Survivor Bias in Long-Term Return Data
The 10% average return of U.S. stocks over the past century is cited so often it feels like a law of nature. But there's a silent asterisk: that return applies only to companies, sectors, and markets that survived. Every company that went bankrupt, every sector that collapsed, every country that lost its capital markets to war or revolution—they're missing from the "average." The graveyard of failed investments is invisible in historical data. This is survivor bias, and it systematically overstates the returns available to investors.
Understanding survivor bias is essential because it challenges the assumptions behind every long-term return forecast, retirement plan, and asset allocation model built on history.
Quick definition
Survivor bias is the systematic overestimation of returns that occurs when historical data includes only the winners—companies, funds, markets, and sectors that survived to be measured. Failed investments, delisted companies, defunct markets, and closed funds are excluded, making the measured average return higher than the true average return to investors who would have invested in all available options.
Key takeaways
- The "10% historical U.S. stock return" excludes no U.S. companies (because U.S. markets survived), but makes a strong implicit assumption: that the U.S. was lucky
- Survivors from other countries and eras had much lower returns; many markets declined 70–90% and never recovered (Russia 1917, Germany 1933, China pre-1990)
- Within the U.S., survivor bias operates at the fund level (closed funds are removed from history), the sector level (energy, utilities have been replaced; new sectors emerge), and the stock level (bankrupt companies earn −100%)
- The true "global investor" return across all markets over the past century is substantially lower than the 10% U.S. number
- Survivor bias inflates expectations for emerging markets most severely, because the future is unknown (today's winners might be tomorrow's failures)
How Survivor Bias Works
Imagine an investor in 1920 deciding between investing in U.S. stocks or German stocks, both of which had posted solid 5-year returns. Here's what actually happened:
Decision tree
U.S. stocks: Continued to grow, weathered the 1929 crash, recovered over the 1940s–2000s. An investor who held through everything earned ~9–10% annualized.
German stocks: Lost 98% of value in the 1923 hyperinflation. Investors who stayed through recovered somewhat in the 1924–1929 period, then lost again in the 1933–1945 Nazi era. An investor who survived (and there weren't many) might have earned 0–1% annualized over the full century, if at all.
In today's data, the German experience is often excluded or downplayed (because it's "extreme") or the data ends before the worst losses (1945–1950s data is harder to find). The U.S. experience is comprehensive because U.S. markets survived intact. The comparison between 10% U.S. and the true global return looks something like 10% U.S. vs. 4–5% global including failures.
This is survivor bias in action: The 10% figure looks normal and repeatable because the survivor (the U.S.) survived. But it was contingent on the U.S. avoiding the fates of other markets.
The Japanese Experience
A more recent example: Japan's stock market in the 1980s was considered the future. Japanese companies were buying U.S. real estate, Japanese stocks were reaching valuations never seen before, and Japan was forecast to overtake the U.S. economically.
In 1989, the Nikkei index peaked at 38,957.
As of 2025, the Nikkei has never recovered to those levels. Japanese investors who invested heavily in 1989 have earned zero capital appreciation over 36 years, with dividend income slowly accumulating to perhaps 2–3% annualized.
In historical datasets from 1980 to 2000, Japan looked like a high-return market (the 1980s boom). But a true time-average investor (1980–2025) earned much less. If you're building a model for future returns and you include data through 2000 (the Japan success story), you're overstating what's available. If you include data through 2025 (the long stagnation), you're getting closer to truth.
This is survivor bias through period selection. Cherry-picking the upside creates the illusion of higher returns.
Fund Closures and "Graveyard Bias"
Within the U.S., the clearest example of survivor bias is mutual funds. Each year, thousands of funds are closed—the worst performers are shut down, merged into better-performing funds, or simply liquidated.
In historical databases, closed funds often disappear. A database of "mutual funds in existence today" that looks back 20 years is skewed upward, because it doesn't include the 40% of funds that were closed because they underperformed.
This is called graveyard bias or backfill bias. When you see "the average mutual fund returned 6% over the past 20 years," that's the average of funds that survived. Funds that closed due to poor performance are invisible.
A corrected study including closed funds typically shows lower average returns, higher dispersion (some funds much better, some much worse), and worse odds of picking a future winner (past outperformance predicts future underperformance more often than persistence).
Sector and Industry Survivor Bias
Industries and sectors can vanish or be replaced.
Railroads in the early 1900s: Were the dominant large-cap investment. "Buy railroad stocks" was conventional wisdom. But over the century, railroads declined as a profitable sector, replaced by automobiles, then aviation, then technology. An investor who overweighted railroads through the 1900s–2000s would have underperformed someone diversified across emerging sectors.
Historical data shows "railroads had a 6% return" (the survivors). But the railroads that failed or merged into obscurity are underrepresented. The true average for all available railroad investments would be lower.
