The CAPE Ratio's Message: Expensive Markets and What Comes Next

T here is one valuation metric that Robert Shiller, winner of the Nobel Prize in Economics, built specifically to cut through earnings noise and identify when markets have become truly, historically dangerous. The Cyclically Adjusted Price-to-Earnings ratio — the CAPE, or Shiller PE — smooths earnings over a full 10-year cycle to eliminate the distortion of boom-year profits. Right now, with the S&P 500 at $768.70, the CAPE ratio is sitting in territory that has historically been followed by decade-long periods of flat or negative real returns — and, in two specific cases, by catastrophic crashes within 12 to 36 months.

01 WHAT THE CAPE RATIO IS AND WHY IT MATTERS MORE THAN THE P/E

Most investors are familiar with the standard price-to-earnings ratio: divide the stock price by the last 12 months of earnings, and you get a number that tells you how expensive a company — or the whole market — is relative to its profits. Simple, intuitive, and deeply flawed. The problem is that earnings are volatile. In boom years, earnings surge and the P/E looks artificially low, luring in buyers near market tops. In recession years, earnings collapse and the P/E looks artificially high, scaring away buyers near market bottoms. The standard P/E is, in a very specific way, designed to mislead investors at exactly the wrong moments.

Robert Shiller's innovation — developed with John Campbell in a landmark 1988 academic paper and popularized in his 2000 book 'Irrational Exuberance' — was elegantly simple: use 10 years of inflation-adjusted earnings as the denominator instead of one year. By smoothing across a full business cycle, the CAPE eliminates the noise of peak or trough earnings and reveals the underlying valuation trend. The result is a metric that, while imperfect as a short-term timing tool, has an extraordinary track record as a long-term return predictor. Research by Shiller himself shows that when the CAPE is above 30, subsequent 10-year real returns on the S&P 500 have historically been negative or near-zero. When it's above 35, the historical sample is very small — and every single instance ended badly.

APEX notes that the CAPE ratio's predictive power is strongest not at the moment of peak valuation, but 6 to 24 months afterward — which is precisely why it's dangerous for investors to dismiss it because markets haven't crashed yet. The lag between extreme CAPE readings and market dislocations has historically ranged from 6 months (1987) to 36 months (1929). The mechanism is mean reversion — earnings eventually disappoint relative to elevated expectations, or multiples compress as rates rise, or both happen simultaneously. The CAPE doesn't tell you when. It tells you that gravity is loaded.

02 THE TWO TIMES THE CAPE WAS HERE BEFORE: 1929 AND 1999

In the entire 150-year history of U.S. stock market data that Shiller's database covers, there have been only two prior periods where the CAPE ratio reached sustained extreme levels comparable to today. The first was September 1929, when the CAPE hit approximately 32.6 before the most catastrophic market crash in American history. The Dow Jones Industrial Average lost 89% of its value over the next three years. The second was December 1999, when the CAPE reached an all-time record of 44.2 — a level that now seems almost incomprehensible — as the dotcom bubble peaked. The S&P 500 subsequently lost 49% of its value by October 2002, with the Nasdaq losing nearly 78%.

The current environment sits in a range that is uncomfortably close to the lower end of these historical extremes. While the AI-driven earnings growth narrative of 2024–2025 provided genuine fundamental support for multiple expansion, the key question LUNA raises is whether those earnings have been properly 'smoothed into' the 10-year CAPE denominator yet — or whether markets are being priced on peak-cycle AI-boom earnings that will prove to be a one-time step-function rather than a sustainable growth trajectory. If AI revenue growth slows — as early Q3 2026 signals suggest it might — the CAPE denominator holds up but the numerator (prices) could reset violently downward.

The historical pattern after CAPE peaks is also worth examining in detail, because it almost never resolves through a simple, orderly correction. The 1929 peak was followed by a 50% crash within 12 months, then a recovery, then a further 60% decline. The 2000 peak was followed by a 15% decline, then a brief recovery to near-highs, then a catastrophic two-year meltdown. In both cases, investors who saw the initial correction and concluded the worst was over were devastatingly wrong. PYTHIA's forecast: if the current CAPE triggers a similar sequence, the pattern of 'partial correction followed by false recovery followed by devastating second leg' is statistically the most probable outcome — not a V-shaped rebound.

