Historical P/E ratios: what they measure and what 140 years of data show

What the P/E ratio measures

The P/E ratio (price to earnings) is the price of a share divided by the company's earnings per share. The US Securities and Exchange Commission's investor site defines it as the current share price divided by earnings per share, where the earnings are those of the past 12 months.^[1] For an index, the same sum is done for the whole index: its price level divided by the earnings of all its companies.

Example: If a share costs 100 and the company earned 5 per share over the past year, the P/E is 20. You pay 20 for each 1 of yearly profit. Turned upside down, the earnings yield is 5 divided by 100, or 5%. The higher the P/E, the lower the earnings yield.

Three versions of the same ratio

CAPE stands for cyclically adjusted price-to-earnings ratio. It divides the price by the average inflation-adjusted earnings of the previous 10 years.^[2] The aim is to smooth out booms and recessions. Aswath Damodaran, a valuation professor at NYU Stern, points out that the choice of version can be used to argue a case: bullish analysts tend to reach for the forward P/E, bearish ones for the current P/E.^[28] So always check which P/E you are looking at.

How to read a P/E

• A high P/E means you pay more for each unit of profit. Investors accept that when they expect profits to grow faster, or when they are content with a lower return. A low P/E can be a bargain – or a sign of doubt about future profits. The number alone does not say which.
• Compare a market with its own past. Accounting rules, sector mix and index construction differ between countries, so a P/E of 20 means different things in different places. Data providers also differ (see chart 5).
• Look at the earnings behind it. A P/E on one year's earnings can look low at a profit peak and absurdly high after a collapse (chart 3). A company with a loss has no meaningful P/E at all, and leaving such companies out biases an average P/E upwards.^[28]
• Put it next to interest rates. When bonds pay little, investors accept a lower earnings yield on shares. See the table further down.
• Know what is inside the numbers. Jeremy Siegel argued that changes in accounting since 1990 have depressed reported earnings in downturns, which pushes CAPE up compared with earlier decades.^[21] And because companies increasingly return cash through buybacks instead of dividends, Robert Shiller added a "total return CAPE" to his data in September 2018 to correct for it.^[22]

What 140 years of US data show

Since 1881 the US CAPE has ranged from 4.78 (December 1920) to 44.19 (December 1999), with an average of 17.42 and a median of 16.13.^[2] The January 1929 reading was 27.1, and the peak that year was 32.56 in September, shortly before the crash.^[29] By January 1932 it had fallen to 9.3. In 1982, Shiller wrote, it "sank below 7 in the summer of 1982, on the eve of a 17-year bull market".^[30]

The shape is the point. Of the 116 Januaries from 1881 to 1996, only one (1929) had a CAPE above 25. Of the 30 Januaries since, 21 have been above 25: every year 1997–2002, 2004–2007, 2015 and 2017–2026. (Our count from the January values in the series.)^[2]

The decade averages show long waves. The January CAPE averaged 11.0 in the 1980s, 24.1 in the 1990s, 28.1 in the 2000s, 25.2 in the 2010s and 34.2 in the 2020s so far.^[2]

On 2 October 2026 the CAPE stood at 41.38.^[2] Already the end-2025 reading of 39.9 was the highest in a quarter century, according to Dimensional.^[19] There are two ways to read this: the market is expensive, or the measure has drifted up because of accounting changes and buybacks (see above). Dimensional adds that many of the earlier extreme readings, those above 35, fell between 1999 and 2001, and that back-to-back 10-year windows share 119 of 120 monthly returns, so the data holds far fewer independent observations than it seems.^[19]

For the ordinary P/E on trailing earnings, multpl.com gives an average of 16.23 and a median of 15.08 over the same history.^[3]

Why the ordinary P/E can mislead

January 2009 shows why CAPE exists. Reported earnings had collapsed in the financial crisis, so the S&P 500's P/E on trailing earnings was 70.9 – not because shares were expensive, but because the denominator had shrunk. Something similar happened in January 2002, after the dot-com crash (46.2).^[3] CAPE, which averages ten years, hardly moved. It is not immune, though: Damodaran notes that after booms (when earnings peak) or crises (when earnings collapse), CAPE and the ordinary P/E can give very different pictures.^[20]

