Stock Market Returns by Decade After Inflation

Key takeaways

• Stock market returns by decade since 1930 ran from 15.76% a year after inflation in the 1950s to minus 2.83% a year in the 2000s, a spread of 18.59 points.
• The 1930s lost 0.25% a year in nominal terms and still made 1.82% a year in real terms, because US consumer prices fell 2.04% a year across the decade.
• The 1970s did the reverse. A nominal 6.05% a year became minus 1.22% real, and minus 4.84% for a sterling investor, as UK prices rose 12.40% a year.
• Starting valuation explained 41.6% of the gap between the nine completed decades. Real GDP growth explained 6.6%, and the fastest-growing decade was nowhere near the best.
• Only 9 of the 87 rolling ten-year windows since 1930 lost money after inflation, and the worst of them started in 1965 rather than in 1929 or 1999.

Stock market returns by decade, after inflation

You already know what each decade is supposed to have taught investors. The 1930s wiped people out, the 1970s were dead money, the 2000s were a lost decade, the 1990s were a party. Here is what the series actually says, in US dollars, before tax and dealing costs.

The return series is the value-weighted return on all CRSP firms listed on the NYSE, AMEX or NASDAQ, published by Kenneth French at Dartmouth. His annual file gives stock market returns by year from 1927 to 2025, so the first complete decade here is the 1930s and the last row covers the six years to the end of 2025. The deflator is the US consumer price index for all urban consumers, December to December, which lines up with those calendar years.

Three things come straight off the table. The best decade for stocks was the 1950s, at 15.76% a year real. Only two of the nine completed decades lost money after inflation, and they sit three decades apart. And across the whole 96 years the market compounded at 10.09% a year nominal and 6.77% real, while the nine completed decades average 6.73%. Not one of them landed between 6% and 10%. The long-run average is a number no decade delivered.

The 1930s made money, because prices fell faster than the market did

The crash is real, and most of it isn't in the decade. From the start of 1929 to the end of 1932 the US market lost 68.8% of its nominal value. The NBER dates that contraction from August 1929 to March 1933, 43 months, the longest in its records since the 65-month contraction that ended in March 1879. The single worst calendar year in the French file is 1931, at minus 43.77%, worse than 2008's minus 36.65%.

Now add the deflator. The December CPI index stood at 17.2 in 1929 and 13.2 in 1933, a fall of 23.3%. Over the ten years from 1930 to 1939 US consumer prices fell 2.04% a year. A nominal return of minus 0.25% a year against a falling price level is a real return of plus 1.82% a year, or 19.8% over the decade. 1933 alone returned 56.79%, the best year in the file.

That isn't a defence of the 1930s. It's a statement about where the damage sat. It sat in four years, 1929 to 1932, and the decade label straddles them. If you want the recovery arithmetic rather than the decade arithmetic, we've set out bear market recovery time on both a price and a real basis separately.

The 1970s is the decade nominal returns lie about

6.05% a year sounds like a workmanlike decade. It was a loss. US prices rose 7.36% a year from December 1969 to December 1979, so the real return was minus 1.22% a year and 11.5% of purchasing power went over the ten years. Nothing in it looks like 1931. The decade's worst year, 1974, returned minus 27.82%, which is bad and not historic.

This is the hard case for treating equities as inflation cover over any particular ten years. What matters to a holder is the real return after inflation, and through the 1970s that was negative while the nominal number stayed positive. The same test applied to metals is in our piece on the gold inflation hedge.

Compare the 2000s. Nominal minus 0.38% a year, inflation 2.52%, real minus 2.83% a year and minus 24.9% cumulatively. That is the lost decade, and it is the only one of the nine that lost money on both measures at once. Two losing decades, two different mechanisms: the 1970s was an inflation problem, the 2000s a valuation problem.

The economy was almost no guide to the decade

If the market tracked the economy, the 1930s would be bottom of the table and the 1940s top. The BEA's real GDP series starts in 1929. It has the 1930s growing 0.97% a year, the slowest decade in the series, and the 1940s growing 5.59% a year, the fastest. The 1940s returned 3.93% a year real and the 1930s 1.82%. Both sit near the bottom of the return table.

Across the nine completed decades the correlation between real GDP growth and real stock returns is 0.259, an R-squared of 0.067. Growth accounts for 6.6% of the difference between decades. The 2000s grew 1.90% a year and lost 2.83% a year; the 2010s grew 2.40%, barely faster, and made 11.65%. Whether the industries inside the index reshuffle on the same boundaries is the separate question we take up in sector returns by decade.

What you paid at the start explains more than what the economy did

Robert Shiller's cyclically adjusted price/earnings ratio, usually called CAPE, divides the index price by the average of ten years of inflation-adjusted earnings. It's the standard way of asking whether a market is expensive without letting one year of earnings decide. Here is what it read at each decade boundary, against the decade that followed.

The two cheapest starts were followed by the best decade in the table and the fourth best. CAPE 10.53 in December 1949 gave 15.76% a year, and 8.75 in December 1979 gave 11.20%. The dearest start by a wide margin, 44.20 in December 1999, was followed by the worst decade at minus 2.83%. Across the nine the correlation is minus 0.645, an R-squared of 0.416, so roughly 41.6% of the variation.

