Ten Investing Numbers You Should Not Memorize

A list of “10 numbers every investor should know by heart” went around this week, ending with the promise that the ten are “worth more than most financial advice you will ever pay for.”¹ Financial social media rewards figures precise enough to sound authoritative and simple enough to screenshot. Precision and accuracy are different things, and a historical statistic can be technically correct while becoming terrible advice the moment its market, period, denominator, or assumptions go missing.

So we checked all ten against the original research, recomputed several from raw data, and graded each one twice: once for whether the number is accurate, and once for whether the advice attached to it is sound. Those two grades often diverge, which is the whole point. This is an audit of ten claims, not of the person who posted them. One of the ten is genuinely useful and deserves the memorization. Two fail arithmetic you can check on a phone. Two cite sources that, as far as we can determine, do not exist.

The scorecard

Accuracy grades the number as stated. Advice grades the instruction the post attaches to it. A claim can be true and still earn a poor advice grade, which happens twice below.

The list earns a C− as financial trivia and a D+ as investor guidance. That gap is the finding. Taken as a quiz, most of these numbers are in the neighborhood of something real, and a reader who repeated them at a dinner party would rarely be caught out. Read as instructions, which is how they are presented and how they will be used, they point a reader toward spending decisions, withdrawal rates, inflation assumptions and market-timing behavior that the underlying research does not support. A collection sold as “worth more than most financial advice you will ever pay for” should clear a higher bar than being roughly memorable.

1. “6.9% real annualized return of stocks over 200 years”

The kernel of truth. Jeremy Siegel’s long-run US series is the most cited dataset in investing, and it does show something close to this: 6.8% a year after inflation from 1802 through 2023, against 3.3% for bonds and 2.5% for bills.² Note the number is 6.8%. We could not locate a source for 6.9%.

What is missing. Three substitutions happen silently. One reconstruction of US history becomes “stocks.” A historical outcome becomes an expected return. And the winning market becomes the representative one. Edward McQuarrie rebuilt the early record from newly digitized sources going back to 1793 and found the pattern does not hold: Siegel filled 1802 to 1824 with price data plus an imputed flat 6.4% dividend yield, and with actual dividend records the 19th-century equity premium was roughly zero. As reported, stocks returned 6.68% against bonds at 6.98% in the 19th century, then 8.85% against 2.32% in the 20th.³ His conclusion is blunt: “Asset returns in the US in the 20th century do not generalize.”

Globally the figure is lower. Dimson, Marsh and Staunton put world equities at 5.2% real a year from 1900 to 2024, and global equities at just 3.5% real for 2000 through 2024.⁴ US equities returned 6.6% real since 1900, which is the survivor’s result, not the typical one.⁵ Forward-looking estimates sit lower still: Vanguard’s June 2026 model projects 4.2% to 6.2% nominal for US equities over the next decade, roughly 2% to 4% real.⁶

Why it matters. Assume 6.9% real instead of 5% and a 30-year projection grows 7.40 times instead of 4.32 times. The disputed decimal is worth a factor of 1.71 in your plan.

Defensible rewrite. World equities returned about 5% real a year since 1900 and US equities about 6.6%; use a range near 4% to 5% real for planning, and test lower. What real return should you assume for stocks? works through the planning number, and Do 200 years of stock returns still matter? covers the McQuarrie corrections.

2. “Divide 72 by your return to find years to double”

The kernel of truth. This one is sound, and it is the only item on the list we would tell you to memorize. Exact doubling time is ln(2) / ln(1 + r). At 8% that is 9.006 years against the rule’s 9. At 12% it is 6.116 against 6.

What is missing. Only the boundaries. The approximation is within 1% for rates of roughly 6% to 9% and drifts outside that: at 2% it says 36 years when the answer is 35.0, and at 20% it says 3.6 when the answer is 3.8. It also assumes a constant return, which no stock portfolio delivers; volatility drags the compound result below the average. For continuous compounding the exact constant is 69.3, and 72 wins for everyday use mostly because it divides cleanly by 2, 3, 4, 6, 8, 9 and 12.

Defensible rewrite. The Rule of 72 is an excellent mental shortcut for fixed rates near 8% and a rough guide elsewhere. It answers when a fixed rate doubles, not when a volatile portfolio will.

