Stocks Usually Win. "Usually" Is Not a Financial Plan.

U.S. stocks did not beat 20-year U.S. Treasuries between 1803 and 1871. They did not beat them from 1929 to 1949. They did not beat them from 1968 to 2009. Three multi-decade periods, the longest 68 years long, when the textbook “stocks for the long run” story failed to deliver. Stocks usually win. “Usually” is not a financial plan.

This guide is about how seriously to take that historical record and what to do about it. The argument is against the overconfident framing that says any 30-year window in any market will reward stocks over bonds, not against equity investing itself. Recent academic work (Edward McQuarrie, 2024) suggests the textbook “stocks for the long run” chart materially overstates the pre-1926 equity advantage. Combine that with the Arnott chart above and the conclusion is uncomfortable but accurate: realized equity premia are regime-dependent, and your specific 20- to 40-year investing window can coincide with a bad regime.

The chart

The chart above is our recreation of the analysis Robert Arnott published in 2009 in “Bonds: Why Bother?”, extended through year-end 2025. Arnott’s original chart began in December 1801 and ended February 2009 (near the post-GFC stock trough), using data from Standard & Poor’s, Ibbotson Associates, the Cowles Commission, and Schwert. We start at 1928 because that’s where Damodaran’s public dataset (and most modern reliable U.S. return data) begins. The pre-1928 portion of the older charts is the same data Edward McQuarrie’s 2024 reconstruction shows was misstated for decades, so excluding it cleanly sidesteps a debate covered in the next section. Three multi-decade spans where the chart shows bonds beating stocks:¹

• 1803-1871: 68 years. A stock investor who started in 1803 fell to roughly one-third of a bond investor’s wealth before equities finally caught up in 1871.
• 1929-1949: 20 years. The Great Depression and the inflationary 1940s combined into two decades where 20-year Treasuries beat the U.S. stock market.
• 1968-2009: 41 years. The end of Bretton Woods, two oil shocks, the dot-com bust, and the Global Financial Crisis stretched into a 41-year window where bonds beat stocks. The chart ends in February 2009.

These are not cherry-picked one-week comparisons. They span entire generations of investors. A 35-year-old who began saving in 1968 retired in 2003 having earned less from a buy-and-hold U.S. stock portfolio than from buying long Treasuries.

A note on methodology. Arnott’s 41-year (1968-2009) span used 20-year U.S. Treasury bonds and ended at the February 2009 stock trough. Our chart above uses 10-year Treasuries (more representative of typical retail bond exposure) and year-end snapshots. With those choices, Arnott’s 1968-2009 span doesn’t show up the same way: the cumulative stocks-vs-bonds ratio rose modestly from 9.06× in 1968 to 13.38× by year-end 2008. Equity premium suppressed (about 1%/yr instead of the expected 4-6%/yr), but stocks still marginally outperformed 10-year bonds. The visible multi-decade bond-winning span in our 10-year data is 1999-2017, an 18-year window where the dot-com bust, the GFC, and the slow recovery left the cumulative ratio underwater versus its 1999 high until 2017 finally re-touched it. Long bonds got a bigger boost from the 1981-2020 falling-rate regime than 10-year bonds did, which is why Arnott’s 20-year version of the chart shows the bond-winning span lasting an additional 23 years. Either dataset lands on the same lesson: multi-decade windows of equity-premium failure are a real historical pattern.

Endpoint bias matters

The Arnott chart ends in February 2009, within weeks of the post-GFC stock-market low. From that bottom, U.S. equities have done extraordinarily well. The 41-year “bonds beat stocks” window is the most dramatic of the three because it ended at a generational equity trough. Extending the analysis through 2024 with the post-2009 rally would substantially shorten that window. The right reading of the chart is therefore not “bonds always win” and not “stocks fail half the time.” Realized equity premia are regime-dependent, and your personal investing window can coincide with a bad regime.

That phrasing matters because the standard rebuttal to Arnott is “but stocks won eventually.” True for the chart in aggregate. False if your specific savings horizon happened to be 1968 through 2003. The right question for an individual investor is not “did stocks win on average over 200 years?” The right question is: can my plan survive a regime where they don’t win during the years I actually need them to?