Energy sector: In 1980, oil was expected to reach $200/barrel. Energy was the dominant sector. A portfolio overweighted to energy in 1980 would have been a disaster for decades. Today, technology dominates, and many assume tech will dominate forever. But future data might show tech was the bubble (just as energy is now seen as overweighted in the 1980s), and the historical 15%+ tech returns will be revised downward in 20 years when analysis includes the era when tech crashed.
The problem: past survivors become expected winners, and new investments in them are often poor-performing. Survivor bias makes you chase the last generation's winners.
Emerging Market Survivor Bias
This is the most critical domain for modern investors. U.S. and developed-market returns are well-documented precisely because those markets survived and maintained good record-keeping. But emerging markets—where future growth is expected—have severe survivor bias.
China's stock market (Shanghai Stock Exchange): Founded in 1990. Historical returns from 1990 onwards show solid growth. But this doesn't account for the 1900–1949 period when China had markets (Shanghai was a major trading center), those markets collapsed completely in wars and the revolution, and records were destroyed or lost. If you include the full history from 1900 to 2025, the time-average return is substantially lower than the post-1990 "recovery" suggests.
Brazil, Argentina, Russia: Each had periods of hyperinflation, currency crises, war, or political collapse that destroyed market value. Historical returns calculated from the survivors (investors who held and recovered) overstate what was available to everyone who invested.
Survivor bias in emerging markets is this: We expect emerging markets to deliver 8–10% returns (matching or exceeding U.S. returns) because past survivors did. But we have biased samples. The true long-term return across all investors, all markets, all time periods, including countries and markets that failed, is lower.
The "Extreme Event" Dismissal
A common reaction to survivor bias is: "Okay, so I shouldn't invest in markets that might fail spectacularly (like 1940s Germany). I should stick to stable, safe markets like the U.S."
But this reasoning reveals the problem: By selecting only the markets that didn't fail, you're implementing survivor bias in real time. You're saying "I'll invest in winners only," which means you're planning to avoid the losers—but you don't know which are which. Emerging markets with 8% forecast returns might include future 30-year collapses, bringing time-average returns to 2–3%.
The deeper insight: The 10% U.S. return is not repeatable. It required specific historical circumstances (no invasion, maintained property rights, inclusive institutions, steady immigration, technological dominance in new industries). Future U.S. returns might be 6–7% if none of those conditions hold as strongly. And if you're picking emerging markets expecting 8% returns, you're assuming they'll replicate the U.S. trajectory, which is itself a form of survivor bias—assuming future winners will look like past winners.
Inflation Adjustment and Real Returns
Survivor bias also hides in nominal vs. real (inflation-adjusted) returns.
The 10% nominal U.S. stock return sounds impressive. But over the past century, inflation has averaged ~2.5–3%. Real returns (adjusted for inflation) are roughly 7–7.5%, not 10%.
But there's bias here too: currency survivor bias. The 7.5% real return in U.S. dollars assumes the dollar maintained purchasing power. For investors in other countries, the foreign currency might have weakened (depreciating against the dollar), further reducing real returns.
A German investor who bought U.S. stocks in 1950 (in Deutschmarks converted to dollars) would have had returns amplified by the dollar's strength. But a German investor who bought German assets had returns hurt by the Mark's weakness (until the 1970s revaluation). Historical data showing "U.S. stocks returned 7.5% real" doesn't account for the currency effect that varies by investor nationality.
Research Documenting Survivor Bias
Academic research has extensively documented survivor bias in historical returns. Elroy Dimson, Paul Marsh, and Mike Staunton at the London Business School conducted one of the most comprehensive studies, analyzing equity returns across 19 countries over the past 113 years.
Their key finding: U.S. returns of 9.5% were exceptional; global average was closer to 5%. Many countries experienced extended periods of very low or negative returns.
The study appears in their book Triumph of the Optimists and in papers published in Journal of Financial Economics. They demonstrate how including delisted companies, closed funds, and failed markets substantially lowers historical average returns.
The U.S. Securities and Exchange Commission's Office of Investor Education publishes guidance on realistic return expectations and historical data limitations: <https://www.investor.gov/>. FINRA research on investor expectations and historical returns is available at <https://www.finra.org/investors>. The Federal Reserve publishes historical economic data and research on inflation-adjusted returns: <https://fred.stlouisfed.org/>. The Social Security Administration and U.S. Treasury provide historical inflation data essential for adjusting nominal returns: <https://www.ssa.gov/> and <https://www.treasury.gov/>.
What This Means for Your Expectations
For U.S. stock planning: The 10% historical return is real for U.S. markets through 2024, but it's not a forecast for U.S. returns going forward. Plan on 6–7% nominal (adjusting down for valuations), or 3–4% real.
For international diversification: Expecting emerging markets to deliver 8% returns (matching historical U.S.) assumes they'll replicate the U.S. path. More defensible: 4–5% nominal, expecting some survivors benefit but also realistic risk of failure in some positions.
For bonds: Historical bond returns are lower than equity returns, but the comparison includes failed bond defaults (countries that defaulted, companies that went bankrupt). Current bond yields (3–4% for Treasury bonds) are a better forward estimate than historical averages.
