A century of daily prices says the dip you can see is a bear market one time in three, that waiting for it costs more than it saves, and that the bottom only arrives when the people buying the dip give up. That last part is the interesting one, because it is also a theory of how slumps end.
On Thursday 5 November the Nasdaq fell 5.5%, its worst day since the tariff panic of April 2025, and closed more than 10% below the record it had set a month earlier. The S&P 500 lost 4.4%. By mid-afternoon, according to Vanda Research, retail investors had sold a net $358 million of individual US stocks, the most on any day since 17 March 2020. One of Vanda's strategists called it “a total capitulation by retail investors.”
What made the day unusual was not the fall. It was the selling. For six years, every drop in the US market has been met by the same reflex from ordinary investors: buy the dip. It worked in 2020, it worked in 2023, and it worked spectacularly in April 2025, when the market fell 12% in four days and was back where it started within a month. In a September survey, 84% of Charles Schwab's active traders said they would buy a sharp sell-off if one came. One came, and they sold.
I had been reading Maggie Mahar's Bull!, her history of the 1982–2000 boom, and its introduction makes a claim about dips that I could not get out of my head. The 1929 crash, she writes, was not the crash. The Dow fell from 381 in September 1929 to 199 in November, then rallied 48% by April 1930. “The low of November 1929 was a false bottom, the rally of 1930 a sucker rally. In 1930 the bear trap sprang shut.” From there the market lost another 86% by July 1932. And then the line that is really an argument about economics: “It seems that a new market cannot begin until the last bull's heart has been broken, and typically, it takes more than one crash to do the job.”
If that is right, buying the dip is not just a bet that can lose. It is the thing that keeps a bear market alive. So I did what I did for the trading essay: collected the data and the studies, and checked. Daily prices for the S&P 500 back to 1927, the Dow back to 1896, the Nikkei back to 1949 and a dozen other indices; Robert Shiller's monthly series back to 1871; the papers on why people cannot sell; and the hundred-year argument between economists about whether a slump should be allowed to run its course. The short version:
Start with the simplest version of the question. The market is 10% below its high, which is the point at which the financial press starts saying “correction” and people start saying “on sale”. What happens next?
I took every day since December 1927 on which the S&P 500 first closed 10% or more below its highest close of the previous year, and followed each episode until the index made a new high. There are 46 of them. Each bar below is one.
Two in three times, the dip was more or less the dip. The median episode fell another 8% after the day it crossed −10%, bottomed within two months and regained its old high in about seven months. Someone who bought on the −10% day was ahead a year later in 70% of the episodes, by a median of 13.5%. Measured on the US index alone, the reflex has a sound basis.
One in three times, it wasn't. Seventeen of the 46 episodes went on to a fall of 20% or more, ten to a fall of 30% or more. In those, the “dip” was a quarter or a third of the way down. The mean further fall across all 46 episodes was 14%, because the bad cases are very bad: 1929, 1937, 1973, 2000 and 2007 each cost dip-buyers another 40 to 85% after the point where the market looked cheap. Once the index was 20% down, the odds got worse rather than better: 59% of those episodes went on to −30%, and the median wait to get back to the old high was almost two years.
There is no signal in the dip itself that tells you which third you are in. Carson Group's Ryan Detrick counts the same thing with a different rule, 13 of 39 corrections since 1950 becoming bears, and gets the same answer.
Verdict on “this is the dip”: unknowable in advance. Two in three stop near where you are. One in three has barely started.
The one-in-three cases are where Mahar's argument lives. Inside a long bear market the market does not fall in a straight line. It falls, rallies hard enough to convince people the worst is over, and falls again. Each rally is a dip that was bought.
I counted them. In the three great collapses of the last century, every rally of 10% or more from a low that was later undercut is marked below. The labels show the ones of 20% or more.
The Dow had eight such rallies on the way from 381 to 41, five of them larger than 20%. The Nasdaq had eight on the way from 5,049 to 1,114, including one of 45% in the autumn of 2001. The Nikkei, over the twenty years from its 1989 peak to its 2009 low, had eleven. One of them, from 2003 to 2007, took the index up 140%. It then fell to a new low. If you had bought every Nikkei dip of 10% for two decades you would have been right about the direction eleven times and lost money anyway.
This is the data behind an old Wall Street rule, Bob Farrell's eighth: bear markets have three stages, a sharp fall, a reflexive rebound, and a long grind down on the fundamentals. The rebound is the part that does the damage, because it is the part that looks like a recovery.
Ironically, these are the very responses that fuel bear market rallies, making it dangerously difficult to tell when a market has finally scraped bottom. If investors simply bowed their heads and accepted defeat, bear markets would last no more than a few months. Everyone would sell, and that would be that. But human nature, once again, intervenes. Men resist disaster. This is why even the Great Crash of 1929 did not happen in a day, a month, or a year.
