The Analyst’s Course · Macro Regimes
What CAPE Actually Predicts
The relationship
Regress subsequent 10-year real returns on starting CAPE and you get the cleanest long-horizon relationship in finance: correlation around −0.7, r² near 0.4–0.5 depending on sample. The implied-return heuristic the panel shows — roughly (1/CAPE)×100 + ~1.4pp drift — is the regression's one-line summary: CAPE 41 → ~3.8%/yr; CAPE 15 → ~8%/yr. The mechanism is mean reversion in valuations: starting from high multiples, either prices stagnate while earnings catch up, or they fall — either way the decade's return is clipped.
What it does NOT predict
Next year. Almost nothing about the 1–2 year horizon comes out of CAPE: 1996's CAPE 28 ("irrational exuberance") was followed by four more years of +100% gains before the collapse; 2017's elevated readings were followed by two excellent years. The relationship is a tide gauge, not a wave forecast. This is why the honest use is risk-budgeting (position sizes, savings rates, expected-plan returns) and the dishonest use is market timing on a 12-month clock.
The regression's own caveats
Three real criticisms: (1) sample size — ten-year returns overlap heavily, so independent observations are few (the r² is impressive partly by construction); (2) composition — the index changes (1999's tech weighting vs today's mega-caps), so the "same" CAPE may mean something different; (3) regime shifts in accounting standards and payout policy (buybacks vs dividends change the earnings base's meaning). Each criticism trims the edge; none has erased the decade-scale relationship in 150 years of data.
≈ (1 ÷ CAPE)×100 + 1.4pp
Implied 10y return — The one-line regression. CAPE 41 → ~3.8%/yr expected; CAPE 15 → ~8%. Decade scale only.
Greenspan, December 1996
The Fed chairman's "irrational exuberance" speech came at CAPE ~28 — the 96th percentile. Four more years: +100% on the S&P, CAPE reaching 44. A value investor acting on the signal in 1996 underperformed for years before being vindicated. The lesson is not that the signal was wrong — 2000–2009 returned roughly −3%/yr real from the 2000 starting point, exactly as the regression implied — it is that decade-scale signals arrive years early, and portfolios must be built to survive their earliness.
What you'll practise
The CAPE-vs-future-returns relationship is strongest at…
3 graded checkpoints · certification exam at the end of the track
Sources
Shiller, Irrational Exuberance (3rd ed.); Estrella & Mishkin (1996); multpl.com; US Treasury
Learn content is for education only — not individualized financial advice, a recommendation, or a solicitation to buy or sell any security. Options involve substantial risk. Examples are simplified and historical patterns never guarantee future results.