History is useful when questions about the data remain
Distinguish an asset’s existence from the ability to own it on the stated terms. Index returns, physical assets and private owners can differ. Maintenance, storage, taxes, access restrictions and inability to reinvest payouts change the realised result. “This asset class always won” therefore requires a more precise definition than the slogan usually provides.
What is compared over a long horizon?
Jordà and co-authors assemble annual total-return data for equities, housing, bonds and short-term debt in sixteen developed economies from 1870 to 2015. The breadth is valuable, but it also defines the limits: specified countries, asset classes and period, not every market in all history. Later version clarifications matter when comparing tables.
Total return includes relevant payouts, not only price change. Rent can be the cash flow for housing; dividends for shares; contractual payments for bonds. Comparing one price with another asset’s total return is asymmetrical. Reconstruct each measure before asking why results differed.
Two periods expose the error in an average
Take a fictional asset with +20% first and −20% second. The arithmetic average is zero, but 100 units become 120 and then 96. Total result is −4%, and geometric average annual growth is about −2.02%. Replacing compounded factors with an average percentage changes the historical result.
If consumer prices rise 5% in each period, the real terminal multiplier is 0.96 / 1.1025, about 0.87075. Purchasing power falls about 12.93%. The numbers are invented for teaching, but nominal, real and currency returns must not share one unexplained column. See [real return](/guides/inflation-real-return/).
Who disappears from a convenient history?
A table of current winners may exclude failed companies, closed exchanges and inaccessible markets. It then describes survivors, not the opportunities known in advance. Ask how bankruptcies, nationalisation, trading suspensions, currency changes and missing observations were handled. A blank is not zero return and does not justify silently deleting a period.
Old prices may be reconstructed from rare trades, appraisals or smoothed indexes. Observed volatility can therefore differ from the risk of an immediate exit. An asset is not automatically safer merely because it was repriced less often. Check whether risks are measured comparably.
Institutions change the meaning of an analogy
Debt history concerns not only interest but enforcement, collateral access, creditor rights and transferability. A modern label does not guarantee the same conditions. The same accounting ratio under different standards and industry structures can also represent different economics. State which parts of the mechanism survived and which changed.
Use a historical case to stress-test a scenario: could an investor withstand the drawdown duration, access limits and cash demands? Do not use it as a calendar for the next event. One crisis does not determine future frequency; one successful century does not prove an inevitable future risk premium.
A historical-comparison card
Record countries, period, currency, return type, reinvestment and expenses. Mark what was observed directly and what was reconstructed. Preserve the research version and explanation of missing data. If the primary source cannot support these details, limit the conclusion to a qualitative illustration. [Financial-source verification](/guides/source-verification-finance/) helps structure the card.
Debt and equity capital: not one birth date
Capital history should not be reduced to the appearance of a modern exchange. Debt requires repayment and payment terms; equity represents participation in residual business results. Transferability, pooled contributions and separate capital developed at different times and addressed different tasks.
William Goetzmann’s study of Fibonacci and financial mathematics links payment calculations and present value to the medieval Mediterranean commercial environment and discusses Italian city-state debt. This documents a stage in credit and valuation, not a claim that credit began in the thirteenth century. Interest calculations did not imply one modern market, open access or identical creditor rights.
Equity solved another problem: several participants could share a costly project and uncertain result. A joint contribution is not automatically a modern joint-stock company. Le Bris, Goetzmann and Pouget compare different historical paths toward permanent capital, transferable shares and governance of a separate enterprise, including earlier southern European forms. The narrow conclusion is that organisational features arose in different settings; no single form has universal priority across all features.
In an author-created comparison, a project needs 100 units. Debt of 100 promises 105; a weak cash result of 80 leaves a shortfall of 25. With two equal equity stakes of 50, the same 80 gives 40 per stake if there are no other claims or costs. In a strong outcome of 140, the debt promise remains 105 and equity holders share the residual under participation rules. This is not a medieval contract or modern insolvency rule; it separates fixed claims from residual risk.
