Does a CAPE Above 40 Mean a Crash Is Coming? What Shiller's Own Data Says About the Next Ten Years

October 29, 2026

Current Shiller CAPE: 40.71 (data month: October 2026)

Percentile: 98.9th of 1,750 monthly readings since January 1881. Nineteen months were higher: eighteen between March 1999 and September 2000, plus August 2026 (40.99).
Long-run mean 17.80 · median 16.61 · all-time high 44.20 (December 1999).
Source: Robert Shiller, ie_data.xls on shillerdata.com, latest row as checked November 6, 2026; the percentile is computed from that file.
Last updated: November 6, 2026 · next check: early December 2026.

Short answer: a Shiller CAPE above 40 has never, in 145 years of data, said anything useful about when a decline would start — but it has said a great deal about the ten years that followed. In Robert Shiller's own spreadsheet the ratio stood at 40.71 for October 2026, the 98.9th percentile of 1,750 monthly readings going back to January 1881, and all 21 earlier months that ever reached 40 were followed by a negative ten-year real return. The first of those 21 months was also followed by a +12.5% year.

This page is kept as a standing reference and is rebuilt from Shiller's raw file rather than from any site that republishes it. Everything below — the band table, the one-year table, the percentile — was computed on November 5, 2026 (current value re-checked November 6, 2026) directly from the ie_data.xls workbook on shillerdata.com, which is the file Shiller himself maintains. For a peak-by-peak narrative of 1929, 2000 and 2022 instead of the distribution, see What Followed CAPE's Three Highest Prior Peaks.

Where the ratio stands (standing box, refreshed monthly)

Shiller CAPE, October 2026: 40.71. Total-return CAPE: 43.33. Excess CAPE yield: 0.56%.
High of this episode so far: 40.99 (August 2026). All-time high: 44.20 (December 1999).
Rank since January 1881: 98.9th percentile of 1,750 monthly readings.
Long-run mean 17.80, median 16.61, low 4.78 (December 1920). The October 2026 reading is 2.29 times the mean.
Source file read on November 5, 2026. Shiller publishes monthly, so this box moves once a month, not daily.

Key takeaways

October 2026 reads 40.71, which is 2.29 times the 1881–2026 mean of 17.80 and the 98.9th percentile of the whole series.
Only 27 of 1,750 months (1.5%) have ever printed 40 or higher: January 1999 through September 2000, and May through October 2026.
Across 1,630 completed ten-year windows, starting CAPE above 30 produced a median real total return of minus 1.1% a year; starting CAPE above 40 produced minus 3.4% a year, negative in 21 of 21 windows.
Over one year instead of ten, the same data gives a median of plus 4.7% after readings above 35 — the down-year rate rises from 32% to 45%, which is a tilt, not a warning siren.

Figures checked: November 5, 2026.

What Shiller's own spreadsheet says as of October 2026

The file has 1,750 monthly CAPE observations, January 1881 through October 2026. Sorting them puts the current reading of 40.71 above 1,731 of them. The mean of the whole series is 17.80 and the median 16.61, so the market is priced at 2.29 times its average multiple and 2.45 times its median one. The lowest reading in the file is 4.78, in December 1920.

Two details matter for anyone comparing this page with a number seen elsewhere. First, Shiller's series is monthly: the price input is an average of daily closes, and the current month carries whatever price was available when the file was last rebuilt. Sites that publish a CAPE that moves every afternoon are computing their own daily variant, not reproducing this series, which is why a daily figure near 41 and this month's 40.71 are both defensible and neither is the other. Second, the same row of the file carries a total-return CAPE of 43.33 and an excess CAPE yield of 0.56% against a 10-year Treasury of 5.24% — that last figure is the one that has actually deteriorated most this year, because the bond yield rose while the multiple stayed flat.

Every starting CAPE band since 1881, and the real return that followed

This is the table the metric deserves and almost never gets. For every month with both a CAPE reading and a full ten years of subsequent data — 1,630 windows, January 1881 through October 2016 — the starting CAPE is sorted into a band and the realised ten-year real total return (dividends reinvested, inflation removed, annualised) is measured from Shiller's real total return price column.

