To Be Popular You Must Already Be Popular: On the Dangers of the Bandwagon Effect

Do you ever wonder how some performer, dance, or catchphrase suddenly becomes popular? Often it is a powerful combination of random chance and social influence. This is illustrated by a brilliant experiment involving music downloads conducted by Matthew Salganik, Peter Dodds, and Duncan Watts. For their study, the researchers created an artificial music market, involving thousands of participants who were visitors to a website popular with young people. The participants were given a long list of previously unknown songs from obscure bands. They were each asked to listen to any songs that interested them and to decide which songs (if any) to download.

About half of the participants were asked to make their decisions independently, based on the names of the bands, the titles of the songs, and their own judgment about the quality of the music. The other half could see how many times each song had been downloaded by other participants. The key question was whether that information would affect people’s decision to download.

Each participant in this second group was also randomly assigned to one of eight possible “worlds,” each of which evolved on its own; those in any particular world could see only the downloads of those in their own world. you might predict that in the end, social influences would not really matter, and that quality (as measured by the choices of those in the control group) would win out. Salganik and his colleagues asked these questions: Would people be affected by the choices of others? Would different songs become popular in the different worlds? Would people be nudged by what other people did?

There is not the slightest doubt. In all eight worlds, individuals were far more likely to download songs that had been previously downloaded in significant numbers, and far less likely to download songs that had not been as popular. For that reason, initial popularity greatly mattered; it could make all the difference between success and failure. While the least popular songs in the control group never went to the top, and while the most popular songs never fell to the bottom, almost anything else could happen. The songs that did well or poorly in the control group, where people did not see other people’s judgments, could perform very differently in the “social influence worlds.” In those worlds, the ultimate success or failure of songs depended strongly on whether they attracted initial popularity. The identical song could be a hit or a failure simply because of the judgments of the people who heard it early in the process. For that reason, the success of songs was quite unpredictable and varied considerably across worlds.

What Salganik and his coauthors found was an “informational cascade,” which occurs when people receive information from the choices of others. Suppose that there is a group of eight people, deciding whom to hire for a new position in a small business. The three candidates are Adam, Barbara, and Charles. If the first speaker says that Adam is clearly best, the second might agree, not because she prefers Adam, but because she trusts the first speaker and it is not clear that he is wrong. Once the first two speakers have spoken in favor of Adam, they have created a strong nudge on his behalf, and the third speaker might simply go along. The fourth speaker, and those who follow, might agree too, at least if they do not feel strongly; they are in a cascade. The popularity of music (and movies, and books) is often a result of this sort of cascade effect. Of course, informational cascades can be accompanied by “reputational cascades,” in which people go along with others not because they have learned from them, but because they do not want to incur their wrath or disapproval.

The music downloads experiment has implications for unpredictable change in many other domains, including business and politics. Building directly on that experiment, sociologist Michael Macy of Cornell University and his collaborators asked whether the visible views of other people could suddenly make particular political positions popular among Democrats and unpopular among Republicans—or vice versa.

Here’s how the experiment worked. All participants (consisting of thousands of people) were initially asked whether they identified with Republicans or Democrats. They were then divided into ten groups: two “independence” groups and eight “influence” groups. In the independence groups, participants were asked what they thought about twenty separate isles—without receiving any information about how members of either political party stood on those issues. In the eight influence groups, participants could see whether Republicans or Democrats were more likely to agree with some political claim. The authors carefully selected issues for which it would not be obvious which side parties would favor. For example: “Companies should be taxed in the countries where they are headquartered rather than in the countries where their revenues are generated.”

The authors hypothesized that in the influence condition, it would be especially hard to predict where Republicans and Democrats would end up. If the early Republican participants in one group ended up endorsing a position, other Republicans would be more likely to endorse it as well—and Democrats would be more likely to reject it. But if the early Republicans rejected it, other Republicans would reject it as well—and Democrats would endorse it. That’s exactly what happened! Across groups, Democrats and Republicans often flipped positions, depending on what the early voters did. As the researchers put it, “Chance variation in a small number of early movers” can have major effects in tipping large populations—and in getting both Republicans and Democrats to embrace a cluster of views that actually have nothing to do with each other.

These findings help explain how members of both parties flip over short periods of time, and also how issues suddenly, and surprisingly, become polarizing across political lines. In many domains people are tempted to think, after the fact, that the success of a musician, actor, author, or politician was inevitable in light of his or her skills and characteristics. Beware of that temptation. Small interventions and even coincidences, at a key stage, can produce large variations in the outcome. Today’s hot singer is probably indistinguishable from dozens and even hundreds of equally talented performers whose names you’ve never heard. We can go further. Most of today’s political leaders are hard to distinguish from dozens or even hundreds of others whose candidacies badly fizzled. Much the same can be said for professors and for companies and products of all kinds. Social influences matter, and so does luck.

