The Shiller CAPE Manual

Research note: every current number on this page is computed at render from our maintained copy of Robert Shiller's monthly dataset (1881–present, refreshed daily from shillerdata.com). The forward-return studies run on a real total-return index we reconstruct from the dataset's nominal price, dividend and CPI columns. Source data retrieved 2026-08-04. How we use AI.

At 41.2, the Shiller PE points to historically weak long-horizon returns — not to an imminent crash. This page computes both halves of that sentence from the full 1881–present record.

What the Shiller PE actually measures

The Shiller PE — formally the cyclically adjusted price-to-earnings ratio, CAPE, or P/E10 — divides the S&P composite price by the average of the last ten years of inflation-adjusted earnings. Robert Shiller and John Campbell built it in the late 1980s to fix the standard PE's worst habit: one year of earnings swings with the profit cycle, so the market looks cheapest exactly when profits are peaking and most expensive in the pit of a recession, when earnings have collapsed. Ten years of real earnings smooths the cycle out and makes 1929, 1966, 2000 and today comparable on one scale.

The price of that comparability is sluggishness. A decade-long denominator moves slowly, which is why the CAPE is a valuation gauge for the coming decade and a famously terrible timing tool for the coming quarter. This page treats it accordingly: we compute what actually followed every level of the CAPE since 1881, and we are precise about what those histories can and cannot license you to conclude.

Where it stands now

As of August 2026, the CAPE reads 41.2 (prior month 40.6), its 5th consecutive monthly increase. The latest month is provisional: Shiller builds each reading on the month's average price, so the newest row firms up after month-end and the trailing three months can restate. That is the 98.9th percentile of 1,748 monthly readings, higher than the September 1929 peak (32.6) and the November 2021 peak (38.6). The only months in a century and a half that were more expensive — 18 of them — all sit inside the dot-com top, which crested at 44.2 in December 1999. Today's reading stands 3.0 points below the most famous valuation extreme in market history.

The full history, 1881–today

Two structural facts jump out of the long chart. First, the modern era runs richer than the old one: the median reading since 1990 is 26.4 against the 16.6 full-record median — and even within the 1990+ sample, today sits at the 95.7th percentile. The structural-regime objection is real, and it takes today's reading from historically extreme to merely modern-era extreme. Second, the readings above 30 — the tinted zone — used to be a once-a-generation event. Before 2017 the market spent 57 months there in total. It has spent 79 months there since.

Does a high Shiller PE predict a market crash?

No — and the record is worth stating carefully, because both the doomer reading and the dismissal are wrong. From every monthly reading above 30 since 1881, the median worst drawdown over the following three years was -24% in real total-return terms, against -16% from the average starting month, and the worst case was -77% (the aftermath of 1929). Crashes from high valuations were deeper. But they were not scheduled: the CAPE first crossed 30 in June 1997, and the market returned another +60.9% in real terms over the 30 months before the December 1999 top.

The current episode makes the timing point even harder to escape. Since its first cross in July 2017, the reconstructed real total return through July 2026 is +158.1% — +11.1% annualized — with a worst drawdown of -24% along the way. Nine years of “expensive” produced returns most decades would envy. That is the strongest evidence on this page that the CAPE is not a timing signal.

The honest summary: a CAPE above 30 has told you the decade would be poor and the drawdowns along the way would be deeper. It has never told you the month, the year, or even — in the current case — the near-decade.

What the CAPE says about the next 10 years

This is the study the CAPE was built for. For every month since 1881 we take the starting CAPE, then measure the S&P's real total return over the following ten years — dividends reinvested, inflation removed — using the index we reconstruct from Shiller's own price, dividend and CPI columns. (Sanity check: that index compounds at 6.7% per year across the whole record, matching the canonical long-run figure.) Grouped by starting valuation:

The association is broadly inverse, though not perfectly monotonic: cheap starting points delivered +11.0% a year at the median, mid bands drifted lower — with the 25–30 band actually edging the 20–25 band — and the 30 and above band, where today's 41.2 sits, delivered -1.1% with 60% of windows ending underwater. The Eras column is the disclosure that matters most: monthly windows overlap, so the 57 months above 30 collapse into just 2 completed historical eras — 1929 and 1997–2002 — rather than 57 independent samples. And every window that starts after 2016 is still open, so the current 79 qualifying months have contributed nothing to the table yet. If the next decade turns out fine, this table softens; that is exactly the falsifiable bet the CAPE makes.

