Market History
Time does not remove risk. It changes its shape.
Since 1871, the S&P 500 has returned about 7.1% a year after inflation with dividends reinvested. Almost no investor actually earned that number, because almost no one held for exactly that period. What matters is the range of outcomes over the time you actually hold.
Published · Insight 010
The method
Using Robert Shiller's monthly S&P data, we computed the annualized real total return for every possible holding period of 1, 3, 5, 10, 15, 20 and 30 years, starting in every month from Jan 1871 onward. That is 1,858 one-year windows down to 1,510 thirty-year windows.
"Real" means after inflation: a negative number means the investor lost purchasing power, even if the dollar balance grew.
The results
| Holding period | Worst | Median | Best | Lost purchasing power |
|---|---|---|---|---|
| 1 year | -58.1% (Jun 1931) | 8.9% | 151.3% (Jun 1932) | 30.4% of windows |
| 3 years | -35.2% (Jul 1929) | 7.6% | 39.2% (Sep 1926) | 21.2% of windows |
| 5 years | -13.2% (Sep 1929) | 7.4% | 33.3% (Sep 1924) | 18.6% of windows |
| 10 years | -5.9% (Mar 1999) | 7.1% | 20.0% (May 1920) | 11.0% of windows |
| 15 years | -2.1% (Dec 1905) | 6.9% | 15.6% (Jul 1984) | 4.4% of windows |
| 20 years | -0.2% (Jun 1901) | 6.9% | 13.6% (Apr 1980) | 0.1% of windows |
| 30 years | 1.9% (Jun 1902) | 6.8% | 11.2% (Jun 1932) | 0.0% of windows |
Annualized real total return across every rolling window, monthly start dates, Jan 1871 to Oct 2026. Windows overlap heavily, so these describe the record rather than independent odds.
Over one year, the market lost purchasing power in 30% of windows. Over twenty years, it happened in 0.1%. Over thirty, never.
What stands out
01 · The median is steady
The typical real return changes little with the holding period. What changes is how far the bad and good outcomes sit from it.
02 · A lost decade is real
The worst ten-year window started in Mar 1999, near the peak of the dot-com boom. A decade of holding did not guarantee a gain after inflation.
03 · Long is not the same as safe
Even the worst thirty-year window, starting Jun 1902, stayed positive, but it grew wealth far more slowly than the median.
04 · Starting point matters
The weakest results cluster after major tops such as 1929 and 1999, and the strongest after deep bottoms. Starting valuation and timing shaped outcomes for decades.
Read with care
Neighboring windows share most of their months, so the 1,630 twenty-year windows contain only about eight non-overlapping twenty-year stretches. The percentages describe what happened, not the probability of what will happen.
The data are monthly averages of the index, not month-end closes, and they exclude fees, taxes and trading costs. The early decades come from reconstructed index data and a smaller, very different U.S. market. Real investors also add and withdraw money over time, which changes their experience.
A long holding period has historically narrowed the range of real outcomes. It has not removed the chance of a disappointing one.
Try it
The S&P 500 Return Calculator shows the result for any two months since 1871, before and after inflation. The Drawdown Recovery Calculator shows how long every major decline took to recover.
Keep reading
Insight 009
When the average depends on ten good months.
Insight 005
When the losers disappear from the history you test.
Tool 001
Returns between any two months since 1871.
Tool 002
How deep the hole was, and how long to climb out.