The Siegel narrative under McQuarrie scrutiny

Jeremy Siegel’s Stocks for the Long Run is the modern canonical case for equities. Its famous chart shows U.S. stocks compounding at roughly 6.8% real per year from 1802 to 2023, versus about 3.3% real for bonds. That 3.5-percentage-point gap, compounded over decades, is the entire intuition behind “just buy and hold.”²

Edward McQuarrie’s 2024 Financial Analysts Journal paper reexamined the underlying historical data and found a problem. The pre-1926 portion of the U.S. record (the half of the 200-year chart that supposedly proves stocks always win) relied on data sources that overstated stock returns and understated bond returns. Survivorship bias, missing dividends on early bond data, and selection effects in early stock indices all pushed the Siegel-era estimates to favor equities.² McQuarrie’s reconstruction, with 3-5x more stocks and 5-10x more bonds in the pre-1926 sample than older datasets used, suggests U.S. stocks and bonds had roughly similar returns for much of the 19th century.³

CFA Institute’s research brief frames the implication precisely: there can be long periods, “some as long as an individual investor’s adult life,” when stocks and bonds deliver similar returns. The McQuarrie work doesn’t invalidate the post-1926 U.S. equity advantage, which is the part most retail investors actually rely on. It does undermine the idea that the textbook 6.8% real / 3.3% real gap is a stable property of equity markets across all of history. The pre-1926 era looks more like a coin flip than a one-sided story.

The counterevidence: Kansas City Fed, 1926-2002

The bull case for equities in the modern era is real. A 2005 Federal Reserve Bank of Kansas City study examined U.S. data from 1926 to 2002 and found that for holding periods of 26 years or longer, stocks did not underperform bonds in any starting year of the sample.⁴ This is the core empirical foundation of advice like “in your 20s and 30s, just buy stocks.”

But the same study includes two cautions that get forgotten in secondhand summaries: the historical pattern may not repeat, and many investors’ effective holding period is shorter than the textbook 30 years (because of late-career income instability, sequence-of-returns risk near retirement, withdrawals that shrink the time-weighted exposure, and the simple fact that most people don’t start saving in their 20s). So the bull case is conditional on (a) a U.S.-specific dataset and (b) a horizon you may not actually have.

Combining the three pieces (Arnott’s chart, McQuarrie’s reconstruction, and the Kansas City Fed’s caveats), the defensible claim is: historical records outside the familiar post-1926 U.S. sample show multiple multi-decade periods where stocks failed to beat bonds, and newer research undermines the idea that the stock advantage becomes certain at any practical human horizon.

What “bonds” actually means

The Arnott chart compares stocks to 20-year U.S. Treasury bonds. Not the Bloomberg Aggregate. Not a generic intermediate-term bond fund. Not T-bills. Not TIPS. Not cash. The distinction matters because long Treasuries are a specific instrument with specific behavior, and conflating them with “bonds” generally is one of the most common mistakes in consumer-facing investing writing.

• Long Treasuries deliver excellent returns when starting yields are high and rates fall. The 1981-2020s era was the perfect environment for this: 10-year yields fell from ~15% to under 1%, producing four decades of capital gains on top of high coupons. That is a major reason the 1968-2009 bond-beats-stock window is as long as it is.
• Long Treasuries are brutal when rates rise. 2022 is the recent reminder. With short rates rising sharply, long-duration Treasury indices fell roughly 30% in nominal terms (worse in real terms) over the year. That same year, U.S. stocks fell about 18%. Both halves of a stock-and-long-bond portfolio fell sharply at the same time. The diversification benefit that long bonds historically provided depends on the rate regime.
• TIPS, I Bonds, intermediate Treasuries, and short T-bills all behave differently from long Treasuries in different regimes. None of them are a free lunch. If your case for owning bonds depends on long Treasuries hedging stocks the way they did in 2008, you’re betting on a specific monetary regime.

The cleanest way to use this guide’s lessons in practice is to ask, for each bond position you hold, what specific risk it’s meant to hedge. Long Treasuries hedge growth shocks in low-inflation regimes. TIPS hedge inflation shocks. Short T-bills and savings hedge liquidity needs. A bond allocation that doesn’t map to any specific risk is just “some non-stock thing,” which is not a strategy.

The realized premium versus the expected premium

The deepest reframing this research enables is to distinguish two very different claims. The expected equity premium is the long-run compensation investors demand for bearing equity risk. There’s good reason to believe it’s positive: if it weren’t, no one would hold stocks. The realized equity premium is what an actual investor earned over an actual window. The two can diverge enormously.

The UBS Global Investment Returns Yearbook (Dimson, Marsh, and Staunton’s ongoing 21-country dataset since 1900) confirms the expected-premium story globally: equities have outperformed bonds, bills, and inflation in all 21 countries with continuous histories since 1900.⁵ That is real evidence that equity ownership is rewarded on average, across countries, over very long horizons. But the same dataset shows substantial national variation in realized premia, and many countries had multi-decade windows where the realized premium was zero or negative.