Maggie Mahar, Bull! (2003), Introduction
The “responses” she means are the ones she heard from investors in 2002 and 2003: “Everything I own has gone down too much, I can't sell now,” and “The market is coming back, this is a buying opportunity.” The pattern repeated in the 1970s. The Dow fell to 631 in 1970, rallied to over 1,050 by January 1973, then collapsed. “Only then,” she writes, “did investors learn not to buy on dips. It would be another eight years before a new bull market began.”
Verdict on “the rally means the bottom is in”: false in every major bear market on record. The biggest rallies in the data all failed.
So far this is about what happens after a dip arrives. The strategy most people mean by “buy the dip” is different: keep some money aside, wait for the market to fall, and then put it in. That can be tested without any forecasting, and three groups have tested it with the same result.
AQR, the quantitative asset manager, built 196 versions of the rule on sixty years of S&P 500 data, varying how big a dip, how long it had to last, and how long to hold afterwards. Over 60% of them had worse risk-adjusted returns than simply holding the index. The average alpha was half a percent a year and not distinguishable from zero. The strategies did lose less in crashes, but only because they were sitting in cash. (AQR sells the trend-following products it compares the rule against, so read it as an interested party with checkable numbers.)
Nick Maggiulli, who writes the Of Dollars and Data blog, asked a sharper question: what if you knew exactly when every bottom would come? Even then, over 40-year windows starting between 1920 and 1979, waiting in cash for the perfect dip underperformed simply buying every month more than 70% of the time. The cash waits too long while the market rises.
I replicated his realistic version on Shiller's monthly data back to 1871, in real terms with dividends reinvested. You put aside $100 a month for twenty years. The monthly buyer invests it immediately. The dip buyer hoards it until the market is a given percentage below its high, then invests everything and keeps investing until the market makes a new high, at which point the hoarding starts again. There are 1,630 overlapping twenty-year windows.
Waiting for a 10% dip beat monthly buying in 32% of windows. Waiting for 20%: 45%. For 50%: 45%, but in 43% of windows no such dip ever came and the money sat in cash for twenty years. The median shortfall is small for shallow thresholds, 1% or less, because dips of 10 or 20% arrive often. It is 7 to 8% for the deep ones. And the means are far worse than the medians, because when waiting goes wrong it goes wrong for a decade.
Vanguard reaches the same place from another direction: given a lump sum, investing it at once beat spreading it over a year in 68% of cases on global data since 1976. Coutts, the private bank, compared buying on the first day of each year from 1990 to 2024 with waiting each year for a 10% correction: 12.1% a year against 6.6%, partly because half the years never had one.
Verdict on “buy the dip beats just buying”: false. The market usually rises while you wait, and the dips that feel like bargains are mostly cheaper than yesterday, not cheaper than tomorrow.
Everything above is about the S&P 500, and the S&P 500 is the most forgiving market in the world. “The dip always comes back” is a true statement about a capitalisation-weighted index of the largest companies in the country that won the twentieth century. It is not true in general.
The Nikkei took 34 years. The Euro Stoxx 50 and Italy's FTSE MIB took about 25, the same as the Dow after 1929. Britain's FTSE 100 and the Nasdaq took 15. Shanghai peaked in October 2007 and was 40% below that level this week; Hong Kong peaked in January 2018 and is 26% below. Someone who bought the Shanghai dip in 2008, after a 70% fall, has waited eighteen years and is still waiting.
Dividends and inflation change the American story in both directions, and the piece would be dishonest without them. With dividends reinvested and prices adjusted for the deflation of the 1930s, an investor who bought at the 1929 peak was back to even by November 1936, seven years later, not 25. But the August 2000 peak took twelve and a half years to regain on the same basis, and the 1973 peak took twelve. The honest US claim is that the total-return index has always recovered within seven to thirteen years, so far.
The sharper problem is inside the index. The reason the S&P 500 always comes back is that a few companies carry it. Hendrik Bessembinder's study of every US stock since 1926 found that 58% of them earned less than Treasury bills over their lifetimes, and that all the wealth the market created above bills, $35 trillion, came from 4% of companies. Ninety companies produced half of it. J.P. Morgan looked at the Russell 3000 since 1980 and found that 40% of stocks suffered a fall of 70% or more from which they never recovered; two-thirds underperformed the index. The index recovers. The median stock does not.
