Methodology
How these figures are made
A data site is worth what its conventions are worth. Here is exactly what is measured, how, and with what limits — including the inconvenient ones.
Scope of the series
Seven series of annual returns (calendar year), ending 31 December 2025. These are indices, not ETFs: they bear no management fees, no trading costs and no tax.
| Series | Period | Currency | Return type | Indicative yield |
|---|
Three reading traps, and what is done about them here
Price return ≠ total return. Four series in the panel (Nasdaq-100, semiconductors, STOXX Europe 600, gold) are tracked excluding dividends, the others with dividends reinvested. Comparing the two directly understates the former by roughly their annual dividend yield — from 0.7 % for the Nasdaq-100 to 3.2 % for Europe. Over thirty years, 3 points a year is a factor of 2.4 on the final capital. That is why every row of the comparison table carries a price or total badge, and why the site offers no automatic “best ETF” ranking: the ranking would depend on a convention, not on reality.
Currency changes the result, sometimes a lot. Series are expressed in their calculation currency, mostly the dollar. A euro-based investor also lived through the euro/dollar move, which can add or remove several points a year: the same year of a world index can be markedly better in euros than in dollars, or the reverse, depending on which way the currency went. Over very long periods the effect fades but never disappears. So do not read the returns shown as your performance in euros.
A fall computed on annual data is always too gentle. The series are annual: the “maximum fall” computed on them only sees year-end points and misses everything in between. The S&P 500 shows −37 % for 2008, but fell 57 % between its October 2007 peak and its March 2009 trough. The Crises & cycles page gives the real peak-to-trough magnitudes, documented separately; the drawdown charts on the Indices & ETFs page stay on the annual basis and say so.
Formulas used
| Measure | Calculation |
|---|---|
| Annualised return (CAGR) | The n-th root of the product of (1 + rᵢ), minus 1. This is the geometric mean: it accounts for compounding, unlike the arithmetic mean, which systematically overstates the real outcome. |
| Volatility | Sample standard deviation of annual returns (denominator n − 1). A volatility computed on annual data is lower than one computed on monthly or daily data. |
| Maximum fall | The largest gap between a year-end value and the highest year-end value reached before it. See the third warning above. |
| Return / risk | (CAGR − 2 %) ÷ volatility. A simplified Sharpe ratio, with the risk-free rate fixed at a flat 2 % and an annual volatility. Useful for ordering, not for publishing. |
| Growth of €10,000 | Successive compounding of the annual returns, with no further contribution, no fees and no tax. |
The calculator engine
- Monthly steps. Capital compounds every month at the equivalent of the annual net rate, and contributions are added at the end of each period (month, quarter or year). That is the more conservative assumption: paying in at the start of the period would produce a slightly better result.
- Fees. Net rate = (1 + gross return) × (1 − fees) − 1. Fees apply to the holdings, so their cost grows with the capital and not with your contributions.
- Indexation. The contribution is multiplied by (1 + indexation rate) at the start of each new year.
- Inflation. The final capital is divided by (1 + inflation)years. It does not enter the compounding: it is a conversion into today’s purchasing power, applied afterwards.
- Tax. Charged once at exit, on the gains only. Deliberately simple: no annual social charges, no allowances, no wrapper-specific treatment. Country presets set the rate; some countries instead tax holdings every year, which the calculator models as a separate annual levy.
- Scenario range. The standard deviation of the annualised return over n years is σ ÷ √n. The bounds shown are ± 1 standard deviation around the central return, roughly two cases out of three — under the debatable assumption that returns are independent from one year to the next.
- Historical simulation. Your contributions are replayed on the actual sequence of annual returns of the chosen index, starting from the selected year. The “annualised return on your contributions” is the internal rate of return of the monthly cash flows, solved by bisection — it differs from the index return because your contributions were not all exposed for the same length of time.
Known limitations
- Annual granularity. No intra-year data: a true maximum drawdown, a rolling correlation or a monthly volatility cannot be measured on the long-run layer.
- Survivorship bias in the choice of indices. The panel keeps well-known indices because they came through the last thirty years well. The performance of markets that disappeared, stagnated or closed to foreign investors is not here.
- A sector is not a portfolio. Semiconductors show the panel’s best annualised return over thirty years. That is true in hindsight; nothing says it will hold for the next thirty, and sector concentration carries a risk cost the average return does not show.
- Rounded crisis figures. The peak-to-trough falls on the Crises & cycles page are documented orders of magnitude, rounded to the percentage point, in local currency and mostly excluding dividends.
- No real-time data. The site is entirely static: nothing is loaded from a market server by the page itself. Market prices are refreshed by a script you run.
Sources
Annual returns come from index providers and public compilations cross-checked against each other. Where a value differed between sources, the index provider’s value was kept.
- MSCI World and MSCI Emerging Markets — MSCI, net returns in dollars; recent years cross-checked against Investing in the Web.
- S&P 500 — S&P Dow Jones Indices, total return; cross-checked against Macrotrends.
- Nasdaq-100 — Nasdaq; annual series cross-checked against Slickcharts.
- PHLX Semiconductor (SOX) — Nasdaq OMX; 1995-2025 annual table from History of Market.
- STOXX Europe 600 — STOXX; 1999-2024 annual series from 1Stock1, 2025 value derived from the tracker fund’s return less the dividend yield.
- Gold (spot ounce) — World Gold Council; 2000-2024 annual table from The Gold Price, 2025 value cross-checked in the financial press.
- Crisis magnitudes and recovery times — closing prices of the indices concerned, cross-checked in market literature.
Series last verified: August 2026. A value looks wrong to you? It can be changed on one line in
assets/js/data.js — the rest of the site adapts on its own.Cumulo is an educational tool. It is neither investment advice, nor a recommendation, nor a solicitation. Past performance does not predict future performance and an equity investment can lose capital. The characteristics of the funds mentioned (fees, assets under management, tax eligibility, replication method) must be checked in the issuer’s key information document.