FinObservatory

Equity valuation

What the market costs, and what that has actually predicted

Robert Shiller’s file carries CAPE and the realised next-ten-year annualised real return in the same row, so the claim can be tested against its own source rather than against a redrawn chart. It is tested below, on the 1,605 months where both numbers exist, and then tested again on the 14 windows that do not overlap each other. Damodaran’s 2026-01-05 snapshot supplies the cross-section: cost of capital, beta and four multiple families for 94 industries.

35.2
CAPE, 2024-09
1,679 / 1,725
Monthly readings at or below it
16.5
Median CAPE since 1881-01
0.286
R-squared on 14 independent windows
5.4 points
Residual SD of that fit, per year

The level

CAPE is 35.23 in 2024-09, the last month of the file. 1,679 of the 1,725 monthly readings since 1881-01 are at or below that. The median of the series is 16.52, its trough 4.78 in 1920-12, its peak 44.20 in 1999-12.

0510152025303540451900192019401960198020002020median 16.51999-12 peak 44.22024-09 35.2

Source: Robert J. Shiller, Irrational Exuberance monthly data (Yale) CAPE is real price divided by the mean of the trailing 120 months of real earnings. Methodology

Does a high CAPE predict a low return?

Shiller publishes the answer in his own file. The column real_10yr_ann_stock_return holds the annualised real return the market went on to deliver over the ten years after each month. CAPE is null for the first 120 months of the file and the forward return is null for the last 120, so 1,605 months carry both, 1881-01 to 2014-09. That is the entire sample.

Regressed on the CAPE earnings yield (1 / CAPE), those 1,605 months give an R-squared of 0.274. The sign is the one the folklore predicts. The magnitude is not: 72.6% of the variation in ten-year outcomes is left unexplained by the starting valuation.

1,605 overlapping monthly windows14 non-overlapping windowsfit on 1,605 overlapping monthsfit on 14 non-overlapping windows
-10%-5%0%5%10%15%20%051015202530354045CAPE at the start of the window2024-09 CAPE 35.2

Source: Robert J. Shiller, Irrational Exuberance monthly data (Yale) Both curves are y = a + b / CAPE. Overlapping fit: a = 0.0038, b = 0.8908. Non-overlapping fit: a = 0.0092, b = 0.8108. Methodology

Why that scatter overstates its own evidence

Two consecutive rows of the forward-return column describe ten-year windows that share 119 of their 120 months. The lag-1 autocorrelation of that column is 0.993; twelve months apart it is still 0.857. The 1,605 points in the scatter above are not 1,605 independent observations, and the t statistic computed as though they were, 24.6, means nothing.

Take the first testable month and every 120th month after it and 14 strictly non-overlapping windows remain: the first begins 1881-01, the last ends 2021-01, and no two of them share a month. On those 14, R-squared is 0.286 and t is 2.19 on 12 degrees of freedom.

The R-squared barely moved: 0.274 on 1,605 overlapping months, 0.286 on 14 independent ones. Overlap did not inflate the fit. What it inflated is the confidence: t falls from 24.6 to 2.19. The two-sided 5 percent critical value of Student’s t with 12 degrees of freedom is 2.179, so the relationship clears the bar by 0.016.

Every specification, both samples

Predictor of the next 10 years’ real returnSamplenrR-squaredt
CAPEoverlapping1,605-0.5100.260-23.74
CAPEnon-overlapping14-0.4800.230-1.89
CAPE earnings yield (1 / CAPE)overlapping1,6050.5240.27424.60
CAPE earnings yield (1 / CAPE)non-overlapping140.5350.2862.19
log CAPEoverlapping1,605-0.5310.282-25.12
log CAPEnon-overlapping14-0.5320.283-2.18
Excess CAPE yield, against the 10-year excess returnoverlapping1,6050.5200.271n/a
Excess CAPE yield, against the 10-year excess returnnon-overlapping140.5160.266n/a

Source: Robert J. Shiller, Irrational Exuberance monthly data (Yale) t is r * sqrt(n - 2) / sqrt(1 - r squared). Reported for the overlapping sample only to show how far wrong it goes. Excess CAPE yield is Shiller's own column, the CAPE earnings yield minus the real ten-year bond yield. Methodology

What 14 observations can and cannot carry

Refit the same regression 14 times, each time dropping one of the 14 independent windows. R-squared moves between 0.106 and 0.376; t moves between 1.14 and 2.57. The two-sided 5 percent critical value at 11 degrees of freedom is 2.201, and 11 of the 14 refits fall below it. Dropping the 1921-01 window alone takes R-squared to 0.106.

