Do 200 Years of Stock Returns Still Matter? Yes, but Not as a Forecast

The familiar long-run chart shows U.S. stocks compounding at roughly 6 to 7 percent real for two centuries, through canals, railroads, electricity, automobiles, telephones, computers, the internet, and AI. The chart is real. The number is roughly right for one country across a long sample. The question worth asking is whether that single number deserves to anchor a financial plan.

Two hundred years of stock history matters, but not because it produces a clean 6-7% real constant. It matters because it shows equity returns are regime-dependent. Pre-Civil-War U.S. bonds beat stocks. Twentieth-century U.S. stocks beat bonds by a wide margin. World ex-U.S. equities have returned roughly two-thirds of the U.S. figure over the past 120 years. The same long-run average can produce wildly different lived investor experiences depending on country, era, and starting valuation.

This post is about how to use the long record, not whether to use it. The right framing treats long-run history as a base rate and a stress-test source, not a forecast engine.

What the Cleanest Long-Run Datasets Actually Show

Long-run return research draws on four datasets, each better at different things.

• Shiller’s online dataset covers monthly U.S. stock prices, dividends, earnings, and CPI back to 1871. Pre-1926 dividends and earnings are reconstructed from Cowles data. Shiller online data.
• Dimson, Marsh, and Staunton (DMS), now published annually as the UBS Global Investment Returns Yearbook, covers stocks, bonds, bills, inflation, and currencies across 35 markets, most traceable to 1900. UBS Yearbook 2026 reports $1 invested in U.S. equities in 1900 grew to $124,854 nominally by year-end 2025. In real terms, U.S. equities compounded at 6.6 percent per year, long bonds 1.6 percent, bills 0.5 percent. UBS Global Investment Returns Yearbook 2026 public summary.
• CRSP covers U.S. security-level data starting in 1926 and is the cleanest modern U.S. equity dataset. Most academic U.S. asset-pricing research uses CRSP as its base.
• Jordà, Knoll, Kuvshinov, Schularick, and Taylor (the “Rate of Return on Everything” project) extend coverage to 16 advanced economies from 1870 to 2015 and add residential housing as an asset class alongside equities, bonds, and bills. Jordà et al. 2017, FRBSF WP 2017-25.

Each dataset is a different lens. UBS notes that the first quarter of the twenty-first century has produced lower stock returns than the twentieth, with global equity investors earning 3.5 percent annualized real from 2000 through 2024. That is positive, and well below the iconic 20th-century U.S. number.

Why the 1800s Data Are Hard

Pre-1870 U.S. data are not a clean ledger. The stocks-for-the-long-run chart that most readers have seen rests on a small sample of institutions whose composition and quality have been challenged by recent academic work.

Edward McQuarrie’s 2024 paper in the Financial Analysts Journal reconstructed pre-1862 U.S. stock and bond returns from primary newspaper archives, restoring failed institutions that earlier compilations had omitted. The starkest example: the Second Bank of the United States, which McQuarrie estimates accounted for roughly 30 percent of total U.S. market capitalization in the 1830s. Shares dropped from $120 to $1.50 in the Panic of 1837 and never recovered. Older series simply left it out. McQuarrie, Stocks for the Long Run? (FAJ).

McQuarrie’s own analogy makes the scale visible. To replicate that omission today, you would have to drop Microsoft, Apple, Amazon, Alphabet, and Facebook from the S&P 500, and even those five companies would not account for as high a percentage of S&P 500 capitalization as the Second Bank did at its peak.

The data problems are not limited to one institution. Global Financial Data’s own historical work shows that pre-1870 U.S. indexes relied on limited samples, were not capitalization-weighted, omitted banks and insurance companies, and excluded over-the-counter shares including Standard Oil before it joined the NYSE. Global Financial Data on the 100-share U.S. index. The combined effect of weighting choices, missing institutions, survivorship bias, and reconstructed dividends is large enough that a confident decimal-point comparison between 1810 and 2020 returns is misleading.

Data availability is not the same as data quality. The strongest global evidence begins around 1900. The strongest U.S. public-market evidence improves materially after 1926. The 1800s are useful for expanding the regime sample, not for adding decimal points to forecasts.

The McQuarrie Correction

Once the broader sample is restored, the “stocks always beat bonds over the long run” rule loses its absolutism in U.S. history. McQuarrie’s SSRN working paper Table 6 reports century-level real geometric returns:

• 1793-1893: stocks 5.7 percent real, bonds 6.0 percent real. Bonds outperformed in the first century.
• 1893-1993: stocks 6.3 percent real, bonds 2.8 percent real. The familiar twentieth-century equity-premium era.
• 1993-2019: stocks 6.8 percent real, bonds 4.8 percent real. A narrower premium amid bond bull-market tailwinds.
• 1793-2019 full sample: stocks 6.05 percent real, bonds 4.13 percent real. The 192 basis-point premium is considerably smaller than the post-1926 number most planning tools use.