That matters because the dips people actually buy are in single names. Odean's famous study of 10,000 brokerage accounts found that the losers investors held on to went on to underperform the winners they sold by 3.4 percentage points over the following year. A stock that has fallen for the last six months tends to keep falling for the next six (the momentum effect); the reversal that value investors live on shows up over three to five years. One study by a value manager found that stocks down 60% in a year did beat the market over the next three years as a group, but 13% of them went bankrupt on the way. A deep single-stock dip is a lottery with a fat tail of zeros, not a sale.
Verdict on “it always comes back”: true for one index, with dividends, on a horizon of up to thirteen years. False for Japan, China, most of Europe for a generation, and the typical stock.
The thread running through all of this is a single, well-measured fact about people: we do not like to realise losses. Odean's accounts sold 14.8% of their paper gains each year and only 9.8% of their paper losses, a 50% greater willingness to cash in a winner than to admit a loser. Thaler and Johnson showed in experiments that after a loss, gambles that offer a chance to get back to even become especially attractive. Kahneman and Tversky measured losses as weighing roughly twice as much as gains. Holding a loser because “it will come back” and buying more of it because “it's cheaper now” are the same decision made twice.
The behaviour is not confined to stocks. David Genesove and Christopher Mayer studied Boston condominiums through the 1990s slump. Owners who would have sold at a loss set asking prices 25–35% of the loss higher than otherwise, sold at slightly higher prices, and were far less likely to sell at all. That is why housing markets “freeze” in a downturn rather than fall: volume collapses, prices stay sticky, and the adjustment is dragged out over years instead of months.
Institutions do it too. Through the 1990s, Japanese banks kept lending to borrowers who could not repay rather than recognise the losses, keeping “zombie” firms alive and starving healthy ones of capital, as Caballero, Hoshi and Kashyap documented. Sweden, hit by a comparable property and banking bust in 1991, made its banks write down their losses before the state would recapitalise them. Sweden's crisis lasted about three years. Japan's lasted a generation.
Mahar found the same reluctance in the mutual-fund data of 2002. Gail Dudack, a strategist she quotes, pointed out that redemptions had been tiny next to the sell-offs that had ended previous bear markets: “You need cash to fuel a new cycle. Until you get the sell-off that creates liquidity, a new cycle can't begin.” Morningstar's Don Phillips put it less kindly: “People have talked about how steadfast the individual investor has been. But I think it's been more paralysis than steadfastness.”
Here is the economics. If a bear market ends when sellers are exhausted, then everyone who steps in to buy before that point absorbs supply at a higher price, gets hurt, and becomes a seller later. The bottom is postponed. Dip-buying, in this view, does not cushion the fall; it lengthens it.
At the level of prices, the evidence fits. The American Association of Individual Investors has polled its members every week since 1987. The most bearish reading in the survey's history, 70.3%, came on 5 March 2009. The market bottomed on 9 March. The bears with the most dip-buying rallies were the longest and deepest; the one with none, 2020, was over in five weeks. And the price at which bear markets end has been rising as the reflex has strengthened.
At the 1932 low the market traded at 5.6 times ten-year earnings; in 1982 at 6.6; the median across twelve bottoms is 12.9. The bottoms of 2020 and 2022 came at 25 and 28. In October 2026 the ratio stood at 40.7, higher than at any point except the final months of 1999. If a cycle ends when valuations reset, the last two cycles never ended. Prices were not allowed to get cheap, which was very good for anyone who owned stocks and is a problem for anyone who is waiting to buy them cheaply.
Part of the reason is that the state now buys the dip. Anna Cieslak and Annette Vissing-Jorgensen went through the Federal Reserve's transcripts and found that since the mid-1990s a 10% fall in the stock market between meetings predicts a cut of about a third of a percentage point in interest rates at the next one, and that the committee's discussions of falling markets predict easing in a way that discussions of rising markets do not predict tightening. Japan tried it more directly: in the early 1990s the government funnelled pension and postal-savings money into shares to hold the Nikkei above 17,000. It did not hold. The index bottomed at 7,055 in 2009.
So is the conclusion that we should let it fall? This is where the argument gets a hundred years old. In the early 1930s the position that a slump does necessary work, and that cushioning it only postpones the cure, was mainstream. Herbert Hoover recorded his Treasury Secretary's formula:
“Liquidate labor, liquidate stocks, liquidate the farmers, liquidate real estate.” He insisted that, when the people get an inflation brainstorm, the only way to get it out of their blood is to let it collapse. He held that even a panic was not altogether a bad thing. He said: “It will purge the rottenness out of the system.”
Herbert Hoover on Andrew Mellon, Memoirs, vol. 3 (1952)
Lionel Robbins wrote in 1934 that “measures which postpone liquidation only make matters worse” and that “in the present depression we eschew the sharp purge. We prefer the lingering disease.” Schumpeter called depressions “forms of something which has to be done, namely, adjustment to previous economic change.” The reasoning is Mahar's and Dudack's, applied to the whole economy.