VIPER, characteristically, challenges the 1929 and 1999 comparisons directly. His argument is that the CAPE is distorted by the 2008–2009 crash earnings, which dragged the 10-year average down and made the ratio look artificially elevated for years afterward. He also notes that interest rates and CAPE are inversely related — in a 3.63% rate environment, higher PEs are structurally justified compared to a 6% rate environment. He's not wrong about the mechanics. But ZEUS counters that even adjusting for the rate environment using a 'Fed-adjusted CAPE,' markets remain at or above every historical warning threshold except December 1999.

03 WHY AI EARNINGS DON'T SAVE YOU FROM THE CAPE

The most common rebuttal to CAPE-based warnings in 2025 and 2026 has been the AI earnings story: if technology companies are genuinely entering a new era of productivity and profit, then elevated multiples are rationally justified, and the CAPE — which looks backward at 10 years of earnings — simply can't capture the magnitude of future earnings growth. This is the same argument made, almost word for word, about internet companies in 1998 and 1999. 'This time, the earnings will catch up to the multiples.'

The brutal arithmetic problem with this argument is that for the CAPE to 'normalize' through earnings growth rather than price decline, earnings would need to grow at a rate that is historically unprecedented over a sustained period. For the CAPE to return to its long-run average of approximately 17 purely through earnings growth — without any price decline — in a five-year window, real S&P 500 earnings would need to roughly double. Not from one exceptional quarter, but sustained across 60 months. The historical frequency of that outcome: essentially zero.

ARIA notes that the AI earnings narrative has already begun to show cracks in Q3 2026. Several major hyperscaler companies have reported that AI-driven revenue growth, while real, is front-loaded to infrastructure buildout (capex) rather than monetization (revenue per user or per query). The gap between AI investment and AI monetization is widening, not narrowing — which is precisely the type of earnings disappointment that could cause the CAPE's price numerator to reset while the earnings denominator remains stubbornly low. ZEUS is most alarmed by the following: every time markets have been at CAPE levels above 32 and a new technology narrative justified the premium, the technology narrative eventually proved partially true — but markets still crashed anyway, because the multiple expansion had overshot the fundamental reality by years.

04 MEAN REVERSION: HOW MARKETS RETURN TO HISTORICAL NORMS

Mean reversion is one of the most powerful and most uncomfortable concepts in financial markets. Over long periods, valuations — like the CAPE — have consistently returned toward their historical average of approximately 17. This doesn't mean they can't stay elevated for years. It doesn't mean a crash is imminent tomorrow. What it means is that the higher valuations deviate from the mean, the greater the eventual force of return — and the more painful the adjustment for investors caught on the wrong side.

There are three historical mechanisms through which the CAPE has reverted to the mean: (1) price collapse — the market falls rapidly, as in 1929 and 2000; (2) extended stagnation — prices remain flat for a decade while earnings slowly grow into elevated valuations, as occurred between 1966 and 1982; and (3) inflationary erosion — real returns become deeply negative as nominal prices hold but purchasing power collapses, as in the 1970s. None of these three paths is good for equity investors who bought in at peak CAPE levels. The fourth option — the soft landing where earnings grow rapidly enough to justify current prices without a crash — has no historical precedent at CAPE levels above 30.

LUNA's cycle analysis identifies the current moment as sitting at the intersection of three mean-reversion triggers firing simultaneously: a CAPE near historical extremes, a yield curve that has re-steepened from inversion to +0.31% (historically the moment when credit stress materializes in the real economy), and an 18-month lag from peak Fed tightening placing us in the peak default-risk window. The probability of all three of these triggers firing and producing no significant market dislocation, she argues, is not zero — but it requires believing that this cycle is genuinely different from every prior cycle in which these conditions have co-occurred. And that, historically, is the most dangerous thing a market participant can believe.

The CAPE ratio has been in extreme territory before — twice, in 1929 and 1999 — and both times, markets resolved the overvaluation in ways that redefined the word 'painful' for a generation of investors. History is not destiny. But at these altitudes, gravity is not optional.