The CFA Institute’s summary of McQuarrie’s work makes the framing explicit: the equity premium is not a stationary constant. Setting the Record Straight. An investor lives through one era, not the 226-year average. The same paper’s rolling-window evidence shows that the post-1942 U.S. experience, where stocks beat bonds in 100 percent of 30- and 50-year windows, has no parallel in the pre-Civil-War sample.

Did Technology Revolutions Raise Stock Returns?

The intuitive story is yes: railroads, electricity, automobiles, and software all transformed the economy and made shareholders rich. The data make a different point.

UBS Yearbook 2026 reports that roughly 80 percent of U.S. listed-firm value in 1900 sat in industries that are now small or extinct (railroads, textiles, iron, coal, steel). Conversely, about 70 percent of today’s U.S. companies come from industries that were small or nonexistent in 1900. The composition of the market has been overhauled.

The striking result: railroads fell from 63 percent of the U.S. market in 1900 to under 1 percent today, and they outperformed both the broad market and their newer technology competitors over the full 1900-2025 sample. A “dying” industry compounded capital better than the obvious winners that displaced it.

The mechanism: in competitive capital markets, expected breakthroughs get capitalized into prices before shareholders capture the surplus. Innovation often benefits consumers (lower prices), workers (higher wages), or new entrants (competitors that eat the incumbent’s margins) more than legacy shareholders. Boring, low-expectation industries with disciplined capital allocation can produce excellent investor returns precisely because nobody pays for excitement.

Hendrik Bessembinder’s cross-section evidence reinforces the point. Do Stocks Outperform Treasury Bills? finds that 58 percent of U.S. common stocks since 1926 produced lifetime buy-and-hold returns below one-month Treasury bills, and that roughly 4 percent of companies account for the entire net gain of the U.S. stock market over the period. Long-run equity returns are not the average company doing well. They are broad diversification capturing a small number of huge winners while most of the cross-section disappoints.

The International Evidence Is Not Optional

U.S.-only history is one country, one century. It is not the only history. McQuarrie’s SSRN paper, citing the 2020 Credit Suisse Global Investment Returns Yearbook, reports that $1 invested in U.S. equities in 1900 grew to $1,937 by 2020, while $1 invested in stocks from the rest of the world grew to $179, less than a tenth of the U.S. value. The 120-year world ex-U.S. real annualized return was 4.4 percent, not the 6.6 percent estimated by Siegel for the U.S.

The CFA Institute summary of cross-country evidence pushes the point further. Of 21 markets in the DMS dataset, most experienced a negative real 20-year stock return at some point in their history, more than half experienced a negative 30-year real return, and roughly one-third had a negative 50-year real return. What Price Risk? These are not exotic edge cases; they are advanced-economy markets.

Goetzmann and Jorion (1999) made the methodological version of this point a quarter-century ago: U.S.-only long-run estimates are upward-biased because the U.S. was arguably the most successful capitalist market of the twentieth century, and survivorship applies across markets, not just within one of them. Goetzmann & Jorion: Global Stock Markets in the Twentieth Century. A planning model that anchors on U.S. 20th-century returns is anchoring on the most successful century of the most successful market.

Jordà’s “Rate of Return on Everything”

The Jordà et al. database is the most thorough long-run global record currently available. It covers 16 advanced economies from 1870 to 2015 and reports real total returns on bills, bonds, equities, and residential housing.

Their full-sample geometric returns (Table 3): bills 0.8 percent real, bonds 1.9 percent real, equities 4.6 percent real, housing 6.6 percent real. The arithmetic-mean equity number is closer to the familiar 7 percent (6.9 percent), but the geometric mean (the rate that drives terminal-wealth math) is meaningfully lower.

Two takeaways for planners. First, the equity-over-bond premium in the global geometric record is roughly 270 basis points, not the 500-plus often quoted from U.S. arithmetic data. Second, even across this best-available global dataset, the risk premium has swung over decade-long periods. Jordà et al. describe “swings in the risk premium at lower frequencies that sometimes endured for decades.” A stationarity assumption does not survive their data.

Housing is included for completeness, not as a recommendation: the 7 percent geometric figure for residential real estate over 145 years is unleveraged, untaxed, with full rent reinvested in real terms, and ignores transaction friction. Real-world owner outcomes diverge from the model. The point is that “risky capital is rewarded” is broader than “public stocks compound at 7 percent.”