That is the honest state of the evidence. The relationship is real in sign, modest in size, and, once the overlapping windows are removed, fragile: removing one decade from the sample costs it 5 percent significance in 11 of the 14 cases.

Window droppednR-squaredtClears 2.201?
1921-01130.1061.14no
2001-01130.2261.79no
1981-01130.2792.07no
1881-01130.2802.07no
1961-01130.2812.08no
1931-01130.2812.08no
1901-01130.2852.10no
1941-01130.2862.10no
1971-01130.2922.13no
1951-01130.2932.13no
1891-01130.2942.14no
1991-01130.3602.49yes
2011-01130.3752.57yes
1911-01130.3762.57yes

Source: Robert J. Shiller, Irrational Exuberance monthly data (Yale) Methodology

By quintile, and at today’s level

Cutting the 1,605 overlapping months into CAPE quintiles gives a mean realised return that falls at every step, from 10.9% a year in the cheapest fifth to 3.2% in the most expensive. The ranges still cross: the worst outcome in the cheapest quintile was 1.2% a year, the best in the most expensive was 13.1%. 0 of the 321 months in the cheapest quintile went on to a negative ten-year real return; 94 of the 321 in the most expensive did.

CAPE quintileMonthsCAPE rangeMean real returnWorstBestNegative
1 (cheapest)3214.8 to 11.110.9%1.2%20.0%0 of 321
232111.1 to 14.47.2%-4.2%15.4%34 of 321
332114.4 to 17.46.6%-4.6%16.1%33 of 321
432117.4 to 21.15.5%-4.0%14.6%32 of 321
5 (most expensive)32121.1 to 44.23.2%-5.9%13.1%94 of 321

Source: Robert J. Shiller, Irrational Exuberance monthly data (Yale) Quintiles of the overlapping sample, so the months within a quintile are not independent of each other either. Methodology

The precedent for today’s CAPE

47 months in the whole series carry a CAPE at or above today’s 35.2, spread over 7 calendar years: 1998, 1999, 2000, 2001, 2021, 2022, 2024. Only 34 of those 47 are old enough to have a completed ten-year window, and all 34 of them sit inside one 36-month stretch, 1998-03 to 2001-02. Those windows returned between -5.9% and 1.1% a year in real terms, mean -2.6%; 29 of the 34 were negative.

Fitted at today’s CAPE, the 14-window regression returns 3.2% a year real over the coming decade. The residual standard deviation of that fit is 5.4 points a year. Shiller’s excess CAPE yield stands at 1.83%, against a ten-year Treasury of 3.84%.

Source: Robert J. Shiller, Irrational Exuberance monthly data (Yale) Methodology

The industry cross-section

Damodaran’s data set is one snapshot. Every row carries the vintage 2026-01-05 and the file has no other date column, so nothing below can be sliced through time. It covers 94 industries in two regions, plus two aggregate rows that are not industries. The 94 US industries hold 5,994 firms and sum exactly to the count the Total Market row reports; the Global set holds 48,156.

Cost of equity here is a formula, not an estimate

Across the 94 US industries, cost of equity equals 3.950% + 4.460% × beta, with a maximum absolute residual of 3.5e-17: exact to machine precision. Across the 94 Global industries the same identity holds with a slope of 5.630%. Read as CAPM, the intercept is the risk-free rate and the slope is the equity risk premium, and both are the same number for every industry in a region. The cross-section of cost of equity therefore carries exactly the information in the cross-section of beta and nothing else. Cost of debt is coarser still: it takes 4 distinct values across the 94 US industries and 4 across the 94 Global ones. Once beta and that bucket are given, the only remaining sources of spread in the cost of capital are leverage and the tax rate.