3. “40% of your total long-term return comes from reinvested dividends”

The kernel of truth. The underlying figures are real. The CFA Program curriculum reports that the S&P 500 compounded at 10.0% with dividends reinvested from the start of 1926 to the end of 2018, against 5.9% on price alone.⁷ Subtract and divide and you get 41%. The cleaner logarithmic decomposition gives 39.9%. So as a share of the compound annual rate, 40% holds up.

What is missing. That number is one of several defensible answers, and the post presents it as the answer. Decompose terminal wealth instead of the annual rate over the same period and dividends plus their compounding account for about 97% of it, because a total-return index grows 7,072-fold while a price-only index grows 207-fold. Hartford Funds publishes both framings in the same document: dividends contributed 33% of total return on average from 1940 to 2025, and 85% of cumulative total return from 1960 to 2025.⁸ One publisher, two attributions, differing only by method and window.

Every row describes the same history. The spread is method, not fact.

The advice is the real failure. “Spend them instead and you burn 40% of your compounding engine” implies that paying a dividend creates return. It does not. A dividend transfers value out of the company, and the share price drops by roughly the dividend on the ex-dividend date, which is why Miller and Modigliani showed dividend policy does not by itself change shareholder wealth.⁹ Spending a dividend reduces future wealth for exactly the same reason selling an equivalent dollar of shares does: money left the portfolio.

Defensible rewrite. Dividends are a large part of historical total return, and reinvesting them during accumulation preserves compounding. In retirement, dividends and share sales are both withdrawals and should be managed together for tax efficiency. See Do most S&P 500 returns really come from dividends? and Dividends are not free money.

4. “2% inflation cuts purchasing power in half over 30 years and destroys 73% over 50”

The kernel of truth. Inflation compounds against you, and that deserves the attention. The specifics are where it collapses.

What is missing. Both numbers are wrong, and you can check them on a phone. Purchasing power remaining is 1 divided by (1 + i) raised to the number of years. At 2%:

The deeper problem is that the two halves contradict each other. Halving purchasing power in exactly 30 years requires 2.34% inflation. Destroying 73% over 50 years requires 2.65%. The claim asserts both at 2%, so it is not merely inaccurate, it is internally inconsistent.

There is also a category error. 2% is the Federal Reserve’s longer-run PCE target, reaffirmed in January 2026, not a historical constant or a promise.¹⁰ US inflation has averaged 2.9% a year since 1900.⁵ At that historical rate a 50-year loss is roughly 76%, which is close to the claimed 73%. The post has approximately the right number attached to the wrong rate.

Defensible rewrite. At 2% inflation, purchasing power halves in 35 years; at the 2.9% the US has actually averaged, it halves in about 24. Plan across a range of inflation paths, and remember your household’s inflation rate depends on your own spending. See the inflation hedges guide for what to do about it.

5. “Miss the 10 best trading days per year and you lose half your returns”

The kernel of truth. A small number of days really do carry an enormous share of long-run returns, and the best ones really do cluster in bear markets. About 78% of the market’s best days occurred during a bear market or the first two months of a bull market between 1995 and 2024.¹¹ That part of the post is correct and worth internalizing.

What is missing. The words “per year.” The underlying study measures the 10 best days across an entire multi-decade sample, not annually. Javier Estrada examined 15 markets and 160,278 trading days and found that missing the best 10 days of the whole period left portfolios 50.8% less valuable, while avoiding the worst 10 left them 150.4% more valuable.¹² Missing 10 days per year over four decades would mean missing hundreds of days; missing just the best 100 turns an 8.2% annualized return into −2.8%. The claim as worded understates its own case by an order of magnitude.

The omitted mirror image matters more. Avoiding the worst days is worth about three times as much as catching the best ones, and both are equally impossible to identify in advance. The statistic is deployed to argue for staying invested, which is sound advice, but it does so by showing you only one side of a symmetric coin. Our full treatment, including a calculator that lets you remove best days, worst days, or both, is in Missing the market’s best days.

Defensible rewrite. A handful of days across decades drive a large share of returns, and they cluster inside the scary stretches when people are most tempted to sell. Since the worst days cluster there too, and neither can be identified in advance, choose an allocation you can hold through the turbulence.

6. “0.98 correlation between the S&P 500 and earnings growth over 30 years”

The kernel of truth. Over long horizons, stock prices are anchored to fundamentals. A share is a claim on future cash flows, and that is not in dispute.

