Crises & cycles

The hard moments, and who paid for them

A long-run return of 8 % a year is not earned by making 8 % every year. It is earned by living through eight drawdown episodes in thirty years — two of which halved some indices — without selling at the wrong moment.

Size of the major falls, index by index

Maximum peak-to-trough fall on closing prices — not on annual data

The redder the cell, the deeper the fall; blue marks the rare assets that rose during the episode. These figures are measured from peak to trough and are therefore markedly larger than falls computed on year-end values.

The eight episodes in detail

Each fall has its cause, its speed and its digestion time. Pick a reference index.

How long did the fall last?

Length of the decline, from peak to trough

Depth and duration are two distinct risks. The Covid crash was the fastest of the modern era — one month — and one of the most violent; the dot-com bust was half as brutal per month but stretched over two and a half years. It is the second shape that does the most damage to investor behaviour.

The great favourable stretches

The four periods that produced most of the last thirty years of performance. Each one followed — or narrowly preceded — a major crisis.

What a stock-market year actually looks like

—

What thirty years of data say, and do not say

The fall is the entry price

No equity index in this panel produced its long-run return without living through at least one 40 % fall. An investment that never dropped 30 % never returned 8 % a year either. Return is not a reward for patience: it is compensation for risk actually borne.

Concentration accelerates both ways

Semiconductors show the panel’s best annualised return and, simultaneously, its worst fall: −85 % between 2000 and 2002. A portfolio exposed to the sector must be sized for that possibility, not for its average.

The order of years matters as much as their average

Two investors with the same average return and the same contributions can end up with very different capital depending on the order in which the good years arrived. Test it yourself in the calculator’s historical simulation.

The drawdown magnitudes and recovery times on this page are rounded orders of magnitude, measured on index closing prices in local currency and mostly excluding dividends. They are meant to calibrate a risk tolerance, not to reconstruct an exact performance. See the methodology.

How to read this page

The heat map. Each row is a drawdown episode, each column an index, each cell the maximum fall suffered during that episode. The redder the cell, the deeper the fall; blue marks the rare assets that rose while the rest fell. Read across, it shows whether a crisis hit everywhere or only certain segments — the difference between 2008 and 2000 then leaps out.

How long the falls lasted. A 50 % fall repaired in eighteen months and a 50 % fall that takes seven years to erase are not the same event, though they show the same number. It is duration, far more than depth, that determines whether you hold on.

The distribution of years. A histogram of annual returns. It is the most useful view on the whole site for calibrating an expectation: "average" years are remarkably rare. Markets spend their time doing markedly better or markedly worse than their average, almost never the average itself.

Why these figures are harsher than elsewhere

The falls on this page are measured peak to trough, on closing prices — not on year-end values.

The difference is considerable, and it explains the gap with most tables found elsewhere. An index ending 2008 at −37 % touched a far lower point during the year. The annual figure describes what an accountant saw; the peak-to-trough figure describes what an investor watching their account statement lived through. It is the second one that makes people sell.

These measurements cover indices in local currency and, for most, excluding dividends. They are therefore rounded orders of magnitude, meant to calibrate a risk tolerance — not to reconstruct an exact performance. The methodology page sets out the conventions used, episode by episode.

What this page cannot tell you

It does not say when the next fall will come, nor how far it will go. Eight episodes in thirty years is too small a sample to draw a law from — and each had its own causes, which do not repeat identically.

Nor does it say what you would have felt. A table showing "−54 % over twenty-eight months" reads in three seconds; living it means opening your statement every month for more than two years and watching the line fall, without knowing whether the bottom has passed. No dataset conveys that, and yet it is what decides most sales at the worst moment.

What it does give is valuable: an order of magnitude for what you must be prepared to live through. That is the figure to carry into the volatility setting of the calculator, rather than an optimistic return. And if you want to understand why the order of the years matters as much as their average, the guide on compound interest demonstrates it arithmetically.