The same yardstick across markets today

On 30 September 2026 the MSCI USA index traded at 25.4 times trailing earnings and 19.4 times forward earnings. MSCI World stood at 22.7 and 17.9, Japan at 19.4 and 15.6, Europe at 16.7 and 14.2, the UK at 14.0 and 12.1, and emerging markets at 14.9 and 9.7.^[4]–^[11]

Forward and trailing numbers can tell different stories. FactSet's forward P/E for the S&P 500 was 19.0 on 2 October 2026, just below its 10-year average of 19.1 and its 5-year average of 19.8.^[23] On that measure US shares look average for the recent past; on CAPE they look close to a record. Neither is wrong. They answer different questions: the forward P/E compares the price with next year's forecast profit, CAPE with a decade of past profit.

The United States is the largest country in MSCI World, so its valuation weighs heavily on a global fund (see What is MSCI World?).

Each market against its own history

Research Affiliates' data, compiled by Monevator (30 September 2025), puts the US CAPE at 39.3 against its own long-run median of 16.5 – 2.4 times as high. India stood at 34.6 against 22.7. Japan (23.4 against 31.1) and Brazil (9.6 against 13.5) were below their own medians.^[12] These readings are a year older than the other figures on this page, so the picture may have shifted. Japan's history includes one of the most extreme bubbles on record, which we look at next.

Japan: when the starting price was the problem

The Nikkei 225 closed at 38,915.87 on 29 December 1989. Invesco, using Datastream indices, puts Japan's CAPE at 93 at that peak, against 49 for the US at its peak in early 2000.^[15] The Nikkei 225 did not close above its 1989 level until 22 February 2024 (39,098.68) – just over 34 years later.^[16] It is a price index, so dividends over the period are not counted. By February 2021 Invesco put Japan's CAPE at around 22.^[15]

Japan is the standard warning that a high starting valuation can cost a generation of returns in a single market. It is also a reason to own many markets rather than one.

Does a high P/E predict low returns?

On average, over long horizons, yes. Shiller and Jivraj's 2017 study, as reported by Evidence Investor, sorted US history by the starting CAPE. When it began below 9.6, the following 10 years averaged 9.8% a year after inflation. Between 16.1 and 17.8 the average was 5.4%. Above 26.4 it was 0.9%.^[14]

But the fit is loose. A regression of 10-year US real returns on CAPE, using 1871–2020 data, gives an R² of 29%. That means the starting CAPE explains less than a third of the variation, and Dimensional notes that it implies a market-timing signal with about an 18% chance of beating buy-and-hold.^[18] Campbell and Shiller found an R² of 30% over ten years but only 1% over one year.^[17] Damodaran puts it plainly: the predictive power is low, especially with one-year returns.^[20]

The honest summary: valuation says something about the likely range of returns over the next decade, and almost nothing about next year.

P/E and interest rates

*CAPE on 2 October 2026; the yield is the September 2026 monthly average. December CAPE values from historyofmarket.com^[25] and multpl.com^[2]; yields are monthly averages from the Federal Reserve's FRED database.^[24] The earnings yield is 100 divided by CAPE – our calculation.

In 1981 shares offered an earnings yield close to the bond yield, and they were cheap. In 2021 the earnings yield (2.6%) was above the 10-year yield (1.47%), which is why Damodaran argues that bringing in low Treasury rates can turn "overvalued" into "undervalued".^[20] Today the order is reversed. Be careful with this comparison: CAPE earnings are adjusted for inflation while Treasury yields are not, and the "Fed model" built on the comparison describes how markets price shares but is no better than long-term P/E at predicting returns.^[26] Shiller's own refinement, the excess CAPE yield, subtracts the real (after-inflation) 10-year yield from 1 ÷ CAPE.^[27]

What it means for you

For someone who saves monthly in a broad index fund, valuation is context, not a signal. It is a reason for realistic expectations: from a high CAPE, the average decade has delivered less than the long-run average, and the spread of outcomes is wide. It is not evidence that you should wait. Waiting takes two correct decisions – when to leave and when to return – and the data above show how weak the signal is over short horizons. Saving the same amount each month already buys more units when prices are low and fewer when they are high. See lump sum or monthly for the research on timing.