to clear the fee before it adds anything. At $0.24B in AUM the fund is small; small-AUM ETFs carry wider bid-ask spreads and, in the tail, closure risk if assets don't grow. And a rules-based sector rotation generates more turnover than a market-cap index, which can raise the tax-cost ratio in a taxable account relative to a buy-and-hold core. None of these is disqualifying. All of them belong in the comparison.

The editor does not hold CAPE at the time of writing.

FAQ

Does a high CAPE ratio mean I should sell stocks? No. High starting valuation lowers the base rate of forward long-run returns, but expensive markets have historically stayed expensive for many years. Using CAPE as an exit signal has produced long stretches of costly underinvestment. It is better used to temper return expectations than to trigger trades.

Is the DoubleLine CAPE ETF a market-timing fund? No. It stays fully invested in U.S. equities and rotates among sectors based on each sector's valuation relative to its own history, with a momentum overlay. It never makes the all-or-nothing call on the aggregate market.

Why are there no 5-year or 10-year return figures? The fund launched on 31 March 2022, so roughly four years of live history exist — not enough to populate a five-year CAGR or a five-year drawdown statistic. Any figure over this window reflects a single macro regime.

How does earnings yield compare to the 10-year Treasury right now? With the 10-year Treasury at 4.58% (FRED, asof 2026-07-14), the equity earnings yield has a higher bar to clear to justify the equity risk premium than it did during the zero-rate era. A low market earnings yield is more uncomfortable when real risk-free yields are positive.

Can valuation predict next year's return? Essentially no. The predictive relationship is a long-horizon phenomenon — its explanatory power is meaningful at ten to fifteen years and negligible at one year.

Key takeaways

• CAPE and earnings yield forecast long-horizon returns with real but modest and noisy skill; they say almost nothing about the next twelve months.
• The DoubleLine CAPE ETF rotates sectors on relative valuation — it is not the discredited "sell the expensive market" trade.
• A four-year live record across one unusual regime is not evidence about behavior across cycles; treat the null long-run statistics as the headline fact.
• Use valuation to set your expected-return assumption, and let rebalancing bands act on price without requiring a forecast.
• Cost, small AUM, and turnover are real frictions that the strategy's edge must clear before it adds value.

Methodology: price, NAV, and distribution data from yfinance (pulled 2026-07-16); expense ratio, AUM, inception, and strategy mechanics from the DoubleLine issuer fact sheet; macro rates from FRED (10-year Treasury and VIX asof 2026-07-14, fed funds and CPI asof 2026-06-01). Return-predictability discussion draws on the published academic literature on cyclically-adjusted valuation and long-horizon returns. Window analyzed: fund inception (2022-03-31) through 2026-07-16.

This article is for educational purposes and does not constitute personalized financial advice. See our full Disclaimer.