Every episode above 30, computed

A qualifying month starts a new episode only after more than twelve months below 30 — a mechanical rule, so nobody can cherry-pick the eras. There have been 3 in 145 years:

The first two episodes resolved the same way at different speeds: 1929's two months above 30 preceded a -77% real collapse, and the dot-com episode (55 qualifying months, 30 from first cross to peak) gave back -45% over the five years after its 44.2 top. The third row is the one nobody can grade from the peak yet: it holds 79 qualifying months spread across a 110-month span (its sub-30 interruptions each lasted under a year), and its peak — 41.2 — is the latest data point on the page. Graded from its first cross instead, it is the +158.1% run described above. (*The current episode's first-cross-to-peak count is still growing.)

The interest-rate rebuttal, using Shiller's own fix

The strongest objection to CAPE doomism is that stocks compete with bonds, and a 41× earnings multiple means something different at 2% real yields than at 6%. Shiller agreed — which is why he publishes the excess CAPE yield: the CAPE earnings yield minus the real 10-year Treasury yield. It is the spread stocks offer over bonds after inflation.

The chart is why we resist flat 2000 comparisons even while making them. At the December 1999 peak the excess CAPE yield was -1.1% — the cyclically adjusted earnings yield sat below the real Treasury yield, meaning stocks carried no valuation-based premium over bonds at all. (An earnings yield is a valuation gauge, not a total-return forecast, so read it as pricing rather than prophecy.) Today the spread is 1.0% — the 19th percentile of the full record and the 15th of the modern era: a thin cushion, but a positive one. On the pure CAPE lens we are 3.0 points from 2000; on Shiller's own rate-adjusted lens we are not there. Both statements are true, and the tension between them is the current market in one sentence.

Where the CAPE will mislead you

• As a sell signal. Crossing 30 in June 1997 preceded three more years of huge gains; crossing 30 in July 2017 has preceded 6+ years of them. Nothing in this page's tables supports acting on a threshold cross.
• Against one eternal average. Accounting standards, payout policy, sector mix and real rates all drift across 145 years. A reading “above the long-run mean” has been true for most of the last three decades — comparing within eras and in percentiles is the defensible version.
• With overlapping-window confidence. The 57 months above 30 with complete 10-year outcomes collapse into two eras. Treat the band table as two case studies plus arithmetic, not 57 independent draws.
• Ignoring the denominator's lag. Ten-year average earnings still carry the 2020 pandemic collapse; as weak years roll out of the window, the CAPE can fall with prices flat. Part of every CAPE move belongs to the denominator's calendar rather than the market's mood.
• Alone. The CAPE is one of ten gauges in our Bubble Tracker, which replays the full framework at the 2000, 2007 and 2022 peaks — 2007 is the sobering one, because the CAPE looked merely warm while the excess lived in credit and housing.

The last 12 months

Full history: shiller-cape.csv (1,748 monthly rows, 1881–present). August 2026 is provisional — Shiller builds each reading on the month's average price, and the trailing three months restate in place as they finalize (this row last updated 2026-08-04). The live chart lives on the Shiller CAPE page.

How we checked it

Every number above is computed from the dataset when this page renders — none is typed in. The percentile is today's reading ranked against all 1,748 months. The forward-return study reconstructs a real total-return index from Shiller's nominal price, dividend and CPI columns (reinvesting dividends monthly, deflating by CPI); we validated it against the canonical long-run result — it compounds at 6.7% a year since 1881 — before trusting any band statistic. Dividends publish a month or two behind prices, so the total-return index currently ends in July 2026; months after that carry a CAPE but no graded return. Episodes use a stated mechanical rule (a year below 30 ends one), and windows are only graded where they are complete, which is why the current episode shows dashes from its peak. The underlying series is Shiller's revised dataset, not real-time vintages: the trailing three months can shift slightly after publication.

Frequently asked questions