Déjà bubble: Can bubbles repeat?

Bitcoin is on a roll. It previously peaked above $60K in November 2021, then fell to under $20K in 2022, and now has quintupled to around $100K. One interpretation is that we had a bitcoin bubble in 2021 which burst in 2022, and today we have a new and bigger bubble. A competing explanation is that bitcoin’s price today is rationally responding to new information, including a more favorable regulatory environment.

Now, bitcoin’s price gyrations raise many questions, but I want to focus on just one: If you believe that there was a bubble in asset X that collapsed in 2022, is it possible to have a second bubble in asset X in 2025? Can bubbles repeat promptly?

If you conceive of markets as having the ability to learn from experience, you might think this couldn’t happen. If “Mr. Market” lost all his money on bitcoin in 2022, surely three years later he wouldn’t buy it again? Once burned, twice shy? And all those crypto bros who went bust in 2022, shouldn’t they be sadder but wiser?

According to this learning-based view, bubbles can never occur rapid fire. Therefore, whatever’s happening today in bitcoin cannot be a bubble. This view also has implications for the stock market. If you think that the U.S. stock market, like bitcoin, had a bubble that burst in 2022, then it’s impossible to have another bubble in 2025. Therefore, we can all take a vacation from the difficult task of market valuation; we’ve acquired bubble immunity.

I disagree with this view. Bubbles do sometimes repeat within the span of a few years. Historically, the Mississippi Bubble in France ended around 1720 just as the South Sea Bubble appeared in Britain. And in the past few decades we’ve often see blatant overpricing that appears, gets corrected, and then promptly re-appears. In the lab, we can observe bubbles that occur in repeated experiments run with the same participants. All these facts are consistent with theoretical models that can produce recurring waves of bubbles.

Experience does not entirely eliminate bubbles. Once burned, not totally shy. Sadder, not much wiser.

Anthropomorphizing the market is a mistake. “The market” is not one big person who can learn from experience; instead, the market is multiple individuals with diverse experiences. Those buying crypto today include new market participants, crypto fanatics whose beliefs are immutable, and those who lost money in 2022 and want to go double-or-nothing.

A better model for financial markets is infectious disease, as suggested by Shiller (1984) and Shiller (2017). Consider COVID. “America” is not one big person who got COVID in April 2020 and then acquired immunity. Instead, America consists of different individuals, and COVID occurred in discrete waves over different subpopulations over multiple years. Similarly, America is not one big person who bought bitcoin in 2021.

Even if we look at specific individuals, learning is not always a useful framework for understanding behavior. If someone goes to the casino and loses money, are we surprised to see them returning later to the same casino? Do we think they have amnesia? No. Similarly, many individuals got COVID multiple times; immunity to COVID is not a permanent condition.

The view that bubbles cannot rapidly repeat

Galbraith (1994) suggests that “financial memory” is a useful concept, and bubbles might be expected to be separated by 20 years:

For practical purposes, the financial memory should be assumed to last, at a maximum, no more than 20 years. This is normally the time it takes for the recollection of one disaster to be erased and for some variant on previous dementia to come forward to capture the financial mind. It is also the time generally required for a new generation to enter the scene, impressed, as had been its predecessors, with its own innovative genius.

Galbraith’s suggestion that bubbles should be separated by 20 years is spookily consistent with the timing of the U.S. stock market bubbles of 2000 and 2021. It’s certainly true that past experiences shape behavior, and those investors who personally lost money in prior markets might take less risk; see, for example, Greenwood and Nagel (2019). For this reason, I agree that it might be less likely to have a bubble in 2025 if you just had one in 2021. But it’s not impossible.

Cochrane (2001) asserts that:

… if there is one testable implication of crowd psychology, it surely must be that a new bubble does not start just as the last one crashes.

I can’t agree. If “crowd psychology” means social contagion models as proposed by Shiler (1984), then we do have models that can produce multiple waves. It’s true that the original mathematical model of epidemic dynamics, Kermack and McKendrick (1927), produced just one peak of infection. However, more complicated models are able to produce recurring waves. Indeed, there’s a special term for the initial occurrence: the “herald wave.” The main wave of the influenza epidemic in autumn 1918 had a smaller herald wave in spring 1918. Perhaps we should think of bitcoin in 2021 as the herald wave for our current bubble. Similarly, the price bubble model of Hirshleifer (2020) features “vigorous oscillations,” that is, multiple waves of rising and falling prices.

Cowen (2023) makes the argument that bitcoin is not a bubble because bubbles don’t typically repeat:

If in the past you have argued that crypto is a bubble, can it be the bubble is back yet again? Typically bubbles, once they burst, do not return in a few years’ time. You still will find Beanie Babies on eBay, but they are not surrounded by any degree of excitement. Similarly, the prices of Dutch tulip bulbs appear normal and well-behaved, as that bubble faded out long ago. Bitcoin, in contrast, has attracted investor interest anew time and again.

He’s right that, unlike beanie babies, bitcoin has had surprising longevity and repeated price gyrations. But is it logical to assert that because bitcoin is much crazier than beanie babies, bitcoin must be rational? That’s like asserting that because most murderers do not eat their victims, Hannibal Lecter must be innocent. In any event, as an empirical matter, we do observe recurring huge bubbles in the Chinese stock market, as I discuss later.

