Two Contagions

In 1841 Charles Mackay told the story that every market crash since has been compared to. During the Dutch tulip trade of the 1630s, he wrote, “even chimney-sweeps and old clothes-women, dabbled in tulips,” and when prices collapsed the country’s commerce “suffered a severe shock, from which it was many years ere it recovered.” His preface gave the lesson its famous form: people “go mad in herds, while they only recover their senses slowly, and one by one” (Mackay 1852).

The historian Anne Goldgar went to the archives and found a much smaller event. She could identify only 37 people who spent more than 300 guilders on bulbs, about a master craftsman’s yearly wage. The buyers were mostly merchants and skilled artisans, connected by family, religion or neighborhood, and “sellers mainly sold to people they knew.” She never found a chain of buyers longer than five. Since the bulbs were in the ground and payment wasn’t due until May or June, most losses in the February 1637 crash were on paper. No one drowned in a canal, she found no bankrupt who could be traced to tulips, and the Dutch economy was left unaffected (Goldgar 2018; Goldgar 2007).

So two things spread in 1637, and they spread differently. The trade moved through people who knew each other, stayed small, and broke when trust between them broke. The story of the trade moved by broadcast and never stopped. That split is the most useful thing the old manias can teach anyone reading a market narrative now: the story of a mania spreads like news, the mania itself spreads like a habit, and loudness measures the story.

How the legend traveled

Goldgar traced the route. After the crash, the Dutch tradition of satirical songs took over, and pamphlets mocking the traders were sold. Later seventeenth-century writers picked them up, then a German author of a popular history of inventions, translated into English, and then Mackay, much of whose tulip chapter comes from that book. From Mackay the story passed into finance textbooks and, eventually, every bubble headline.

The economists who looked at prices read the episode more calmly still. Peter Garber argued that very high prices for rare new bulbs, falling fast as they multiplied, were normal in the bulb trade, and that only the last weeks of common-bulb trading look like a bubble at all (Garber 1989).

The legend’s best-known author missed the mania in front of him. Andrew Odlyzko showed that Mackay, as editor of the Glasgow Argus during the Railway Mania, wrote in 1845 that there was “no reason whatever to fear for any legitimate railway speculation” (Odlyzko 2012). Mackay’s chapter on tulips was a story, which is easy to repeat. Telling which railway would pay was a judgment, and he got it wrong.

Strasbourg, 1518

The dancing plague is the other favorite example of mass madness, and the historian John Waller’s account has the same two layers. Chronicles indicate that about 400 men, women and children danced for days, with dozens of deaths. The outbreak came after a decade of famine, sickness and severe cold. Many believed that Saint Vitus could curse sinners with compulsive dancing, so each new dancer made the curse more believable to the next. The city authorities made it worse by gathering the dancers in some of the most public spaces in town (Waller 2009).

The behavior spread inside a frightened community, through belief that others visibly confirmed. The story spread by chronicle, and the figures we have come from those chronicles. Waller also explains why the popular explanation, poisoning by ergot (a grain fungus), fits poorly: the dancers behaved according to the conventions of the Saint Vitus legend, which points to belief rather than a drug.

Two kinds of contagion

Sociology has a name for each half. Mark Granovetter showed in 1978 that whether a crowd joins something depends on how individual thresholds are spread out, how many others each person needs to see doing it first, so two groups with similar average views can end up in very different places (Granovetter 1978). Damon Centola and Michael Macy then separated simple contagions, which one exposure can pass on, from complex ones, which people adopt only after confirmation from several contacts. Risky, costly or unproven behaviors are complex, and for those the long-distance ties that speed up gossip barely help (Centola and Macy 2007).

Centola tested it with 1,528 people in an online health community, built so the only thing that differed was how the network was wired. A health behavior reached 53.77% of people in tightly clustered networks and 38.26% in random ones, and spread more than four times faster through the clustered ones (Centola 2010). Reinforcement from overlapping contacts beat reach. A later experiment found a tipping point for changing a group’s convention: the largest committed minority that failed was 21% of the group, the smallest that succeeded 25% (Centola et al. 2018).

A story about a mania is a simple contagion. One telling is enough to pass it on, and repeating a cautionary tale costs nothing; it even makes the teller look prudent. Buying bulbs on credit, or joining a dance people believed was a curse, needed confirmation from people close by.

Loud is not deep

Big numbers don’t tell you which kind you’re looking at. Sharad Goel and colleagues measured the shape of about a billion sharing events on Twitter and found that the correlation between how popular something became and how far it traveled person to person was generally low. Two cascades of the same size could be a single broadcast or a long chain of people passing it on (Goel et al. 2016).

Same size, different shape

Broadcast One source reaches everyone directly.