Starting CAPE	Windows	Median 10y real	Mean	Worst to best	Negative
Under 10	229	+11.0%	+11.3%	+1.8% to +20.0%	0 of 229
10 to 15	461	+7.8%	+7.6%	-4.2% to +17.6%	45
15 to 20	514	+6.2%	+6.0%	-4.6% to +16.1%	46
20 to 25	260	+4.9%	+4.6%	-4.0% to +14.2%	63
25 to 30	109	+5.6%	+5.9%	-1.2% to +11.8%	5
30 and above	57	-1.1%	-1.0%	-5.9% to +4.5%	34 of 57
of which 35 to 40	13	-0.6%	-0.8%	-3.8% to +1.1%	8
of which 40 and above	21	-3.4%	-3.7%	-5.9% to -2.6%	21 of 21

The gradient from the top of the table to the bottom is the whole case for the metric, and it is a strong one: cheap starting points delivered double-digit real returns with no negative outcome in 229 tries, and expensive starting points delivered roughly nothing. But two caveats belong next to the numbers rather than in a footnote. The 25-to-30 band looks oddly better than the 20-to-25 band, because those 109 months are mostly 1927–1929 and 1996–1998, and the second group ran into a decade that ended well. And the overlap problem is severe: consecutive monthly windows share 119 of their 120 months, so the 21 observations above 40 are not 21 independent experiments. They are one episode, January 1999 to September 2000, counted 21 times.

The only other time the ratio was this high

That is worth being blunt about, because it cuts both ways. Readings of 40 or more have occurred in exactly 27 of 1,750 months, or 1.5% of the record, and they come in two clusters: 21 months from January 1999 to September 2000, and the six months from May to October 2026 that we are living in. There is no third case. Anyone who tells you what happens after a CAPE of 40 is describing one stretch of one bubble in one country.

Month	CAPE	Next 10 years, real, annualised	Note
January 1999	40.58	-4.4%	First month ever above 40. The next 12 months returned +12.5% real.
March 1999	41.44	-5.9%	Worst ten-year outcome in the whole file.
December 1999	44.20	-3.2%	All-time high reading.
January 2000	43.77	-3.0%	The reading most often held up as the current reading's twin.
September 2000	41.90	-3.1%	Last month above 40 until 2026. The next 12 months returned -29.8% real.
October 2026	40.71	window closes October 2036	Sixth month of the current cluster.

Put the first and last rows side by side and the timing problem is impossible to miss. Both months printed about 41. One was followed by the best year of the bubble; the other by a 30% real loss. The ten-year column, meanwhile, is remarkably tight — every value between minus 2.6% and minus 5.9%. The decade was knowable in a way the year was not.

The same question over one year instead of ten

Running the identical calculation over a 12-month horizon, using all 1,738 months that have a following year of data, is the cleanest way to see what the metric does and does not carry.

Starting CAPE	Months	Median next 12 months, real	Mean	Worst to best	Down years
All months	1,738	+8.7%	+8.5%	-58.1% to +151.3%	31.6%
30 and above	126	+8.0%	+4.1%	-29.8% to +31.2%	34.9%
35 and above	60	+4.7%	+0.4%	-29.8% to +23.8%	45.0%

Read the median column and the one-year signal nearly vanishes: even after readings above 35, the typical next year was positive. Read the mean and the down-year column and something real appears: the average falls from +8.5% to +0.4% and the share of losing years climbs from 31.6% to 45.0%, because the left tail gets much fatter. That is the honest shape of it. A high CAPE does not tell you a bad year is coming; it tells you that the distribution of next years has moved from roughly two-in-three good to roughly coin-flip, with the worst cases worse.

Where 1929, 2000 and 2022 actually sit

Episode	CAPE then	What the next ten years paid
September 1929 peak	32.56	-1.4% a year real. January 1929, at 27.08, still managed +0.5% a year — eight months of valuation made the difference between flat and negative. The Dow's own nominal peak of 381.17 (September 3, 1929) was not regained until November 23, 1954, but an investor who bought the exact top and reinvested dividends through the deflation was back to breakeven around late 1936, per Mark Hulbert's reconstruction using Shiller's data.
January 2000	43.77	-3.03% a year real, or -26.5% cumulative, measured from Shiller's own total return series. The widely quoted "lost decade" figure of -0.95% a year is the nominal S&P 500 total return for 2000–2009 reported by Dimensional; the gap between the two is simply 2000s inflation, and mixing them is the most common error in this debate.
January 2022	36.94	Window still open until January 2032. Through October 2026, 4.75 years in, it has paid +50.5% cumulative real, about +9.0% a year. It is not a finished observation and should not be used as one.