For a household, the realized premium is what funds retirement. The expected premium is reassuring but pays no bills. Plans that depend on the expected premium showing up on schedule are fragile to the gap between the two. That gap is what the calculator below quantifies.

Stress-test your plan

Set your portfolio, contributions, allocation, and an equity-premium-failure scenario. The calculator runs a deterministic year-by-year simulation under both the textbook baseline (7% real stocks, 2% real bonds) and a stress scenario where stocks underperform the baseline by your chosen amount for the chosen number of years. You see the dollar gap.

The four deep risks (and why one calculator isn’t enough)

Equity-premium failure is one form of long-horizon risk. William Bernstein’s Deep Risk framework identifies four sources of permanent purchasing-power loss that long-horizon investors actually face: inflation, deflation, confiscation, and devastation.⁶ The goal of portfolio design isn’t to find the one asset class that always wins. It’s to build a portfolio that survives different forms of failure. A compact mitigation map:

The full per-risk treatment, with hedgeability framework and cost-of-mitigation analysis, lives in Summitward’s Four Deep Risks guide. The point relevant to this guide is that the equity-premium-failure scenario the calculator above stress-tests is only one of the four. A plan that survives an equity-premium failure but is annihilated by hyperinflation or confiscation is still fragile. Diversifying across asset classes and geographies is the response to a world where you can’t predict which deep risk shows up first.

Five concrete moves the calculator implies

If your plan only works under the textbook baseline, here are five ways to make it more resilient:

1. Save more (or withdraw less). The most underrated lever. A higher savings rate or a lower planned withdrawal rate dramatically narrows the baseline-vs-stress gap because contributions and reduced spending are independent of any equity premium showing up. The 4% rule is a starting point, not a law; see the Safe Withdrawal Rate guide for modern alternatives like guardrails and CAPE-based rules.
2. Liability-match known future spending. If you know you need $80k/year in years 2030-2034, owning a TIPS ladder or a defined bond cash flow that matches those years removes equity-premium dependency for that liability. Same logic for near-term needs covered by short Treasuries.
3. Diversify globally. The Arnott chart and McQuarrie data are both U.S.-centric. Adding international equities means tomorrow’s extreme winners and tomorrow’s decent equity-premium regimes don’t have to come from one country. See The Case for Global Equity Diversification.
4. Reduce employer-stock and concentrated single-stock positions. Concentrated equity is equity-premium risk on steroids: not only does it depend on the equity premium showing up, it depends on a specific company being one of the rare big winners. See Concentration Risk and Most Stocks Lose to T-Bills for the math.
5. Don’t treat any single allocation as contractual. 60/40 is a reasonable shape; so are target-date funds; so is 100% equity for early-career savers. None of them is guaranteed to win every regime. See 60/40, Target-Date Funds, or 100% Stocks Forever? for the framework.

What this guide is and isn’t saying

It is not saying:

• Stocks are bad. Equities have been the best long-run wealth-building asset class globally for 124 years. The case for owning them is unchanged.
• You should market-time. The Arnott periods are visible in hindsight; nobody knew in 1968 that the next 41 years would favor bonds.
• Bonds always win. The chart endpoint is February 2009. Recent equity returns have been strong.

It is saying:

• The phrase “stocks for the long run” describes a historical empirical regularity in U.S. data, not a law of finance.
• Multi-decade windows where stocks lose to bonds are a real historical pattern, not theoretical edge cases.
• If your retirement plan only works when the textbook equity premium shows up on schedule, the plan is fragile.
• Diversification, across asset classes, geographies, and risk types, is the rational response to that fragility, not market timing.

Frequently asked questions

Doesn’t the post-2009 stock rally invalidate the Arnott chart?

It changes the headline numbers. Extending the chart through 2024 substantially shortens the 1968-2009 bond-winning window. But it doesn’t erase the 1929-1949 or 1803-1871 spans, and it doesn’t erase the McQuarrie revision of pre-1926 data. The historical record still contains multi-decade windows of equity underperformance against long Treasuries.

What real return assumption should I use for planning?

7% real for U.S. stocks and 2% real for bonds (the calculator’s defaults) are reasonable round-number long-run U.S. averages from the 1926-onward dataset. They are not forecasts. Many forward-looking models (AQR, Vanguard, Research Affiliates) suggest lower expected real returns going forward, especially for U.S. equities at current valuations. If you want to be conservative, run the calculator with a higher stress (say −2% to −3% per year for the next decade). The point isn’t the specific number; it’s whether your plan can survive a meaningfully lower regime.