Source: Aswath Damodaran, NYU Stern, current industry data Methodology

US industry cost of capital runs from 4.36% (Utility (General)) to 10.66% (Software (Internet)), median 7.09%. Global runs from 4.99% (Banks (Regional)) to 15.31% (Semiconductor Equip), median 7.67%.

The PE column that is not the market’s PE

Damodaran ships two different things both called PE. The current, trailing and forward columns are averages across firms; the aggregate columns are total market capitalisation divided by total net income. For the US Total Market row they read 72.2 and 34.2. The second is what a dollar of the market’s earnings costs. The first is not.

The same file says why an average across firms is not the market’s ratio: 57.2% of the 5,994 US firms had negative trailing net income, and 56 of the 94 US industries had more than half their firms losing money. Trailing PE is null in exactly the 3 US industries in which every firm lost money, and nowhere else in the US (0 mismatches in either direction). An average-of-firms PE is an average over whichever firms had positive earnings.

The aggregate ratio comes in two forms. Across all 5,994 US firms it is 34.2; across the profitable ones only it is 26.6.

Source: Aswath Damodaran, NYU Stern, current industry data Methodology

All 94 industries

US figures. Every row links to the full US and Global detail.

IndustryUS firmsBetaCost of capitalAggregate PEP/BEV/EBITDAMoney-losing firms
Advertising521.217.81%460.54.615.179%
Aerospace/Defense790.957.60%78.17.933.451%
Air Transport231.196.72%26.12.810.061%
Apparel350.947.13%41.23.913.963%
Auto & Truck331.469.38%148.57.751.882%
Auto Parts351.348.18%47.01.99.149%
Bank (Money Center)150.764.98%15.01.6n/a7%
Banks (Regional)5680.404.98%15.01.1n/a14%
Beverage (Alcoholic)140.816.48%24.31.98.364%
Beverage (Soft)270.646.33%26.96.918.270%
Broadcasting240.475.09%n/a1.77.771%
Brokerage & Investment Banking321.176.08%24.42.8n/a22%
Building Materials411.117.85%14.73.611.829%
Business & Consumer Services1550.897.23%30.85.316.257%
Cable TV90.745.20%8.71.36.444%
Chemical (Basic)291.016.22%18.51.16.976%
Chemical (Diversified)40.855.23%n/a1.18.7100%
Chemical (Specialty)590.977.25%48.12.414.759%
Coal & Related Energy161.078.41%52.22.219.581%
Computer Services641.097.83%30.24.716.555%
Computers/Peripherals361.359.71%36.234.126.261%
Construction Supplies401.158.29%21.95.516.630%
Diversified200.887.30%13.41.89.880%
Drugs (Biotechnology)4961.148.49%n/a8.251.590%
Drugs (Pharmaceutical)2280.987.85%58.96.618.685%
Education320.786.75%39.32.612.050%
Electrical Equipment1121.258.99%n/a6.535.075%
Electronics (Consumer & Office)80.877.63%n/a3.9n/a100%
Electronics (General)1140.977.85%56.75.025.960%
Engineering/Construction481.218.69%36.95.521.840%
Entertainment920.837.13%n/a4.824.483%
Environmental & Waste Services530.957.43%37.76.617.679%
Farming/Agriculture351.137.27%24.22.416.766%
Financial Svcs. (Non-bank & Insurance)1760.975.00%22.93.789.941%