It lost. Milton Friedman and Anna Schwartz showed that the US money supply shrank by a third between 1929 and 1933 while the Federal Reserve watched, and the depth of the Depression became the standard example of what happens when you let things liquidate. At Friedman's ninetieth birthday in 2002, Ben Bernanke, then a Fed governor, ended his speech: “Regarding the Great Depression. You're right, we did it. We're very sorry. But thanks to you, we won't do it again.” Nobody who has run a central bank since has disagreed in practice. Carmen Reinhart and Kenneth Rogoff's survey of banking crises found that equity prices fall 56% over three and a half years on average after one, and house prices 35% over six, which is roughly what unrestrained liquidation looks like.
So the thesis splits. At the level of individual decisions, it is true: a loss not taken is a loss deferred, and a bear market full of hopeful buyers lasts longer than one that is sold out. Sweden against Japan is the clean experiment, and recognising the losses quickly is what shortened the cycle. At the level of policy, the conclusion the 1930s drew from the same observation, that the authorities should let everything fall, is the one thing economists across the spectrum agree was a catastrophe. What has actually happened since 1987 is the opposite of liquidation. Central banks and retail investors together have made bear markets shorter and bottoms dearer, and nobody knows what a cycle that never clears does in the end, except that on the way up it looks like 1999 and on the way down it looks like Japan.
Verdict on “buying the dip delays the bottom”: half true. True of prices and of people. The policy lesson the 1930s drew from it was wrong.
A few facts about the cycle itself, since the buy-the-dip debate tends to be conducted without them. Since 1928 the S&P 500 has had 27 bear markets, averaging a 35% fall over about nine and a half months, and 28 bull markets averaging a 112% rise over two and three-quarter years. Bears came every year and a half between 1928 and 1945, and every five years since. There have been 27 bears and 15 recessions: a bear market is not a recession forecast. The famous joke, that the market has predicted nine of the last five recessions, is roughly right.
In fourteen of fifteen recessions the market bottomed before the economy did, by a median of four months. Waiting for the economic news to improve before buying therefore means buying after the bottom. But the one exception is instructive: after the 2001 recession ended in November, the market kept falling for another eleven months. And in every case the low was only visible afterwards. Hartford Funds calculate that 42% of the market's best days in the last twenty years came during bear markets and another 36% in the first two months of a new bull, before anyone knew it was one. That is the usual argument for staying invested. It is also the precise description of the days a dip-timer spends in cash.
The history of retail dip-buying is short and mostly unhappy, with one recent, memorable exception.
In 2000, as John Griffin and his co-authors showed from Nasdaq trading records, the March reversal “was accompanied by a broad sell-off from institutional investors but accelerated buying by individuals, particularly discount brokerage clients.” Individuals bought the dip from the institutions and rode it down 78%. In 2022, retail investors were net buyers almost every week of the year; in March, Bank of America argued that this was a positive signal. By December, Vanda estimated that retail portfolios had lost $350 billion and that the average account was down about 30% against 17% for the index. Then came April 2025: the market fell 12% in four days on tariff news, retail bought roughly $5 billion the day after the announcement and kept buying, and the index was back at its old level by 2 May. That is the dip everyone remembers, because it is the one that paid within weeks.
It is also, in the data above, the normal case: two dips in three stop there. The reflex is rational as long as the market stays in the two-thirds. It was in the two-thirds in 2018, 2020, 2023 and 2025. The last time it was in the other third, in 2022, dip-buyers lost twice what the index did. And on 5 November 2026, with the index 3% off its high and valuations at 1999 levels, the people who had promised to buy the dip sold it instead. Whether that was the start of the capitulation Mahar describes, or just a bad Thursday, is unknowable from here, which is rather the point.
If the question means “should I keep buying a broad index when it falls”, the answer from a century of data is yes, in the sense that you should keep buying it when it rises too, and for the same reason. If it means “should I hold cash and wait for a fall”, the answer is no: that lost in most twenty-year windows since 1871 and in most of AQR's 196 variants. If it means “is a 10% fall a signal that the bottom is in”, the answer is that it has been one two times in three, and the third time cost everything the other two made. And if it means “should I buy more of the stock that has halved”, the answer is that the typical stock that halves does not come back, and the ones that do take years.
The thing I had not understood before reading Mahar is that these are not four separate questions. They are one question about whether we can bear to be wrong. The dip-buyer and the loser-holder and the zombie bank are all refusing the same loss, and the market cannot finish falling until enough of them accept it. That is a good reason to be the person who buys every month and never has to decide. It is also a reason to be suspicious of your own confidence on a Thursday afternoon when the index is 4% down and the whole internet is saying the same three words.