Is the 60/40 portfolio dead?

2022 was a terrible year for 60/40 because stocks and long bonds both fell sharply. That’s a regime where neither half of the portfolio diversified the other. But over longer horizons, 60/40 has been a reasonable shape for risk-averse investors. The deeper question is what 60/40 is supposed to do: smooth volatility, protect purchasing power, or maximize wealth? Different answers lead to different allocations. See 60/40, Target-Date Funds, or 100% Stocks Forever? for the full treatment.

What about the “stocks for the long run” bull case generally?

It’s real. The post-1926 U.S. equity premium is well-documented and reflected in DMS’ global data too. This guide is not anti-equity. It’s anti-overconfidence about realized premia on your specific 20- to 40-year window. Most investors should still own significant equity exposure; this guide argues for explicit awareness that the bet may not pay off on schedule.

Is Bessembinder’s “most stocks lose to T-bills” the same point?

No, it’s a different point. Bessembinder’s research is about individual stock returns within the equity universe: most individual stocks lose to T-bills lifetime, but a small minority of extreme winners drive aggregate market wealth. That’s an argument for broad diversification within equities. This guide is about aggregate equity vs aggregate bond returns over multi-decade windows. Different question, complementary lesson. See Most Stocks Lose to T-Bills.

Related guides

• Stocks Are Always Risky covers the deeper risk catalog (Russia/China/Japan total-loss cases, the Pastor-Stambaugh long-horizon variance result, DMS global evidence).
• The Four Deep Risks of Investing unpacks Bernstein’s framework with one section per risk and the hedgeability table this guide summarized.
• 60/40, Target-Date Funds, or 100% Stocks Forever? treats the allocation choice itself: who each strategy is for and what each is solving.
• Safe Withdrawal Rate covers the spending side of the same problem: how much you can take out depends on whether the equity premium shows up, and modern guardrail rules adapt to that uncertainty.
• The Case for Global Equity Diversification extends the diversification argument across countries.
• Do You Need Managed Futures? covers one of the diversifiers that tends to help when the equity premium fails for decades, with a sleeve decomposition calculator and a candid review of trend-following funds.
• Do 200 Years of Stock Returns Still Matter? steps further upstream: the data-quality story behind McQuarrie’s correction, the Jordà global record, and a calculator that compares terminal wealth across five historical regimes.

Sources

1. Arnott, R. (2009). “Bonds: Why Bother?” Research Affiliates — Journal of Indexes. The cumulative relative-performance chart spanning December 1801 through February 2009 with named bond-winning periods (1803-1871, 1929-1949, 1968-2009) uses data from Standard & Poor’s, Ibbotson Associates, the Cowles Commission, and Schwert.
2. McCaffrey, M. (2024). Stocks for the Long Run? New Evidence, Old Debates (CFA Institute Research and Policy Center research brief summarizing Edward McQuarrie’s 2024 Financial Analysts Journal reexamination of pre-1926 U.S. stock and bond returns).
3. Kitces, M. (2024). In The Long Run, Stocks Outperform Bonds, Or Do They? Notes McQuarrie’s expanded dataset (3-5x more stocks, 5-10x more bonds than older sources) and the planning implications.
4. Federal Reserve Bank of Kansas City (2005). How Long Is a Long-Term Investment? Found that in the 1926-2002 U.S. sample, stocks did not underperform bonds for holding periods of at least 26 years; cautions this pattern may not repeat.
5. UBS Global Investment Returns Yearbook 2026 (Dimson, Marsh, Staunton). Public summary edition. 21 countries with continuous equity histories since 1900; equities outperformed bonds, bills, and inflation globally on average, with substantial national variation.
6. Bernstein, W. (2013). Deep Risk: How History Informs Portfolio Design. Investing for Adults series. The inflation/deflation/confiscation/devastation framework is also discussed in his Efficient Frontier essays.
7. Damodaran, A. Annual U.S. asset returns dataset (S&P 500, T-bills, 10-year Treasuries, Baa corporates, real estate, gold), 1928-current. Public dataset behind the calculator’s baseline 7% real / 2% real long-run averages.
8. Awealthofcommonsense.com (2020). How Often Do Long-Term Bonds Beat Stocks? Intuition piece on the Arnott result and why a positive equity premium requires the possibility of equity underperformance.