The effects of social influences may or may not be deliberately planned by particular people. For a vivid and somewhat hilarious example of how social influences can affect beliefs even if no one plans anything, consider the Seattle Windshield Pitting Epidemic. In late March 1954, a group of people in Bellingham, Washington, noticed some tiny holes, or pits, on their windshields. Local police speculated that the pits had resulted from the actions of vandals, using BBs or buckshot. Soon thereafter, a few people in cities south of Bellingham reported similar damage to their windshields. Within two weeks, the apparent work of vandals had gone even farther south, to the point where two thousand cars were reported as damaged—these evidently not the work of vandals. The threat appeared to be approaching Seattle. The Seattle newspapers duly reported the risk in mid-April, and soon thereafter, several reports of windshield pits came to the attention of local police.

Before long, those reports reached epic proportions, leading to intense speculation about what on earth, or elsewhere, could possibly be the cause. Geiger counters found no radioactivity. Some people thought that an odd atmospheric event must have been responsible; others invoked sound waves and a possible shift in the earth’s magnetic field; still others pointed to cosmic rays from the sun. By April 16 no fewer than three thousand windshields in the Seattle area were reported to have been pitted, and Seattle’s mayor promptly wrote to the governor as well as President Dwight Eisenhower: “What appeared to be a localized outbreak of vandalism in damaged alto windshields and windows in the northern part ofWashington State has now spread throughout the Puget Sound area. . . Urge appropriate federal (and state) agencies be instructed to cooperate with local authorities on emergency basis.” In response, the governor created a committee of scientists to investigate this ominous and startling phenomenon.

Their conclusion? The damage, such as it was, was probably “the result of normal driving conditions in which small objects strike the windshields of cars.” A later investigation, supporting the scientists’ conclusion, found that brand-new cars lacked pits. The eventual judgment was that the pits “had been there all along, but no one had noticed them until now.” (You might have a look at your car right now; if you’ve had it for a while, there’s probably a pit or two, or more. Et’s not the work of space aliens.)

Here’s a less dated example of social influences. In 2012, authorities in Colombia introduced a school-based immunization program to deal with HPV (sometimes called genital warts), with the effect of reaching about 90 percent of the relevant population in its initial year. So far, so good. But in 2014, several adolescent girls in one school seemed to have adverse reactions to the vaccine, and they were admitted to a local hospital. Soon thereafter, videos of adolescent girls with all sorts of symptoms—twitching, fainting, falling unconscious—appeared on social media platforms and in national newspapers. About six hundred cases were reported. The health authorities found that the HPV vaccine was not responsible and that what had happened was a mass psychogenic reaction. This finding did not alleviate public concerns. Fear spread rapidly through relevant communities, and by 2016, HPV vaccine uptake among eligible girls fell to 14 percent for the first dose and 5 percent for the complete course, where the corresponding numbers had been 98 percent and 88 percent in 2012.

The Seattle Windshield Pitting Epidemic and the Colombia psychogenic reactions are extreme examples of unintentional social nudging, but every day we are influenced by people who are not trying to influence ls. Most of ls are affected by the eating habits of our dining companions, whatever their intentions. If you find yourself nudged by your friends’ dietary choices, it is unlikely to be because your friends decided to nudge you. More likely you just thought, Oh, that looks good. Nevertheless, social influences are often used strategically. In particular, advertisers are entirely aware of the power of those influences. Frequently they emphasize that “most people prefer” their own product, or that “more and more people” are switching from another brand, which was yesterday’s news, to their own, which represents the future. They try to nudge you by telling you what most people are now doing, or what people are increasingly doing.

In many nations, candidates for public office, or political parties, do the same thing: they emphasize that “most people are turning to” their preferred candidates, hoping that the very statement can make itself true. Nothing is better than a perception that voters are shifting to a candidate in droves. In the United States, a perception of that kind helped to account for the election of Barack Obama in 2008, for that of Donald Trump in 2016, and for his defeat at the hands of Joe Biden in 2020. When voters flock to one candidate or another, they are each making what they believe to be an independent judgment on the successful candidate’s behalf. Maybe they were, but maybe not; their judgment might be strongly driven by the widespread perception that other people are flocking to that candidate.

__________________________________

Excerpted from Nudge: The Final Edition by Richard H. Thaler and Cass R. Sunstein, to be published September 14 by Yale University Press. Copyright c 2021 Richard H. Thaler and Cass R. Sunstein. Reprinted by permission of the publisher. All rights reserved.