What is missing. Everything needed to check it. We searched for a source and found none. The wording is also incoherent as stated, because it correlates a level, the index, with a growth rate, earnings growth, which are different units. A serious version would have to specify price or total return, aggregate earnings or per share, levels or growth rates, reported or operating earnings, nominal or real, and whether the windows overlap.

A figure that high is a statistical warning sign rather than a discovery. Two series that both trend upward produce enormous correlations mechanically, which is the classic spurious regression result of Granger and Newbold.¹³ Overlapping 30-year windows compound the illusion: adjacent windows share 29 of 30 years, so the effective sample is far smaller than the observation count suggests.¹⁴ We built a simulator for exactly this failure in The R² trap.

The evidence also cuts against the moral. Across 16 countries from 1900 to 2002, the correlation between real per-capita GDP growth and real equity returns was −0.37 with a p-value of 0.16, which means it is not statistically distinguishable from zero.¹⁵ Faster-growing economies have not delivered higher stock returns, largely because growth is financed by new firms and new share issuance that existing shareholders do not own. Long-run returns decompose into shareholder yield, per-share growth, and the change in valuation, and that last term is what makes decades diverge from fundamentals.

Defensible rewrite. Fundamentals anchor returns over long horizons, but not one for one, and the popular 0.98 figure has no source. Cross-country evidence finds no reliable link between economic growth and equity returns. See also the stock market is not the economy.

7. “4% is the safe withdrawal rate, historically lasting 30+ years”

The kernel of truth. William Bengen’s 1994 study found that withdrawing 4% of a portfolio’s starting value, then adjusting that dollar amount for inflation, survived at least 30 years in every US historical period he tested.¹⁶ The Trinity study reached compatible conclusions.¹⁷ As a first-pass screen for whether you are near retirement readiness, 4% is genuinely useful.

What is missing. First, the mechanism is usually misdescribed. The rule withdraws 4% of the initial balance and then adjusts that dollar figure for inflation each year. Withdrawing 4% of the current balance annually is a different strategy that can never run out and produces far more variable income. Second, the headline result excludes taxes and fees, assumes a specific stock-bond mix, and covers exactly 30 years. Third, it is a worst-case US rate, so the median outcome left a large unspent balance.

The international record is far less reassuring. Wade Pfau found that a 4% inflation-adjusted withdrawal would have been safe in only 4 of 17 developed countries.¹⁸ Forward-looking work also lands lower: Morningstar’s December 2025 analysis puts the starting safe rate at 3.9% for a 30-year horizon with a 90% success target, though flexible spending rules raise it to 5.7%.¹⁹ Meanwhile Bengen himself has revised upward over the years and now argues for roughly 4.7%. The honest summary is that the number is contested in both directions and depends on assumptions the tweet never states.

Defensible rewrite. 4% is a reasonable screening heuristic for a 30-year, balanced-portfolio US retirement, before taxes and fees. Your sustainable rate depends on horizon, valuations, taxes, guaranteed income, and how much spending flexibility you have. See why the 4% rule is not enough and sequence of returns risk.

8. “−14% average intra-year drawdown. Every year. Ignore it.”

The kernel of truth. This is the most accurate number on the list after the Rule of 72. J.P. Morgan’s current guide reports an average intra-year drop of 14.2% since 1980, with annual returns positive in 35 of 46 years.²⁰ We recomputed it independently from daily total-return data and got −14.3%, with 37 of 46 years positive. The small difference is instructive: J.P. Morgan uses price returns only, so excluding dividends turns two additional years negative.

What is missing. The word “every” and the instruction “ignore it.” An average is not a recurring event. Our computation puts the median at −10.5% and the range from −2.8% in 1995 to −47.7% in 2008. Twelve of 46 years were worse than −20%, and 22 were milder than −10%. No year is average.

“Ignore it” is fine advice for a young accumulator making regular contributions and bad advice for a retiree funding this year’s spending, anyone buying a house soon, anyone using leverage, or anyone who will sell at the bottom. The drawdown is harmless only if your liabilities let it be.

Defensible rewrite. Expect a double-digit decline in a typical year and occasionally far worse, and structure your allocation and cash reserves so an ordinary drawdown never forces a sale. Portfolio risk is more than volatility covers the better measures.

9. “+36.4% in the 12 months after a midterm election year bottom”

The kernel of truth. Very little. We searched for the origin of this figure and could not find one, so we will not attribute it to anybody.

