Is the Stock Market Overvalued? Nine Valuation Measures Since 1881
US stocks are very expensive by historical standards. The valuation score is 97 out of 100, with all 9 measures in the top tenth of their own history. That has been informative about the next decade’s returns and has given little guidance about the next decline.
The average of 9 percentiles, each ranking one measure against its own history; not a measure of how far stocks are overvalued or of the odds of a decline. Record: 97.4 in August 2026.
Historical median across overlapping monthly starts, after inflation and including dividends; 3 completed independent decades.
Today's reading
As of September 2026, the score of 9 valuation measures stands at 97.3 out of 100, just below its record of 97.4 in August 2026, and Tobin's Q, price to sales and market cap to GDP are at a record. Ranked against earlier history, it reads 96. Since 1901, months scoring 90+ on that version were followed by a median real return of 0.5% a year over the next decade across overlapping monthly starts, against 6.6% for all months; the 3 completed independent decades returned 1.4%, 6.5% and 6.2% a year.
Sources, methodology & freshnessShiller S&P data (1881+), Kenneth French CRSP book-to-market portfolios (1926+), Federal Reserve Z.1 and BEA via FRED (1945+) · Monthly; quarterly and annual inputs carried until their next releaseData as of 2026-09-01 · Open ↓Close ↑
Nine measures in one score since 1881
The solid line ranks each measure against its whole record. The dashed line ranks it only against earlier history. They differ for older peaks because several measures have drifted upward for decades: ranked against today, 1929 scores 76; ranked against its own past, it scores 99. Hover for both scores and how many measures each one averages.
“Earlier history” uses historical release-lag assumptions and today's revised data, including Shiller's earnings and dividends interpolated from quarterly figures; it does not reconstruct what investors could actually see at the time.
The same four measures throughout
Some of the movement above comes from adding measures over time. This version uses only the four that exist since 1881, the CAPE, trailing P/E, dividend yield and distance above trend, so its composition never changes. It reads 96 today, against a record of 99 in July 1999.
The nine measures today
Each measure's latest reading and its rank against its own record, where 100 is the most expensive. Quarterly and annual measures publish with a lag, so their dates differ. Several share a numerator, the market's price or value, so nine agreeing measures corroborate one another without being nine independent pieces of evidence.
Dividend yield ranks in reverse: a lower yield means a higher valuation. Trailing P/E peaked in 2009 because earnings collapsed while prices fell less, the measure's known weakness in recessions. Price to book is a proxy built from Ken French's CRSP book-to-market portfolios; price to sales divides nonfinancial corporate equity by gross value added.
What real returns followed each score?
Real returns here are after inflation and include dividends. Months are grouped by the score ranked against earlier history. The medians use every monthly start, which overlap heavily; the completed independent decades, ten-year windows that do not overlap, show how few separate outcomes stand behind each figure. Scores of 90 and above covered 14% of months.
The completed decades after scores of 90 and above began in 1928 (+1.4% a year), 1959 (+6.5% a year) and 1996 (+6.2% a year); the one from 2018 has not finished. The overlapping-month median of +0.5% sits below all of them: most of its starting months fall in the long expensive stretches of the 1960s and the late 1990s, and the later months of each stretch, nearer the 1968 and 2000 peaks, did worse than the first month, where each completed decade starts. Neither figure is a forecast. The median next-year return after scores of 90 and above, +7.1%, was close to the +8.8% for all months.
S&P 500 real total return from Shiller's price, dividends and CPI. Download everything: valuation_composite.json.
The average of 9 percentiles, each ranking one measure against its own history; not a measure of how far stocks are overvalued or of the odds of a decline. Record: 97.4 in August 2026.
Historical median across overlapping monthly starts, after inflation and including dividends; 3 completed independent decades.
How Stock Market Valuation Works
- 1Collect nine valuation measuresThree come from Robert Shiller’s monthly S&P data since 1881 (the CAPE, price to trailing earnings and price to dividends). Price to book comes from Ken French’s CRSP book-to-market portfolios since 1926. Tobin’s Q, price to sales, market cap to GDP and market cap to profits come from the Federal Reserve and BEA since the late 1940s. The ninth is how far the real S&P 500 total return sits above its long-run trend line.
- 2Rank each against its own historyEach measure becomes a percentile: the share of its own readings below the one being ranked. That puts a P/E and a ratio to GDP on one scale, from 0 for the cheapest reading on record to 100 for the most expensive.
- 3Average the percentilesThe valuation score is the average of the measures available each month. It is not a percentile itself and does not measure how far stocks are overvalued. Before 1926 only the three Shiller measures and the trend exist, so early scores rest on fewer inputs; the page shows how many are in the average and adds a version built from the same four measures throughout.
- 4Rank against earlier historyA percentile against the whole record ranks 1929 against data from 2026. The page also ranks each measure only against earlier readings, with release lags assumed and at least twenty years of history, and grades what happened next on that version. It uses today’s revised data, so it approximates what investors saw without reconstructing it.