For more than three decades, the Shiller CAPE ratio has been one of the most widely cited valuation tools in finance. It has also, since the early 1990s, been persistently, systematically wrong. The question is why. And the answer, it turns out, has significant implications for how advisors and investors should interpret elevated equity valuations right now.
In a June 2026 working paper, Federal Reserve Board economist Dino Palazzo argues that the CAPE ratio's apparent structural break reflects not a permanent shift in equilibrium valuations, not sustained overvaluation, but a measurement problem. Palazzo states that the post-1992 elevation of CAPE "can be largely attributed to systematic earnings distortions arising from accounting changes interacting with intangible capital growth."
The Dog That Did Not Bark
The intellectual foundation here draws on John Cochrane's famous argument: if dividend growth is unpredictable, then valuation ratios must predict returns. The CAPE ratio is built precisely to do that, smoothing earnings over ten years to strip out cyclical noise. The problem is that two regulatory changes corrupted the denominator.
First, FASB Statement No. 2 (1974) mandated the immediate expensing of research and development costs rather than capitalization. As technology and pharmaceutical companies grew to dominate the S&P 500, this rule created a widening gap between GAAP earnings and what Palazzo calls "adjusted earnings." Second, the SFAS 121 standard of 1995 required broader asset impairment recognition and restructuring charges, introducing episodic volatility into reported earnings that CAPE's ten-year average cannot fully smooth away.
The result, Palazzo finds, is that post-1992 CAPE has averaged 27.8 against CAPE-H's 19.3, a 44 percent overstatement. The traditional CAPE was not identifying a new era of permanently elevated valuations. It was identifying its own measurement failure.
Constructing a Better Ruler
Palazzo constructs CAPE-H, a historically comparable version of CAPE, by adding back R&D expenditures and stripping out special items before computing the ten-year average. The adjustment restores what the paper calls "intertemporal comparability" without attempting to model economic earnings or estimate intangible capital stocks. It is a measurement-consistency correction, nothing more.
The results are striking. Traditional CAPE generates an out-of-sample R squared of negative 41 percent for five-year excess returns, meaning it performs substantially worse than simply using the historical average. CAPE-H generates a positive 14.22 percent. The mechanism is specific: CAPE-H restores predictability in price appreciation, while dividend growth remains unpredictable under both measures. Palazzo notes that this 47 percentage point improvement for price appreciation "represents the central mechanism through which CAPE-H improves out-of-sample forecasting performance."
The decomposition is instructive. R&D capitalization accounts for roughly 70 percent of the total error reduction; special items exclusion contributes the remaining 30 percent. Both corrections are necessary.
Cycles, Not Regime Shifts
One of the paper's most important contributions is historical recontextualization. Where traditional CAPE identifies the post-1992 period as unprecedented, CAPE-H identifies four sustained technology-driven high-valuation cycles across 145 years: the industrial trusts era (1885-1907), the computing and electronics revolution (1958-1973), the Internet buildout (1995-2008), and the platform economy era (2013-present). Palazzo observes that "the current episode's duration of 12.9 years sits well within the range established by earlier cycles." The alarm of permanent overvaluation was, in his framing, a false alarm.
The Warning That Now Has Teeth
This is where the paper's conclusion carries real weight for advisors today. During 2011-2020, CAPE ranked in the 75th to 95th historical percentile while CAPE-H ranked in the 40th to 65th range. CAPE was crying wolf. CAPE-H was not.
Since 2021, the two measures have converged. As of December 2025, both sit near the 90th historical percentile. Palazzo states plainly that "current conditions reflect the latest manifestation of these recurring cycles and signal an elevated probability of a market correction in the next five years." CAPE-H predicts a 61.1 percent probability of a five-year correction. CAPE predicts 60.2 percent. The wolf, this time, may actually be at the door.
5 Key Takeaways for Advisors and Investors
- Traditional CAPE has overstated equity overvaluation by roughly 44 percent since 1992 due to accounting distortions, not market excess.
- The corrected measure, CAPE-H, restores meaningful predictive power and should inform long-horizon return expectations.
- High equity valuations in the 2011-2020 era were not as dangerous as CAPE signalled; CAPE-H correctly indicated moderate conditions during that period.
- Both CAPE and CAPE-H now agree: current valuations are genuinely elevated, near the 90th historical percentile.
- Five-year correction risk is now statistically meaningful; advisors should revisit return assumptions and downside scenario planning with clients.
Footnote:
1 Palazzo, Dino. "The CAPE That Cried Wolf." SSRN Working Paper, Federal Reserve Board, 8 June 2026, papers.ssrn.com/sol3/papers.cfm?abstract_id=6900766.