For funds and individual stocks: Past winners are not predictive of future winners (academic research is consistent on this). Paying for performance chases survivors. Low-cost index funds that hold all survivors (not just past winners) produce better outcomes than active management selecting supposed winners.
Common Mistakes
Mistake 1: Assuming global returns will match U.S. historical returns
U.S. returned 10%, so investing globally should deliver 8–9%. Nope. Global average was 5%. U.S. was the exceptional survivor. Emerging markets might be the next winner, or they might repeat the Japan experience (great in one decade, flat for 30 years).
Mistake 2: Overweighting based on past outperformance
That emerging market fund outperformed for 5 years, so you add more. Survivor bias means past winners are often regression candidates (mean reverting). You're often chasing the last cycle, not predicting the next.
Mistake 3: Assuming closed funds just got merged into winners
Closed mutual funds are often merged into winners, but more often shut down because of poor performance. The graveyard bias means "mutual funds that survived" returns are rosier than "all mutual funds investors could have bought."
Mistake 4: Using nominal returns for long-term planning
Nominal 7% stock returns are nice, but real returns (adjusted for inflation) are 4–5%. For retirement planning over 30 years with rising costs, plan on real returns, not nominal.
Mistake 5: Ignoring that the U.S. was lucky
The U.S. maintained property rights, avoided invasion, had stable currency, and innovated continuously. Not all countries did. Some will in the future; some won't. Assuming the U.S. will continue as a survivor is reasonable but not certain. Diversification across countries and asset types is a defense against losing your single-market bet.
FAQ
Q: If survivor bias overstates returns, what return should I plan on?
A: For U.S. stocks: 6–7% nominal (3–4% real). For global stocks: 5–6% nominal (2–3% real). For bonds: 3–4% nominal (0–1% real). These are conservative relative to historical averages, but more realistic after adjusting for survivor bias, higher valuations than historical average, and lower expected inflation than the past century.
Q: Does survivor bias mean I shouldn't invest in emerging markets?
A: No, it means you shouldn't expect 8% returns matching historical U.S. returns. Emerging markets may deliver growth, but with higher volatility and the risk of extended period underperformance. A diversified allocation with some emerging market exposure is reasonable; overweighting based on past returns is not.
Q: Why don't index funds have survivor bias?
A: Index funds include living companies and exclude delisted (failed) companies, so they do have survivor bias. But they're agnostic: they hold winners and loser-to-be equally, without attempting to select. An investor in the S&P 500 in 1980 held IBM (which underperformed) and Walmart (which outperformed). Active funds selecting only winners typically underperform because they overweight past winners who mean-revert.
Q: Did Japan have survivor bias?
A: No, Japanese stocks are well-recorded. The issue is recency bias. The 1980s boom was included in historical data, making Japan look like a 8–10% return market. But an investor who lived through 1989–2025 earned much less. The "survivor" was Japan's market (it didn't disappear), but the survivor didn't thrive.
Q: What about U.S. companies that went bankrupt? Are they in the return data?
A: Historical stock indexes (S&P 500, etc.) include delisted companies in their historical calculations, so bankrupted companies are represented (as −100% returns in the final periods). But very old data (1800s) might not include all failures. For practical purposes, U.S. historical data from 1920 onward is fairly comprehensive.
Q: If 70% of my return is dividends, does survivor bias still apply?
A: Dividends are real, but survivor bias affects them too. A company that went bankrupt paid dividends during its decline, then paid zero forever. Historical dividend data weights toward survivors. Over very long horizons (50+ years), dividend yields have been 3–4% on average; expect similar going forward. But the capital appreciation component is what has survivor bias. Expect lower total returns than historical averages suggest.
Related Concepts
- Selection bias – The broader category of bias introduced by non-random sampling
- Recency bias – Over-weighting recent performance in forecasts
- Reversion to the mean – Why past winners often become future losers
- Diversification – A hedge against betting on a single survivor
- Inflation adjustment – Converting nominal returns to real returns for meaningful comparison
- Expected returns vs. historical returns – Why they differ due to valuation and survivor bias
Summary
The 10% historical return on U.S. stocks is accurate for the United States, but it's a survivor bias that inflates expectations globally. The U.S. market survived wars, recessions, and crises that devastated other markets. Global average returns were 5%, not 10%, because many investors invested in markets that failed or stagnated for decades.
Survivor bias operates at every level: across countries (only the winners' markets have good data), across funds (closed underperformers vanish from databases), across sectors (railroads declined; tech will eventually too), and across stocks (bankrupt companies are −100% but disappear from view).
The implications are clear: Don't plan on 10% returns. Expect 6–7% for U.S. stocks, 5–6% globally, 3–4% for bonds. Don't chase past winners (they're already priced in and often mean-revert). Don't overweight based on recent outperformance (survivor bias says past winners were lucky survivors, not superior pickers). And recognize that the U.S. was exceptional—not because Americans were smarter, but because the country avoided the catastrophes that befell other markets. Future returns depend on repeating that exceptional trajectory, which is uncertain.