Food Processing780.615.79%16.81.89.660%
Food Wholesalers130.876.53%27.14.912.354%
Furn/Home Furnishings270.826.53%85.42.513.752%
Green & Renewable Energy150.866.04%66.41.313.587%
Healthcare Products2040.917.54%46.54.423.476%
Healthcare Support Services1040.876.83%23.62.811.863%
Heathcare Information and Technology1151.118.22%109.94.025.977%
Homebuilding300.917.27%10.21.68.420%
Hospitals/Healthcare Facilities310.806.19%16.75.610.752%
Hotel/Gaming631.087.36%31.712.317.754%
Household Products1100.827.03%24.25.813.772%
Information Services150.927.00%25.83.812.840%
Insurance (General)210.676.34%30.03.518.324%
Insurance (Life)200.645.60%10.21.411.225%
Insurance (Prop/Cas.)570.485.78%16.52.010.114%
Investments & Asset Management2830.666.13%24.82.547.158%
Machinery1050.967.70%26.24.517.543%
Metals & Mining731.048.20%45.64.313.785%
Office Equipment & Services141.337.92%55.62.910.043%
Oil/Gas (Integrated)40.305.07%15.71.78.10%
Oil/Gas (Production and Exploration)1420.726.25%15.21.46.259%
Oil/Gas Distribution230.675.78%21.62.913.443%
Oilfield Svcs/Equip.970.957.04%19.01.99.658%
Packaging & Container191.026.75%12.72.210.837%
Paper/Forest Products60.966.93%11.01.87.050%
Power460.485.01%23.12.313.311%
Precious Metals560.847.47%50.93.617.386%
Publishing & Newspapers190.565.95%21.12.312.953%
R.E.I.T.1900.645.32%56.12.024.334%
Real Estate (Development)140.845.82%23.31.012.279%
Real Estate (General/Diversified)120.816.25%41.71.225.158%
Real Estate (Operations & Services)540.977.41%405.53.137.059%
Recreation491.026.76%96.04.011.757%
Reinsurance10.585.64%18.61.010.90%
Restaurant/Dining640.927.16%36.474.721.447%
Retail (Automotive)340.946.78%27.47.517.656%
Retail (Building Supply)141.549.51%23.4132.215.950%
Retail (Distributors)620.957.22%25.44.016.745%
Retail (General)230.817.27%43.07.220.926%
Retail (Grocery and Food)151.127.24%17.73.99.140%
Retail (REITs)260.625.57%35.72.117.98%
Retail (Special Lines)941.098.01%27.77.316.562%
Rubber& Tires30.534.48%42.70.86.7100%
Semiconductor661.5210.55%77.413.342.762%
Semiconductor Equip311.409.89%35.99.726.245%
Shipbuilding & Marine80.756.69%12.91.58.563%
Shoe111.028.01%22.56.117.236%
Software (Entertainment)771.038.44%34.59.126.271%
Software (Internet)291.6910.66%n/a10.9100.472%
Software (System & Application)3091.289.34%49.19.131.870%
Steel191.067.76%23.61.911.037%
Telecom (Wireless)120.545.48%21.43.510.875%
Telecom. Equipment570.927.72%48.08.427.268%
Telecom. Services390.635.39%14.41.67.579%
Tobacco100.796.94%19.0n/a15.150%
Transportation190.866.72%24.44.816.753%
Transportation (Railroads)40.987.27%21.55.813.625%
Trucking261.017.52%34.62.911.250%
Utility (General)140.244.36%21.71.815.00%
Utility (Water)140.414.93%22.02.015.729%

Source: Aswath Damodaran, NYU Stern, current industry data Aggregate PE is total market capitalisation over total net income across all firms in the industry, not the average of firm PEs. Blank cells are missing in the source. Vintage 2026-01-05. Methodology