Examples of rapidly repeating overvaluation

First, let’s talk about violations of the Law of One Price (LOOP). It’s true that many LOOP violations are single-peaked: a mispricing arises, gets big, and then gradually shrinks. For example, in the case of the Palm/3Com mispricing described by Lamont and Thaler (2003), you could say the market made a mistake in March 2000, realized the error, and slowly corrected itself over the next six months.

However, many LOOP violations are recurring; the Law of One Price gets repeatedly broken, like a shoplifter returning to the same store again and again. Here are three examples:

• What I call “the most inherently idiotic mispricing in the history of finance” involves a closed-end fund and Cuba. Thaler (2016) shows that this fund had a discount around 10% that turned into a premium around 70%, then this premium fell to near zero, and then rose back up, all within six months.
• As shown in Peng, Xiong, and Wang (2024), the China AH premium was 200% in 2008/2009, then fell to zero in 2014, then rose back up to more than 40% in 2016.
• The largest closed-end fund premium in U.S. history, ballpark figures, had a premium over 2000% (intraday) in April 2024, it shrunk to 100% in October 2024, and today it’s back over 1000%.

In addition to LOOP violations, we also see other obviously crazy overpricing that corrects and then repeats in the space of a few months. Three more examples:

• GameStop peaked in January 2021, then mostly corrected back down in early February 2021.[1] Intelligent observers came to the entirely reasonable conclusion that the mania was over and that rationality had returned. For example, Barry Ritholtz said "markets seem to operate almost as a learning machine…They're highly adaptable and tend not to see the same mistake happen frequently."[2] It turns out that this moment was a false dawn of rationality. In the weeks following Ritholtz’s comments, GameStop went up 4X, almost reaching its prior high.
• In Korea, confusion about two unrelated men named “Ban” caused a specific stock to rise 5X in September 2016. When the company clarified the situation, the price went back down. But by December 2016, the stock was back up near its previous high.
• The viral video “Gangnam Style” caused a seemingly irrational rise in the price of a Korean semiconductor stock in October 2012. The stock price mostly corrected, but when PSY released “Gentleman” in 2013, the price soared again.

These cases are perhaps not bubbles, but they are certainly overpricing, and they demonstrate that it’s not always true that when the stock market corrects mistakes, the mistakes stay corrected.

Turning now to bubbles, the China A shares market has had two large and well-documented bubbles, one peaking in 2007 and another peaking in 2015; see Cai, He, Jiang, and Xiong (2021) and An, Lou, and Shi (2022). These bubbles had many classic features: violations of LOOP, the entry of new market participants (as measured by new accounts), and the unholy trinity of the three V’s: high valuation, high volume, and high volatility. If the Chinese stock market could have an eight-year interval between two bubble peaks, isn’t it possible that bitcoin could have a four-year interval?

In addition to real-world evidence, we can also study bubbles in experimental asset markets. In the lab, bubbles do tend to diminish when we repeat the experiment with the same subjects; “shared group experience” does mitigate bubbles. However, in some cases, even running the experiment three times is not enough to produce an efficient market. Hussam, Porter, and Smith (2008) thus conclude that “experience alone is not a sufficient condition to ensure the elimination of price bubbles … a bubble can reignite.”

Why bubbles might repeat

There are many reasons that bubbles might repeat in rapid succession. The first is the arrival of new and inexperienced investors, what I call The Fourth Horseman of the bubble. As P.T. Barnum allegedly said, there’s a sucker born every minute. Friedman (1953) considered whether markets might be destabilized by “a changing body of amateurs” who lose money by buying high and selling low.

An, Lou, and Shi (2022) show in the China A shares market, many buyers in the 2015 bubble were new arrivals who had not participated in the 2007 bubble. One driver of recurring waves of disease is the continuous entry of new individuals, such as babies, who lack immunity to existing contagions.

Another contributor to recurring bubbles is the fact that all bubbles have winners and losers. While the losers might decide to exit the game, the winners may be eager for more, having “learned” that bubbles are profitable. Between 2020 and 2022, bitcoin took a roundtrip from $20K to $60K and back again. Net profits were about zero, but some individuals made large profits and think they can again.

Even those who lost money in 2022 might want to take another bite of the apple. For them, jumping into a second bitcoin bubble would, like a second marriage, be a triumph of hope over experience. As discussed in Maurer (1940), in the world of scam artists, it’s common for certain gullible individuals  to be conned over and over, like Charlie Brown and Lucy with the football.

Another reason that bubbles might repeat is that they may be driven by external forces that also repeat. These include government policy that puts money in the hands of investors (as discussed by Greenwood, Laarits, and Wurlger (2023)) and technological innovation (what Kindleberger called “displacement”). We also see exogenous variation in epidemics, with waves of influenza every winter as the conditions promoting contagion rise.

Bubbles might be especially likely to repeat today thanks to social media, which creates a hothouse environment for ideas to spread across the investor population. If we think of bitcoin investing as a virus, this virus can evolve, mutating and spreading to susceptible minds. Just as the rise of modern air transportation allowed highly contagious diseases to spread across the globe in days, the rise of social media allows highly contagious ideas to infect financial markets.

Modern medicine has produced a vaccine for COVID. Sadly, a vaccine for investor overoptimism remains out of reach.

Endnotes

[1] References to this and other companies should not be interpreted as recommendations to buy or sell specific securities. Acadian and/or the author of this post may hold positions in one or more securities associated with these companies.