An index stays at 2,000. Its annual earnings per index unit fall from 100 to 20. These are invented numbers, all adjusted for inflation.
Its trailing P/E, price divided by the latest year's earnings, jumps from 20 to 100. The price has not moved. Each dollar of the shrunken profit now costs five times as much.
Has the market become wildly more expensive, or has one bad year distorted the comparison? CAPE puts that year alongside nine others before you judge the price.
Average the earnings, not the ratios
The cyclically adjusted price-to-earnings ratio (CAPE), also called the Shiller P/E, divides an index's real price by its average real annual earnings over the previous ten years.
“Real” means adjusted for inflation. Price and earnings use the same purchasing-power units, so an older dollar of profit is comparable with a newer one.
CAPE smooths the earnings, not the price tag. Averaging ten annual P/E ratios or using next year's forecast earnings gives you a different measure.
The history here uses Robert Shiller's original CAPE for the S&P 500, a basket of large US companies. It describes that market, not every stock everywhere.
Work through a profit slump
Give our index nine earlier years of earnings of 100 each, followed by the slump year at 20.
Average earnings are (9 × 100 + 20) ÷ 10 = 92. Divide the 2,000 price by 92 and CAPE is about 21.7.
Same price, different earnings:
| Measure | Earnings used | Price | Result |
|---|---|---|---|
| Trailing P/E | 20 · latest year | 2,000 | 100 |
| CAPE | 92 · decade mean | 2,000 | 21.7 |
The latest year's earnings sit 80% below the earlier level. The decade average sits just 8% below it. One bad year no longer speaks for the whole decade.
These ten annual observations simplify Shiller's method, which uses monthly observations of annual earnings.
That long memory has a cost. If earnings stay at 20 for another year, one more strong year leaves the window. Average earnings fall to (8 × 100 + 2 × 20) ÷ 10 = 84. At the same 2,000 price, CAPE rises to about 23.8.
Smoothing a temporary slump also delays recognition of a lasting decline. The average cannot tell you which kind of slump you are seeing.
High relative to which history?
CAPE needs its own history; a cutoff borrowed from trailing P/E answers a different question. In Shiller's monthly series, the median from January 1980 through December 2024 rounds to 24.0. That is the sample's middle level. December 2024's reading of 37.7 sits well above it.
The line spends years above the sample median. Expensive can stay expensive for a long time.
The median describes what investors paid in this window. Starting the sample earlier would change it. The margin-of-safety exercise exposed the same limit in Harbor's past multiples: a common price is not necessarily a fair price.
The return evidence comes from a separate comparison. In their 2001 study using US data from 1871–2000, Campbell and Shiller found that higher starting CAPEs were associated with lower subsequent ten-year inflation-adjusted returns, including dividends. Outcomes varied widely around that relationship.
For the same future cash payments and selling price, a higher purchase price leaves you with a lower return. CAPE compares the price with past profits; future growth, dividends, and the eventual selling price remain unknown. One reading cannot give you an exact ten-year annual return.
A high CAPE can persist while prices rise; a low CAPE can precede further losses. Neither tells you when a crash will start, what will trigger it, or how far prices will fall.
A second lens uses the economy
The market-cap-to-GDP ratio, also called the Buffett indicator, compares the total market value of listed shares with annual nominal GDP. GDP measures the value added by production within a country during a period; nominal means measured at that period's prices.
For a separate hypothetical economy, use $40 trillion of listed equity value and $25 trillion of annual GDP. The ratio is 40 ÷ 25 × 100 = 160%: share values equal 1.6 times one year's domestic output.
GDP is neither listed-company profit nor the sum of every business's sales. Share values price a claim on future profits; GDP counts production during one period. Both use dollars, but they count different things.
Suppose a large private company lists its shares. Listed market value gains a business whose domestic factories were already contributing to GDP. The ratio can jump without any previously listed stock getting more expensive.
- Foreign profits. Listed companies can earn abroad; domestic GDP counts production at home.
- Private businesses. Countries differ in how much of their economy trades on an exchange.
- Profits and interest rates. A larger share of output going to profits, or lower rates used to value those profits, can support higher share values.
Nothing makes 100% a universal fair-value line. A higher ratio calls for an explanation: prices, earning power, and the businesses being counted can all change it.
Use the gauges as context
The December 2024 reading supports this: “The S&P 500's CAPE was 37.7, well above the 1980–2024 sample's 24.0 median. That gives me reason to question generous long-run return expectations. It gives me no exit date.”
That is long-horizon context. Turning it into “sell before next month's fall” adds a prediction the gauge never made. That extra leap is the problem examined in market-timing behavior.
Before comparing two published readings, check their index, earnings definition, inflation adjustment, date, and historical window. Matching the name alone is not enough.
Neither gauge values your individual holding. A broad market can be expensive while a particular company is reasonably priced, or the reverse. Value traps and growth traps returns to individual businesses, where either a low multiple or rapid growth can mislead.
In short
- CAPE divides real price by average real annual earnings over ten years, not by an average of P/Es.
- Smoothing one bad year also slows recognition of lasting changes in earning power.
- Starting valuations help frame long-run return expectations; they do not schedule a crash.
- Market cap to GDP compares share values with domestic production; 100% is no magic dividing line.
- An expensive market does not make every stock expensive.