What is missing. The bottom is only visible afterward. Measuring returns from the lowest closing price of a year selects the single best possible entry point using information nobody had at the time. This is not a seasonal pattern, it is an arithmetic identity: measuring forward from any year’s low produces a large positive number in any year, midterm or not. The claim also never reports the comparable figure for non-midterm years, which is the only thing that would make it evidence of a midterm effect at all.

When researchers test the pattern properly, it thins out. A 2023 study of 49 industry portfolios from 1926 to 2022 found that the higher raw returns around midterms “all but disappear after returns are adjusted for risk.”²¹ The best-known political-cycle finding, a gap between returns under Democratic and Republican presidents, was published under the title “The Presidential Puzzle” precisely because its authors could not find a risk-based explanation and declined to present it as tradable.²² There have also been only about 25 midterm elections since 1926, which is a small sample for a claim about the “most reliable” pattern in markets.

Defensible rewrite. Election-cycle statistics are sensitive to the dates chosen, and returns measured from an ex-post bottom are not investable. Do not set your allocation by the political calendar.

10. “100% of 20-year rolling periods delivered positive real returns”

The kernel of truth. US stocks have come remarkably close to never losing to inflation over 20 years. We ran it ourselves on annual S&P 500 real returns from 1928 to 2025 and found zero negative windows out of 79.

What is missing. Our own result is a good illustration of why the claim fails. That dataset starts in 1928, so it cannot contain the window that breaks the rule. On Shiller’s monthly series going back to 1871, the 20 years from June 1901 to June 1921 returned −0.22% a year after inflation. Annual sampling hides it, because the calendar-year minimum is positive. Same claim, two defensible datasets, opposite answers, and the entire difference is a start date and a sampling frequency.

Costs settle it. These are gross returns that no investor earns. Apply a 1% annual fee to our own dataset and two windows go negative, 1929 to 1948 at −0.40% and 1962 to 1981 at −0.18%. On the longer Shiller series a 1% fee produces 27 negative windows. Even leaving the negatives aside, “positive” is doing heavy lifting: the worst window in our data returned 0.63% a year for 20 years, which is technically positive and practically a lost generation.

The international record is worse. Dimson, Marsh and Staunton found “only three equity markets other than the United States (with a fourth on the borderline) that never experienced a shortfall in real returns over a 20-year period. The worst 20-year real returns of 11 countries were negative. Indeed, historically, in 6 of the 16 countries, investors would need to have waited more than 50 years to be assured of a positive return.”²³

Finally, the sample is smaller than it sounds. The same authors warn that the US evidence “is based on relatively few nonoverlapping observations and is hence subject to large sampling error.” A 155-year record contains only seven non-overlapping 20-year periods. “100% of 20-year periods” sounds like hundreds of independent trials. It is closer to seven, all drawn from the most successful market in history.

Defensible rewrite. Long horizons have made losses rare for US stocks, without eliminating them, and other countries have gone 20 years underwater in real terms. Time reduces the frequency of bad outcomes; it does not remove equity risk. See Stocks are always risky and But what about Japan?

Check the arithmetic yourself

Three of the ten claims turn on arithmetic rather than data definitions, which means you can settle them in your browser. The tabs below compute exact doubling times against the Rule of 72, purchasing power under any inflation rate, and what your own plan is worth under 3%, 5% and 7% real returns.

What we recommend instead

A better toolkit is small: a few durable relationships, plus the habit of asking what any statistic actually measured before acting on it.

• Plan with ranges. Use roughly 4% to 5% real as a central equity assumption and test 3% and 6%. A plan that only works at the optimistic end is not a plan.
• Separate description from forecast. A return that happened over 125 years is evidence about the range of outcomes, not a yield available today.
• Think in total return. Dividends, buybacks, price changes, taxes and fees all move your wealth. A dollar spent is a withdrawal whether it arrived as a dividend or a share sale.
• Match assets to liabilities. Money needed within a few years should not depend on earning the long-run average. That is what cash, T-bills and short bonds are for.
• Refuse binary timing. The best-days statistic, the election-cycle statistic and the earnings correlation all invite you to switch between stocks and cash on a backward-looking pattern.
• Treat 4% as a screen. Useful for a first estimate of retirement readiness; the real plan needs taxes, Social Security, longevity and spending flexibility.
• Diversify across countries. The three most impressive claims on the list all depend on US exceptionalism holding for another century.