One claim that circulates about the 2022 episode needs correcting, because an earlier version of this page repeated it: it is not the case that the market has been in CAPE-above-35 territory continuously since January 2022, nor that it has tripled since then. The ratio fell to 27.08 by October 2022; of the 58 months from January 2022 through October 2026, 31 printed below 35 and 14 printed below 30. The current cluster above 40 began in May 2026, not in 2022. And the real total return over those 4.75 years is 1.5 times your money, not three times.

The ten-year window that closed this month

There is a useful check available every month that commentary almost never runs: look at the window that just finished. October 2016 started with a CAPE of 26.53 — in the 25-to-30 band, well above the long-run mean, and at the time widely described as expensive. The ten years to October 2026 delivered +11.74% a year real, +203.4% cumulative. That is the best outcome in the 25-to-30 band's entire range and a reminder that "above average" and "above 40" are not the same statement at all. The bands in the first table only turn reliably negative at 30, and only turn uniformly negative at 40.

What a reading this high changes for a long-term portfolio

If the decision in front of me were how to set a planning assumption rather than whether to trade, I would use the band table directly and leave the timing question alone, for three reasons the data supports. First, the ten-year column is where the metric has demonstrated skill, so a retirement projection built on the long-run real equity return of roughly 6.5% is starting from the wrong row of the table; the 30-and-above row says something closer to zero, and the 40-and-above row says negative. Lowering an assumption costs nothing if it turns out to be too conservative. Second, the one-year table shows the cost of acting on the same signal as a trade: the median next year after a reading above 35 was positive, so an exit rule triggered at 35 would more often than not have forgone a gain. Third, the sample is one episode wide at this level, which argues for adjusting expectations by degrees rather than treating minus 3.4% a year as a forecast.

The counter-argument deserves its own sentence rather than a dismissal. The excess CAPE yield in the same file is 0.56%, meaning the earnings yield on this market barely exceeds the real yield on a 10-year Treasury now paying 5.24% — so the alternative to equities is genuinely better paid than it was in 1999, when the comparison was a stretched multiple against a bond market that at least offered a real return. Whether that makes the current reading more dangerous than 1999 or less depends entirely on earnings from here, which no price-to-earnings ratio can tell you.

Tip: A valuation signal only helps if you write down what you decided and why, so you can check yourself years later rather than rewriting the memory. An investment journal is a simple place to log each allocation change, and a good stock market investing book gives the longer history these ratios are drawn from. (These are Amazon Associate links — we may earn a small commission on qualifying purchases.)

FAQ

What is the current Shiller CAPE?

40.71, for October 2026, in the monthly series Shiller publishes himself. The highest reading of the current episode is 40.99, from August 2026. Daily figures near 41 that appear on other sites are separate calculations built on the same method, not values from this series.

Is a CAPE of 40 a sell signal?

The record does not support using it as one. All 21 months that previously printed 40 or more were followed by negative ten-year real returns, which is a strong statement about the decade. Over 12 months the same readings produced a median of +4.7% real, and the very first month above 40 — January 1999 — was followed by a +12.5% year. What does change is the risk of a bad year, from 31.6% across all months to 45.0% after readings above 35.

How often is CAPE updated?

Monthly. Shiller's workbook uses a monthly average of daily closes for the price input, with the current month carrying the latest price available when the file is rebuilt, so the ratio does not move day by day. This page is refreshed against that file once a month; it was last read on November 5, 2026.

Has CAPE ever been higher than it is now?

Yes, but only in 1999 and 2000. The all-time high is 44.20, in December 1999. Across the whole 1881–2026 record, only 27 of 1,750 months reached 40 or above, which is 1.5% of the series.

What ten-year return should I assume from here?

The band table is the most defensible starting point, and it points to roughly zero to slightly negative real returns rather than the long-run average: a median of minus 1.1% a year for starts above 30 and minus 3.4% a year for starts above 40. Treat that as a reason to lower a planning assumption, not as a forecast — the 40-and-above figure rests on one historical episode, and the band from 25 to 30 produced an outcome as good as +11.8% a year.

Sources

This article is for general investor education, not personalized investment advice. It is not a recommendation to buy, sell, or hold any security. Past performance, including the historical episodes discussed above, does not predict future returns. Figures are as of the dates stated and will change. This is not a recommendation to buy, sell, or hold any specific stock, fund, or index, and it does not account for any individual reader's goals, time horizon, or risk tolerance.

Start Here: every analysis by topic · About · Privacy Policy

Search This Blog

US Stock Insights