Why the Right Number Is Closer to 5 Percent than 7 Percent

The popular “stocks return 7 percent real” rule of thumb anchors on U.S. 20th-century history. UBS Yearbook 2026 puts U.S. equities at 6.6 percent real, 1900-2025. Equities outside the United States returned 4.4 percent real over a comparable 120-year span (McQuarrie SSRN p. 13, citing DMS). A market-cap-weighted global composite blends those numbers: with U.S. roughly 60 percent of world market cap and ex-U.S. roughly 40 percent, the blend lands at 0.60 × 6.6 + 0.40 × 4.4 ≈ 5.7 percent real. The historical DMS World index has run closer to 5 percent because the U.S. weight in the index was lower across most of the sample.

A second piece of the gap is methodological. Jordà et al. Table 3 reports 16-economy equity returns for 1870-2015 at 6.89 percent arithmetic real but 4.64 percent geometric real. The 225-basis-point gap is the volatility drag from compounding. Terminal-wealth math uses geometric returns, not arithmetic. A reader who plugs an arithmetic average into a compounding model gets a materially wrong answer. Half of the gap between “7 percent real” and “5 percent real” comes from this single choice of statistic.

See What Real Return Should You Assume for Stocks? for the focused numerical case and a planning-implications close.

Realized Returns May Overstate Forward Expectations

A subtler caveat applies even after correctly using the global geometric mean. Twentieth-century realized stock returns were boosted by what finance researchers call rerating: a roughly one-time set of improvements in corporate governance, accounting disclosure, investability of foreign markets, broader diversification, and falling required returns as risk premia compressed. Some of the past return reflected expanding valuation multiples and shrinking discount rates, not repeatable cash-flow growth.

The Gordon-growth decomposition makes this concrete: long-run nominal return ≈ shareholder yield + nominal earnings growth + the annualized impact of any change in the price/earnings multiple. When multiples expand over a sample window, realized returns exceed what shareholder yield plus growth alone would have produced. That excess does not repeat in a new investor’s forward window unless multiples expand again from the new starting point.

Damodaran’s January 2026 implied equity-risk-premium work (data update 2 for 2026) puts the S&P 500 implied ERP at 4.23 percent over the Treasury bond rate. Combined with current real bond yields, that implies a forward U.S. equity real return materially below the backward-looking 6.6 percent. Vanguard’s 2026 outlook lands in the same neighborhood for the next 5 to 10 years. The implication for planning: forward expectations may be lower than backward realizations even at the correct global geometric mean. See Do Stock Valuations Still Matter? for the CAPE-based forward-return decomposition that captures this directly.

What This Means for Planning

Four moves follow from the evidence.

Use long-run history as a base rate, not a promise

Treating the U.S. 6.6 percent real number as the central forecast for the next 30 years is treating the most favorable century of the most favorable market as the modal future. A defensible planning baseline for a globally diversified equity portfolio sits closer to 4 to 5 percent real. The 6.6 percent figure is the upper bound of what plausibly happened in one historical regime.

Plan in real returns, not nominal

Inflation regimes have shifted across the 200-year record. UBS Yearbook 2026 finds that both equities and bonds have tended to deliver higher real returns when growth is higher and inflation lower. The reflexive claim that stocks are a clean inflation hedge across all environments does not hold up in the data; equities struggled in the 1970s alongside bonds. Real returns are the relevant unit for retirement and FI planning.

Diversify globally

The world ex-U.S. terminal-wealth gap is exactly the survivorship problem that global diversification addresses. Owning the world does not predict that the U.S. will underperform. It hedges the possibility that the U.S. century was the exception rather than the rule. The case is laid out in detail in The Case for Global Equity Diversification.

Stress-test your plan against more than one regime

A plan that only works at U.S.-20th-century returns is not a plan; it is a bet on one country’s century. Run your numbers under the world ex-U.S. regime, the early-republic regime, and a bad-regime scenario. If the plan still survives all three, it is not betting on one country’s century.

Try It: The Historical Regime Wealth Comparator

The calculator below shows the same investment compounding under five real-return regimes drawn from the academic record. The differences between regimes are larger than most retirement planning models account for. The U.S. 1900-2025 bar is the regime most plans implicitly assume; the others are equally real history.

Frequently Asked Questions

Are stock returns from the 1800s reliable enough to use today?