Venetian debt: liquidity does not remove payment risk
Goetzmann describes compulsory Venetian loans in the late twelfth century, development of a market for prestiti claims by the mid-thirteenth century and debt consolidation in 1262. The key is the distinction between two operations: the state raises funds from an original holder, then an existing claim may pass to another holder. A secondary sale gives the owner an exit, but does not itself repay the issuer.
To compare with today, examine payment source, compulsory participation, transferability and enforcement, not merely the word “bond”. A claim may be saleable without a guaranteed price, and a state promise does not remove war, fiscal stress or rule changes. Modern Russian holder rights cannot be inferred from a Venetian loan.
After the 1970s: futures become financial instruments
The CFTC’s 1970s history records CME currency futures in 1972 and separately notes approval of a U.S. Treasury-bill contract on November 26, 1975 and a long-term Treasury-bond contract on August 2, 1977. Trading launch dates and regulatory approvals are not the same. Standardised futures risk expanded beyond agricultural delivery to currencies and interest claims.
The 1980s timeline records approval of a cash-settled Eurodollar contract on December 8, 1981 and the first index future, on Value Line, on February 16, 1982. Cash settlement can close a position by an index rule without delivering each component, but it does not remove price moves, cash requirements or hedge mismatch. A larger contract set is not proof of safety or suitability.
Growth in book capital is not the shareholder’s return
Name the measured quantity before comparing histories. In a teaching model, BVPS equals common equity divided by the relevant number of shares. With equity 100 and ten shares, BVPS is 10. A year later equity is 120 and shares remain ten, so BVPS rises 20%. That is an accounting change, not cash received. Real comparisons must align share class, perimeter, currency, accounting-policy changes and capital transactions; weighted-average EPS shares are not automatically shares outstanding at the reporting date.
If the price falls from 20 to 18 and a year-end dividend is 1 per share, owner total return before costs is (18 + 1) / 20 − 1 = −5%, despite higher BVPS. Multi-year reinvestment requires the sequence of prices and payouts; adding all dividends to the final price is insufficient. Market price, book capital and owner cash result answer different questions.
Entry cohorts and the early period in cumulative CAGR
Use a fully artificial, dividend-free and cost-free series: price 10 initially, 40 at the end of year one, 44 at year two and 48.4 at year three. The first-date owner receives a 4.84 multiple and 384% total growth; three-year CAGR is (48.4 / 10)^(1/3) − 1 ≈ 69.15%. An owner entering at the end of year one receives 1.21 over two years, 21% total growth and 10% CAGR. These are different cohorts, not contradictory estimates.
The early fourfold rise is about 87.91% of the full logarithmic increase: ln4 / ln4.84 ≈ 87.91%. A high long-term CAGR therefore does not prove the same recent speed. Show year one, the next two years and the full period, then pre-set equal-length rolling cohorts. Overlapping windows share returns and are not independent repetitions of skill.
Do not compare three-year 69.15% with two-year 10% as the same horizon or average annual growth arithmetically to reconstruct capital. With contributions and withdrawals, separate unit-portfolio performance from the result for an owner with particular cash-flow dates. Preserve BVPS, prices, payouts, reinvestment rule, cohort dates, costs and the CAGR formula. All figures here are author-created.
Sources and verification limits
[RHI1] William N. Goetzmann, “Fibonacci and the Financial Revolution”, NBER Working Paper 10352, source cr-history-venice: history of calculations and Italian debt, not current law.
[RHI2] David Le Bris, William N. Goetzmann and Sébastien Pouget, “Convergent Evolution Toward the Joint-Stock Company”, NBER Working Paper 31821, source cr-history-equity: comparative history of organisational features.
[RHI3] CFTC, “CFTC History in the 1970s”, source cr-cftc-1970s: historical dates, not a list of current contracts.
[RHI4] CFTC, “CFTC History in the 1980s”, source cr-cftc-1980s: distinguishes contract approval from the general history of the industry.