Person to person Each person passes it to one or two others.

Falsehood doesn’t change this. Soroush Vosoughi, Deb Roy and Sinan Aral found that false stories spread farther, faster, deeper and more broadly than true ones (Vosoughi et al. 2018). When Jonas Juul and Johan Ugander compared cascades of the same size, most of the difference in shape disappeared: false news was more infectious, so it got bigger, but it spread by the same mechanism (Juul and Ugander 2021).

Reading 2021 by structure

The NFT boom of 2021 read like a mania in the headlines, and the trading data shows the two layers again. Matthieu Nadini and colleagues mapped 6.1 million trades made between 2017 and April 2021 and found that the top 10% of traders performed 85% of all transactions (Nadini et al. 2021). The attention was broadcast to everyone. The trading was done by a small, connected group. Volume measured the story; the structure showed who was actually in it.

At the desk

When the desk reads the narrative around a project, it separates reach from adoption. Is the attention coming from one broadcast, a launch post, a listing or a famous account, or from many people passing it on? Are the people who act on it, using the product, building on it or holding it, connected to each other, so that each new one is confirmed by several others? Does the activity hold when the broadcast stops? These are questions about structure, and they say nothing about price. The comparison to a past mania (“this is just tulips again”) deserves the same check, because it is the most broadcast-friendly story in finance and, as Goldgar showed, often a misreading in its own right. The desk’s notes on misreadings in a fork and how claims lose credibility take the same view from other angles.

How this could be wrong

Similar people choose similar things, which can look like contagion. Sinan Aral and colleagues found that more than half of what looked like influence in one large network was explained by people’s similarity (Aral et al. 2009). Clusters of adopters may sometimes be clusters of like-minded people, with no spreading at all. The tipping-point experiment used small online groups, so its 25% is not a universal constant. And the historical counts are what survived in archives and chronicles; Goldgar’s numbers are a floor, and the Strasbourg figures come from accounts written by people who had their own reasons to tell the story.

References

1. Mackay, Charles. Memoirs of Extraordinary Popular Delusions. Richard Bentley, 1841. Quoted from the 2nd edition, 1852.
2. Goldgar, Anne. Tulipmania: Money, Honor, and Knowledge in the Dutch Golden Age. University of Chicago Press, 2007.
3. Goldgar, Anne. "Tulip Mania: The Classic Story of a Dutch Financial Bubble Is Mostly Wrong." The Conversation, February 12, 2018.
4. Garber, Peter M. "Tulipmania." Journal of Political Economy 97, no. 3 (1989): 535–560. · doi:10.1086/261615
5. Odlyzko, Andrew. "Charles Mackay's Own Extraordinary Popular Delusions and the Railway Mania." Working paper, revised 2012. · doi:10.2139/ssrn.1927396
6. Waller, John. "Dancing Plagues and Mass Hysteria." The Psychologist 22, no. 7 (2009).
7. Granovetter, Mark. "Threshold Models of Collective Behavior." American Journal of Sociology 83, no. 6 (1978): 1420–1443. · doi:10.1086/226707
8. Centola, Damon, and Michael Macy. "Complex Contagions and the Weakness of Long Ties." American Journal of Sociology 113, no. 3 (2007): 702–734. · doi:10.1086/521848
9. Centola, Damon. "The Spread of Behavior in an Online Social Network Experiment." Science 329, no. 5996 (2010): 1194–1197. · doi:10.1126/science.1185231
10. Centola, Damon, et al. "Experimental Evidence for Tipping Points in Social Convention." Science 360, no. 6393 (2018): 1116–1119. · doi:10.1126/science.aas8827
11. Goel, Sharad, et al. "The Structural Virality of Online Diffusion." Management Science 62, no. 1 (2016): 180–196. · doi:10.1287/mnsc.2015.2158
12. Vosoughi, Soroush, Deb Roy, and Sinan Aral. "The Spread of True and False News Online." Science 359, no. 6380 (2018): 1146–1151. · doi:10.1126/science.aap9559
13. Juul, Jonas L., and Johan Ugander. "Comparing Information Diffusion Mechanisms by Matching on Cascade Size." PNAS 118, no. 46 (2021): e2100786118. · doi:10.1073/pnas.2100786118
14. Nadini, Matthieu, et al. "Mapping the NFT Revolution: Market Trends, Trade Networks, and Visual Features." Scientific Reports 11 (2021): 20902. · doi:10.1038/s41598-021-00053-8
15. Aral, Sinan, Lev Muchnik, and Arun Sundararajan. "Distinguishing Influence-Based Contagion from Homophily-Driven Diffusion in Dynamic Networks." PNAS 106, no. 51 (2009): 21544–21549. · doi:10.1073/pnas.0908800106