How long did each crash take to come back? It depends what you mean by “back.”

Take the crash everyone knows. The S&P composite closed at its 1929 peak on 7 September 1929. It did not close above that level again until 22 September 1954 — 300.5 months, a quarter of a century. That is the number quoted to frighten people, and it is correct.

Now add the dividends. Companies kept paying them through the Depression, at lower levels, and an investor who reinvested them was made whole in nominal terms by January 1945 — 184 months. Then add deflation: prices of everything fell by a quarter in the early 1930s, so a dollar bought more. Measured in purchasing power, with dividends reinvested, the same investor was back to even by November 1936 — 86 months, about 7.2 years.

Twenty-five years, fifteen years, seven years. One crash, three honest answers, and the thread arguing about it never says which one it means.

This story walks through every episode in the Markets Panic research and shows all three clocks side by side where a free source exists. It does not tell you which clock is yours. That depends on whether you reinvested, whether you were adding or withdrawing money, what you paid in fees and tax, and what your own prices were doing. But it will stop you from comparing someone's price-only thirteen years with someone else's dividends-reinvested two.

Dividends shorten the wait. Inflation can lengthen it more than dividends shorten it.

The 1929 case is the one where the three clocks disagree most, but it is not the only shape. In 1973 the order reverses. The oil-shock bear market regained its price peak in 90.2 months, and with dividends in only 42. But the 1970s were inflationary, not deflationary. After adjusting for prices, the investor who held through 1973 was not back to even until January 1985 — 144 months, twelve years. The index looked recovered on the screen for most of that time. The groceries said otherwise.

The 2000 peak is the other long one, and it hides a trap in the table. The Nasdaq Composite, the index that defines the dot-com episode here, has no free total-return series, so its dividend and inflation columns show the S&P 500 over its own 2000 peak instead — clearly labelled, not a substitute for the Nasdaq. On that basis the S&P regained its 2000 level with dividends in 73.7 months, by 23 October 2006. Inflation-adjusted, it did not: the 2008 crisis arrived before purchasing power was restored, the real total-return series bottomed in March 2009, and the real recovery came only in May 2013 — 153 months. That is the longest wait to real break-even from any month in the 153 years of Shiller's data, and it began at what many people remember as an ordinary, if expensive, time to buy.

The 2007 crisis is the tidy case. Price: 65.6 months to 28 March 2013. With dividends, daily: 53.8 months to 2 April 2012, a year sooner. Inflation-adjusted: 65 months, back to roughly the price clock, because the dividend gain and the inflation loss nearly cancelled. The 2020 shock was so fast that the three clocks barely separate (5.9, 5.7 and 7 months). The 2022 bear market shows the modern pattern: dividends cut the wait from 24.5 to 23.3 months, and the inflation of that year pushed the real recovery out to 25.

Why the forum answers contradict each other

Read the replies to “how long does it take to recover?” with the three clocks in mind and the contradictions resolve. “One to two years” is usually the 2020 shock, or the 1987 crash, on any clock — both were back inside two years. “Four to five years” is 2007 with dividends. “Thirteen years” is the S&P 500 from its 2000 peak measured in purchasing power with dividends — 153 months here — or its price index if you dismiss the brief 2007 recovery that the next crisis erased. “Fifteen years” is the Nasdaq's 181.4-month price wait. “Twenty-five years” is 1929 without dividends. Nobody in those threads is lying. They are remembering one row of this table each, and none of them says which column.

There is a fourth clock the table cannot show: your own. An index regaining its old level is not an investor breaking even. Someone who kept contributing through the fall bought at lower prices and recovered sooner than the index. Someone who was withdrawing — a retiree, say — sold at lower prices and may never recover on the index's schedule. Fees, taxes and, outside the United States, currency movements all shift the date. The three clocks here are properties of one benchmark; the fourth is a property of one life.

How long is long? Every start month since 1871

The episodes above are famous peaks. A fairer question is what happened to an investor who bought in any month, not just the worst one. Shiller's monthly series allows exactly that: for every month from January 1871 to June 2024, we ask how the S&P composite with dividends reinvested had done after one, two, three, five, ten, fifteen and twenty years.

Two things stand out. First, the worst outcomes are all the same event: every nominal worst case is a start month in 1929–31. Second, the answer to “how long is long?” is different in nominal and real terms. With dividends and no inflation adjustment, no twenty-year start month has ever lost money in this series, and the worst fifteen-year outcome was a loss of 5.2%. Adjusted for inflation, 11% of ten-year starts lost purchasing power, the worst ten-year start was March 1999 at -45.7%, and even at twenty years one start (June 1901) finished -4.3% below where it began. The longest wait from any month until the inflation-adjusted index with dividends regained that month's level was 153 months, from August 2000 to May 2013.

So when a reply says “the market always recovers,” it is describing the nominal, dividends-reinvested, twenty-year column of one country's index. That column has indeed never lost. It is also the most generous column on the page, and the one furthest from how most people experience a crash, which is in the first year, in nominal price, on a screen.

Use the clock that matches your decision

If you are deciding whether to sell before a feared crash, the relevant clock is the one that governs your re-entry rule, and that is usually price — the level you will be watching on a screen. If you are deciding whether you can afford to wait, the relevant clock is inflation-adjusted with dividends, because that is what your money will buy when you need it. If you are arguing on a forum, the relevant clock is whichever one the other person is using, and it is worth asking.

The calculator lets you put a portfolio on any of the price paths in the table and choose a re-entry rule before seeing the outcome. It works in price terms, like the screen you would be watching. Three of the long episodes are linked below.

Send a private correction or source concern.