The Myth Ledger · Myth 05 · Overconfident
“A high Shiller CAPE tells you to get out.”
Valuation informs a distribution of long-run outcomes; it does not provide a reliable exit and re-entry calendar.
What survives scrutiny
Paying more for a stream of earnings generally lowers the return one should plan around, all else equal. But “all else” moves: profitability, accounting, sector mix, inflation, discount rates, and the valuation investors later accept. CAPE is more useful for sober planning ranges than binary market timing.
Evidence stack
What the literature can—and cannot—say
Tags distinguish measured evidence, economic foundations, model-dependent results, and important limits.
Economic foundation
Valuations contain long-horizon information.13
Campbell and Shiller’s work supports a relationship between valuation ratios and subsequent long-run returns. This is an expectations statement: expensive claims usually offer less prospective compensation.
Empirical evidence
Out-of-sample forecasting is much harder than fitting history.14
Many return predictors weaken when tested using only information available at the forecast date. A model can explain a historical average and still fail to improve an investor’s live decisions.
Boundary condition
Long horizons do not create independent observations.15
Ten-year returns measured every month overlap for 119 of 120 months. That supplies far less independent evidence than the chart’s number of dots suggests and can overstate confidence.
Technical lens
A forecast needs an interval, not an exclamation point.
CAPE divides price by ten-year average real earnings. It smooths the business cycle but introduces dependence on a trailing decade and accounting history. Even if the conditional mean return is lower at high CAPE, the conditional range remains wide.
CAPE = price ÷ 10-year average real earnings
Concrete example
Being early is a portfolio outcome.
If a market compounds for years while a timer waits, the eventual decline must first erase the foregone gains before cash catches up. A correct valuation concern can therefore produce a losing strategy without a robust sell rule, re-entry rule, and benchmark.
Decision checklist
A decision rule, not a slogan
Optional checkmarks stay in this browser. JavaScript enables saving; the decision rules below are always readable.
- Use conservative, range-based return assumptions in planning when broad valuations are high.
- Rebalance to a risk target instead of moving all-in or all-out on a threshold.
- Diversify internationally, but do not assume a lower CAPE identifies a free bargain; composition and risk differ.
- Precommit the evidence and re-entry rule before making any valuation-driven tilt.
Optional lab: explore the assumptions
The default illustration is readable without JavaScript. Changing its inputs requires JavaScript; outputs are not forecasts.
Counterfactual lab · illustrative
The cost of waiting for the crash
Deterministic assumed path, not a forecast: compare $100 invested with $100 held in cash, then ask whether a market decline is needed to put them level.
- Invested $100 after 3 years at 9.0% annually
- $130
- Cash $100 after 3 years at 3.5% annually
- $111
- End-date market decline needed to match cash
- −14.4%
A valuation concern must overcome path dependence. It needs a sell date, a decline of about 14.4% at this date, and a re-entry executed before recovery. This assumed outcome does not establish a reliable timing rule.
Model boundary · Deterministic returns and a single end-date decline; ignores tax, volatility, sequence, reinvested distributions, and the possibility cash or stocks follow a different path.
Chapter sources: evidence and limits
Original source numbers are retained across chapters. Each finding travels with its limitation.
- 13Stock Prices, Earnings, and Expected Dividends ↗
Campbell, J. Y. & Shiller, R. J. (1988). The Journal of Finance 43(3), 661–676.
Finding: Valuation ratios contain information about long-horizon returns, consistent with prices varying relative to fundamentals.
Limit: Long-horizon predictability is not short-horizon timing precision, and overlapping observations make inference difficult.
- 14A Comprehensive Look at the Empirical Performance of Equity Premium Prediction ↗
Goyal, A. & Welch, I. (2008). The Review of Financial Studies 21(4), 1455–1508.
Finding: Many celebrated return predictors performed poorly out of sample relative to a simple historical-average forecast.
Limit: Forecast evaluation is period- and specification-dependent; weak predictability may still matter for long-range planning.
- 15Long-Horizon Predictability: A Cautionary Tale ↗
Boudoukh, J., Israel, R. & Richardson, M. (2019). Financial Analysts Journal 75(2), 17–30.
Finding: Overlapping long-horizon returns can make statistical relationships look more certain than the independent information supports.
Limit: The critique concerns inference and confidence, not a claim that valuations contain zero information.