Where the valuation story breaks down

The obvious objection to that table is that it reads a mechanical rule into nine data points, and it's a good objection. Two more come out of the data itself. The denominator is a ten-year average of earnings, so one collapse in profits keeps the ratio elevated for ten years after the event that caused it. And the reading has been "expensive" for most of a working lifetime: across the 1,713 monthly observations from January 1881 the series averages 17.40, and 383 of the 405 months since January 1990, or 94.6%, sit above that average.

Look at December 1969 and December 1989. CAPE 17.33 and 17.65, about as close as two readings get two decades apart. What followed was minus 1.22% a year and plus 14.63% a year, 15.85 points apart. The 1930s started on 22.01, the third dearest start in the table, and delivered a positive real decade anyway.

So price sorts the extremes and says very little about the middle. Anyone offering a decade forecast from the 30.33 the 2020s started on is reading a relationship that has worked at 8.75 and at 44.20 and failed in both directions in between. The 2020s are running at 10.47% a year real through six years, which is that failure in progress.

Two different datasets, the same decades

One series can be wrong, so here's a second one. Shiller publishes a real total return index for the S&P composite: a different index, built a different way, deflated by his own price series. Its decade figures are 2.11% for the 1930s against our 1.82%, minus 1.41% for the 1970s against minus 1.22%, minus 3.17% for the 2000s against minus 2.83%, and 14.63% for the 1990s, identical to the French figure. The largest divergence anywhere in the nine decades is 0.90 points, in the 1950s. Same signs, same ranking, two independent constructions. Both are total-return series, and that agreement is the evidence for it: the S&P composite's dividend yield averaged 4.65% across the months from 1930 to 1979 on Shiller's own figures, so a price-only series would have trailed a reinvesting one by roughly that much.

A sterling investor had different decades

Everything above is in dollars, deflated by US prices. A UK reader's outcome depends on sterling and on UK inflation instead. Using the Bank of England's monthly dollar/sterling rate and its spliced monthly UK consumer price index, both December to December, the same dollar returns convert like this.

The 1970s is the clearest case. UK prices rose 12.40% a year against America's 7.36%, so an index that lost 1.22% a year in real dollars lost 4.84% a year in real sterling. The 1940s run the other way. The pound fell from $3.93 at the end of 1939 to $2.80 at the end of 1949, and UK inflation of 3.07% a year ran below America's 5.36%, so a 3.93% dollar decade was 9.90% a year in sterling.

The 1980s make the same point from the other side. The pound went from $2.20 at the end of 1979 to $1.60 at the end of 1989, so a UK holder's real return was 13.89% a year against 11.20% in dollars. In four of the eight decades the sterling figure beats the dollar figure, and in two of the other four it loses by more than a point. Currency isn't a rounding error over ten years, and it doesn't push in a reliable direction either.

Two caveats on that table. The sterling columns are arithmetic rather than a trade anyone could have placed: nobody could buy the whole CRSP universe in 1940, and exchange controls, withholding tax and currency costs all sit outside these numbers. And the Bank's monthly series stop at the end of 2016, which is why the last two decades have no sterling row. For the long-run picture across asset classes rather than across decades, see our historical returns by asset class.

The decade boundary does some of the work

Stock market returns by decade are a convenient slice, not a market cycle. Take every ten-year window instead. There are 87 of them from 1930 to 2025. Nine lost money after inflation, 10.3% of the total. The median window returned 8.01% a year real. The best started in 1949 at 16.95% and the worst started in 1965 at minus 4.22%, a worse ten years than either losing calendar decade managed.

The negative windows also cluster. Seven of the nine began between 1965 and 1973, and the other two began in 1999 and 2000. Move the cut by a year and the label moves with it: the 2000s lost 2.83% a year, the ten years from 2001 made 0.16% a year, and the ten from 2002 made 1.26%. The lost decade is partly a product of where the calendar put its boundary.

What this data cannot tell you

These are stock market returns by decade for one country, over 96 years, which gives nine completed observations. Nine is a small sample, and the correlations above are description rather than estimation. The sample contains one Depression, one inflation of the 1970s kind and one valuation bust, so any lesson drawn from a decade is a lesson drawn from a single episode.

Returns here reinvest dividends gross, with no tax, no dealing costs and no fund fee, and no fund tracked this universe in 1930. The CPI is one measure of the price level, and your own basket isn't the national one. A backtest measures what happened; it isn't a forecast.

What would change the conclusion

Three things, in order of how much they'd move it. The first is inflation. Every row in the first table is really a statement about the deflator: the 1970s lost money with a positive nominal return, and the 1930s made money with a negative one. If US inflation ran at 7.36% again, a 6% nominal decade would be another loss.

The second is the valuation relationship. The 2020s started at a CAPE of 30.33 and are running at 10.47% a year real. Score that unfinished decade as it stands and the correlation across ten decades is minus 0.538 rather than minus 0.645. Four more years at this pace would weaken it further; a repeat of the 2000s would restore it.

The third is the boundary itself. The NBER's shortest contraction on record, February to April 2020, lasted 2 months, and the market finished that year up 24.03%. Cycles no longer line up with calendar decades in any useful way, if they ever did. What's worth watching isn't which decade we're in. It's the two inputs that decided every row above: what you paid at the start, and what prices did while you held.