What this data cannot tell you

  • The two halves of this page are different vintages. Shiller’s file ends 2024-09. The Damodaran snapshot is dated 2026-01-05. No sentence should put a number from one beside a number from the other, and none here does.
  • CAPE is not comparable across countries or accounting regimes. It is a ratio of a price index to a ten-year average of reported earnings. Two markets whose firms write down goodwill differently, or whose index composition differs, produce CAPEs that cannot be ranked against each other. This page shows one market only: the S&P composite.
  • Shiller’s earnings are trailing and smoothed. The denominator of CAPE at any month is the mean of the previous 120 months of real earnings, so a CAPE reading responds to a profit collapse slowly and by construction. earnings_nominal is missing for the last 3 months of the file (it ends 2024-06) while cape runs to 2024-09.
  • The forward-return test rests on 14 independent windows. Everything else is the same data counted more than once. 11 of the 14 leave-one-out refits fail to clear the 5 percent bar. Any claim about significance here is a claim about 12 degrees of freedom.
  • Damodaran cannot be time-sliced. One vintage, 2026-01-05. There is no history in the file, so no industry trend, no re-rating, and no before-and-after can be computed from it.
  • Cells are missing. All 35 data columns these pages display were counted for nulls across the 188 industry-region rows of their tables. The 14 that have any are named in full below. Nulls are shown as “n/a” and never imputed:
    ColumnNullsIndustry (region)
    agg_mktcap_to_net_income_all_firms10 of 188Broadcasting (Global); Drugs (Biotechnology) (Global); Real Estate (Development) (Global); Broadcasting (US); Chemical (Diversified) (US); Drugs (Biotechnology) (US); Electrical Equipment (US); Electronics (Consumer & Office) (US); Entertainment (US); Software (Internet) (US)
    ev_ebit_all_firms9 of 188Bank (Money Center) (Global); Banks (Regional) (Global); Brokerage & Investment Banking (Global); Bank (Money Center) (US); Banks (Regional) (US); Brokerage & Investment Banking (US); Coal & Related Energy (US); Electronics (Consumer & Office) (US); Software (Internet) (US)
    peg_ratio9 of 188Shipbuilding & Marine (Global); Chemical (Diversified) (US); Electronics (Consumer & Office) (US); Insurance (Life) (US); Paper/Forest Products (US); Real Estate (Development) (US); Real Estate (General/Diversified) (US); Reinsurance (US); Rubber& Tires (US)
    ev_ebit_pos_ebitda_firms8 of 188Bank (Money Center) (Global); Banks (Regional) (Global); Brokerage & Investment Banking (Global); Bank (Money Center) (US); Banks (Regional) (US); Brokerage & Investment Banking (US); Coal & Related Energy (US); Electronics (Consumer & Office) (US)
    roic8 of 188Bank (Money Center) (Global); Banks (Regional) (Global); Brokerage & Investment Banking (Global); Financial Svcs. (Non-bank & Insurance) (Global); Bank (Money Center) (US); Banks (Regional) (US); Brokerage & Investment Banking (US); Financial Svcs. (Non-bank & Insurance) (US)
    ev_ebitda_all_firms7 of 188Bank (Money Center) (Global); Banks (Regional) (Global); Brokerage & Investment Banking (Global); Bank (Money Center) (US); Banks (Regional) (US); Brokerage & Investment Banking (US); Electronics (Consumer & Office) (US)
    ev_ebitda_pos_ebitda_firms6 of 188Bank (Money Center) (Global); Banks (Regional) (Global); Brokerage & Investment Banking (Global); Bank (Money Center) (US); Banks (Regional) (US); Brokerage & Investment Banking (US)
    agg_mktcap_to_net_income_profitable_firms3 of 188Chemical (Diversified) (US); Electronics (Consumer & Office) (US); Rubber& Tires (US)
    expected_growth_5yr3 of 188Chemical (Diversified) (US); Real Estate (General/Diversified) (US); Reinsurance (US)
    std_dev_operating_income_10yr3 of 188Bank (Money Center) (US); Brokerage & Investment Banking (US); Electronics (Consumer & Office) (US)
    trailing_pe3 of 188Chemical (Diversified) (US); Electronics (Consumer & Office) (US); Rubber& Tires (US)
    roe2 of 188Retail (Building Supply) (US); Tobacco (US)
    current_pe1 of 188Electronics (Consumer & Office) (US)
    price_to_book1 of 188Tobacco (US)
  • The multi-year beta columns are withheld. yr_2022 through yr_2025 and average_beta_multiyear exist in the source. The US spreadsheet labels the average column “Average (2022-2026)” and the Global spreadsheet labels the same column “Average (2020-24)”, so it does not span the same years in the two regions. It is also not the arithmetic mean of the four year columns: across the 192 rows it equals that mean in 0 of them. Rather than print a label that would be wrong for one region or the other, the five columns are omitted.

Full construction, the trap register and the reproduction checks are on the methodology page.