Reliable enough to use as a regime-expanding stress test, not as a decimal-point forecast. McQuarrie’s 2024 reconstruction restores institutions and failures earlier indexes omitted, and the resulting record shows pre-Civil-War U.S. bonds outperforming stocks at the century level. The data quality improves materially after 1870 globally and after 1926 in the U.S.

Did stock returns stay roughly constant across major technological revolutions?

At the century-aggregate level, real equity returns have stayed in a surprisingly durable range across global samples. At the investor-experience level, no. UBS Yearbook 2026 documents that railroads dominated 63 percent of the 1900 U.S. market and fell below 1 percent today, that 80 percent of 1900 listed-firm value sat in industries now small or extinct, and that the early-21st- century investor has earned 3.5 percent annualized real on global equities, well below the 20th-century U.S. norm. Technological progress and shareholder returns are correlated but not identical.

What real return should I assume for stocks over the next 30 years?

A defensible planning range for a globally diversified equity portfolio is 4 to 5 percent real, with sensitivity testing at 2 percent and 6 percent. This is below the iconic 6.6 percent U.S. figure and above the bad-regime stress case. Vanguard’s 2026 outlook projects U.S. equities at 4 to 5 percent nominal (about 2 to 3 percent real) over the next 5 to 10 years given current valuations. See Do Stock Valuations Still Matter? for the CAPE-based forward-return logic.

Why is U.S.-only history potentially misleading?

The U.S. is one country in a 21-market sample. Of those 21 markets, most experienced a negative real 20-year stock return at some point, more than half experienced a negative 30-year return, and roughly one-third experienced a negative 50-year return. Goetzmann and Jorion call this survivorship across markets: long-run results from the most successful market of the most successful century are upward-biased.

What is the McQuarrie correction to Siegel?

Jeremy Siegel’s long-run U.S. equity series rested on pre-1862 reconstructions that omitted major failures, used limited samples, and were not capitalization-weighted. McQuarrie’s 2024 work in the Financial Analysts Journal restores omitted institutions including the Second Bank of the United States, defunct banks and railroads, and broader market coverage. The resulting full-sample numbers (1793-2019, geometric, real) are 6.05 percent stocks and 4.13 percent bonds, a 192-basis-point premium rather than the larger figure post-1926 produces.

How is this different from CAPE-based forward forecasting?

CAPE-based forecasting (covered in Do Stock Valuations Still Matter?) uses today’s starting valuation to estimate forward returns over the next 10 to 20 years. The long-run history question asked here is upstream: what range of regime parameters should you even consider plausible? CAPE narrows the forecast within one regime; the regime question asks whether you should be planning around one regime in the first place.

Related Guides

• Stocks Usually Win. “Usually” Is Not a Financial Plan. covers the McQuarrie regime-toggling evidence with an Equity Premium Stress Test calculator that lets you vary the premium directly.
• Stocks Are Always Risky covers Bernstein’s four deep risks, DMS data, and Pástor-Stambaugh’s long-horizon variance result.
• Do Stock Valuations Still Matter? handles the forward-return question with a CAPE-based decomposition calculator.
• The Case for Global Equity Diversification covers the U.S.-vs-ex-U.S. valuation gap and the survivorship argument.
• Most Stocks Lose to T-Bills covers Bessembinder’s cross-section evidence on individual stocks.
• Lifecycle Asset Allocation covers age-based equity exposure decisions, which depend directly on forward-return assumptions.
• The Stock Market Rewards Optimists, But Not the Naïve Ones synthesizes Dimson’s 2026 Rational Reminder interview into seven planning lessons and adds an ERP stress tester.

Key Takeaways

• The 1800s data are useful for expanding the regime sample, not for adding decimal points to forecasts. Pre-1862 U.S. data omitted major failures including the Second Bank of the United States, and McQuarrie’s reconstruction restores them.
• Across McQuarrie’s reconstructed 226-year U.S. sample, stocks compounded at 6.05 percent real and bonds at 4.13 percent real, a 192-basis-point premium that is materially smaller than post-1926 averages.
• World ex-U.S. equities returned 4.4 percent real annualized 1900-2020, less than two-thirds of the U.S. figure. Of 21 markets in the DMS dataset, most experienced a negative real 20-year stock return at some point, and one-third experienced a negative 50-year return.
• Technological revolutions did not mechanically raise shareholder returns. Railroads dominated the 1900 U.S. market, fell below 1 percent of it today, and still outperformed the broad market and their tech competitors over the full sample.
• A planning baseline of 4 to 5 percent real for a globally diversified equity portfolio, with sensitivity at 2 and 6 percent, treats long-run history as a base rate rather than a promise. A plan that only works at U.S.-20th-century returns is a bet on one country’s century.