How Summitward helps

Every claim above eventually becomes a question about your household rather than about history. Summitward’s projections let you run your own plan at 3%, 5% and 7% real instead of adopting a number from a screenshot, and the retirement tools replace the 4% heuristic with Monte Carlo simulation across multiple spending policies, including guardrails and variable withdrawals, so you can see what a bad early sequence would do to your specific plan. The portfolio dashboard reports drawdown, pain and ulcer measures rather than a single average decline, and factor analysis shows whether recent performance came from security selection or from exposures you could buy for a few basis points.

Methodology

Figures we computed ourselves come from two datasets already in this project: the Kenneth R. French daily value-weighted US market total return series (July 1926 to May 2026, 26,253 trading days) and Damodaran’s annual S&P 500 real total returns (1928 to 2025). Intra-year drawdowns are peak-to-trough declines in cumulative total return within each calendar year, 1980 to 2025, which is why they differ slightly from J.P. Morgan’s price-return version. Rolling 20-year windows are overlapping and annual; overlapping windows are not independent observations. Terminal-wealth comparisons assume constant returns and no taxes. The script that produces every one of these numbers lives at scripts/analyze_viral_numbers.py in the Summitward repository, so the arithmetic can be checked rather than trusted.

Frequently asked questions

Are all ten numbers wrong?

No, and that framing would be its own kind of dishonesty. The Rule of 72 is genuinely useful and accurate near 8%. The 14% average intra-year decline is real and well sourced. The dividend figure is defensible as a share of the annual compound rate. What fails is mostly the leap from a conditional historical statistic to a universal instruction, plus two cases of arithmetic that does not survive a calculator and two figures with no locatable source.

Why grade accuracy and advice separately?

Because they come apart. The drawdown statistic is accurate and the advice attached to it, “ignore it,” is wrong for retirees and anyone with near-term spending. The dividend attribution is roughly right as a number and the conclusion drawn from it, that spending dividends burns a compounding engine, misdescribes how dividends work. A single grade would hide that split, which is exactly where readers get hurt.

Should I ignore financial statistics on social media entirely?

A more useful filter is to ask four questions of any number: which market, which period, measured how, and compared with what. If a post cannot answer those, the number is decoration. The claims here that survived scrutiny all came with answers; the ones that failed either omitted them or had no source to answer from.

What real return should I actually assume?

Something in the range of 4% to 5% real for a globally diversified equity portfolio, tested against materially worse outcomes. World equities returned 5.2% real since 1900 and 3.5% real since 2000, and current forward estimates for US equities are near 2% to 4% real. More useful than picking the exact figure: check how much your plan changes when the figure moves.

Key takeaways

• Of the ten, the Rule of 72 is the keeper. The 14.2% average intra-year decline is accurate but comes with bad advice attached.
• Two claims fail arithmetic: 2% inflation leaves 55.2% of purchasing power after 30 years, not 50%, and the claim’s two halves imply two different inflation rates.
• Two claims have no source we could find, the 0.98 earnings correlation and the 36.4% post-midterm return, and the latter is measured from a bottom visible only in hindsight.
• The dividend claim is method-dependent: the same history yields 33%, 40%, 85% or 97% depending on the decomposition chosen.
• “Every 20-year period was positive” is false on Shiller’s monthly data, false after a 1% fee, and false in 11 of 16 countries. It also rests on about seven independent periods.
• Ask of every viral number: which market, which period, measured how, compared with what.

Related guides

• What Real Return Should You Assume for Stocks? the planning number behind claim 1.
• Missing the Market’s Best Days the full treatment of claim 5, with a calculator for both sides of the counterfactual.
• Safe Withdrawal Rate: Why the 4% Rule Is Not Enough claim 7 in depth, including guardrails and CAPE-based rules.
• The R² Trap why overlapping windows make claim 6 look far more certain than it is.
• Stocks Are Always Risky the international evidence that breaks claim 10.
• Do Most S&P 500 Returns Really Come From Dividends? the attribution math behind claim 3.

Sources

1. The audited list circulated on X in July 2026 and was quote-tweeted by @egr_investor. This guide audits the claims, not their author.
2. McCaffrey, P. (2025). Stocks for the Long Run? New Evidence, Old Debates, CFA Institute Research Foundation. Exhibit 1 reproduces Siegel’s 1802–2023 real returns (stocks 6.8%, bonds 3.3%, bills 2.5%).
3. McQuarrie, E. F. (2024). Stocks for the Long Run? Sometimes Yes, Sometimes No. Financial Analysts Journal 80(1), 12–28. Century-level figures as reported by CFA Institute.
4. Dimson, E., Marsh, P., & Staunton, M. (2025). UBS Global Investment Returns Yearbook 2025. World equities 5.2% real 1900–2024; global equities 3.5% real 2000–2024.
5. Dimson, E., Marsh, P., & Staunton, M. (2026). UBS Global Investment Returns Yearbook 2026 (public summary). US equities 6.6% real and US inflation 2.9% a year, 1900–2025.
6. Vanguard. Vanguard Capital Markets Model return forecasts, as of 30 June 2026: US equities 4.2–6.2% nominal annualized over ten years.
7. CFA Institute. Analysis of Dividends and Share Repurchases, CFA Program curriculum. Note this is curriculum text and gives no underlying data source for the 10.0% and 5.9% figures.
8. Hartford Funds (2026). The Power of Dividends: Past, Present and Future, using Morningstar and Ned Davis Research data.
9. Miller, M. H., & Modigliani, F. (1961). Dividend Policy, Growth, and the Valuation of Shares. The Journal of Business 34(4), 411–433.
10. Federal Open Market Committee. Statement on Longer-Run Goals and Monetary Policy Strategy, adopted January 2012, revised August 2025, reaffirmed 27 January 2026.
11. Hartford Funds. Timing the Market Is Impossible (CCWP062), citing Ned Davis Research, 1995–2024.
12. Estrada, J. (2008). Black Swans and Market Timing: How Not to Generate Alpha. The Journal of Investing 17(3), 20–34. 15 markets, 160,278 trading days.
13. Granger, C. W. J., & Newbold, P. (1974). Spurious Regressions in Econometrics. Journal of Econometrics 2(2), 111–120.
14. Boudoukh, J., Richardson, M., & Whitelaw, R. F. (2008). The Myth of Long-Horizon Predictability. Review of Financial Studies 21(4), 1577–1605; see also Hodrick, R. J. (1992), Review of Financial Studies 5(3), 357–386.
15. Ritter, J. R. (2005). Economic Growth and Equity Returns. Pacific-Basin Finance Journal 13(5), 489–503. Correlation −0.37, p = 0.16, 16 countries, 1900–2002.
16. Bengen, W. P. (1994). Determining Withdrawal Rates Using Historical Data. Journal of Financial Planning, October 1994.
17. Cooley, P. L., Hubbard, C. M., & Walz, D. T. (1998). Retirement Savings: Choosing a Withdrawal Rate That Is Sustainable. AAII Journal 10(3), 16–21.
18. Pfau, W. D. (2010). An International Perspective on Safe Withdrawal Rates from Retirement Savings. Journal of Financial Planning, December 2010. A 4% real withdrawal was safe in 4 of 17 countries.
19. Arnott, A. C., Benz, C., Kephart, J., & Guo, T. (2025). The State of Retirement Income: 2025, Morningstar. 3.9% base case; 90% success over 30 years.
20. J.P. Morgan Asset Management. Guide to the Markets, US, 3Q 2026, slide 17. Average intra-year drop 14.2%, positive in 35 of 46 years, price returns only.
21. Anderson, W., Białkowski, J., & Wagner, M. (2023). Midterm Elections and Stock Returns. Finance Research Letters 55, 103825. 49 industry portfolios, 1926–2022.
22. Santa-Clara, P., & Valkanov, R. (2003). The Presidential Puzzle: Political Cycles and the Stock Market. Journal of Finance 58(5), 1841–1872.
23. Dimson, E., Marsh, P., & Staunton, M. (2004). Irrational Optimism. Financial Analysts Journal 60(1), 15–25.
24. Shiller, R. J. Stock market data used in Irrational Exuberance, updated monthly series from 1871. shillerdata.com. Basis for the June 1901 to June 1921 window.

Editor’s note

Educational content, not investment advice. Grades are our editorial judgment applied to publicly posted claims, and they assess the claims rather than the person who posted them. Where we could not locate a primary source, we say so instead of guessing at attribution; two of the ten figures fall into that category. Figures we computed are reproducible from the script named in the methodology section. Facts verified against the cited journals, working papers, fund-company publications and central bank documents as of July 2026; forward-looking estimates and working papers may be revised.