Bandwagon Effect: Why a Crowd's Choice Becomes Your Own

Other people's choices carry information. If thousands of listeners saved a song, some of them probably liked it, and a stranger's saved count is a cheap substitute for listening. Following the crowd is therefore not always a mistake; it is a shortcut that works when the crowd knows something you do not. The bias is in what the shortcut does over time. Each person who follows the count adds to it, which makes the next person more likely to follow, and the tally stops tracking the merits of the thing and starts tracking its own history. A small early lead, which may have been luck, compounds into a large one.

Two separate pressures feed the effect. The first is informational: the belief that the crowd is probably right. The second is social: the wish to be where other people are, to have the same references, and not to be the one who backed the loser. Solomon Asch's line-judgment experiments, first reported in 1951, showed how strong the social pressure alone can be. When a group of confederates gave an obviously wrong answer about which of three lines matched a standard, participants went along with them on 36.8 percent of the critical trials in the 123-person series Asch summarized in 1955, against an error rate under 1 percent when answering alone; about a quarter never conformed, and about three quarters went along at least once.

Economists had the word before psychologists had the experiments. Harvey Leibenstein's 1950 paper on bandwagon, snob and Veblen effects described a bandwagon good as one that people demand more of because others are demanding it, which bends the ordinary demand curve. Snob effects run the other way: some people want a thing less once it is common. Both are cases of one person's choice depending on everyone else's.

What the effect produces at the level of a whole market is inequality and unpredictability. Where choices are made independently, popularity tracks quality, roughly. Where each choice is visible to the next chooser, the winners win by far more, and which items win depends on the accidents of the early tally. That is the finding of the MusicLab experiment described next, and it is why the demo on this page is run on two populations rather than one: a single visitor's add or skip says little, but the split between those who saw 12 saves and those who saw 12,418 can show the count doing its work.

Matthew Salganik, Peter Dodds and Duncan Watts published "Experimental study of inequality and unpredictability in an artificial cultural market" in Science in 2006. They built a website offering 48 songs by unknown bands, recruited 14,341 participants, mostly from a teen-interest site, and let each one listen to any songs they liked, rate them from one to five stars, and download the ones they wanted for free.

Arriving participants were assigned at random to one of two conditions. In the independent condition they saw only band and song names. In the social influence condition they also saw how many times each song had been downloaded by earlier participants in their own world, and there were eight such worlds, each starting from zero downloads and evolving separately. A second experiment made the social signal stronger by listing the songs in order of download count rather than in a random grid.

In both experiments, every one of the eight social influence worlds showed greater inequality of success, measured by the Gini coefficient of download shares, than the independent world: popular songs were more popular and unpopular songs less popular when earlier downloads were visible. Inequality was larger in the second experiment, where the signal was stronger. Success was also less predictable. The same song's share of downloads varied more between social influence worlds than between random subsets of the independent participants, and that variation also grew with the strength of the signal. Quality mattered, but only partly: taking a song's rank in the independent world as its quality, the best songs rarely did poorly and the worst rarely did well, while songs in the middle could land almost anywhere.

The design is the reason the result is trusted. Observational data on hit songs can never show what would have happened in a rerun of history; the experiment ran history eight times in parallel with identical songs, identical starting conditions and interchangeable participants, so that the differences between worlds could only come from the chain of choices inside each one. The authors noted that social influence in real markets, with marketing and media attention added, is probably far stronger than the bare download count they tested.

The effect appears wherever a running count of other people's choices is shown next to the choice. The rows below are illustrations, not findings from the studies cited on this page.

The crowd's count is information, so the goal is to use it at the right weight rather than to ignore it.

• Judge first, then look. Form an opinion of the song, the product or the argument before you see how many others liked it, as the independent-world participants did; the count can then adjust your view rather than replace it.
• Ask what the count measures. Saves, likes and downloads measure earlier exposure as much as quality, and the MusicLab result is a reminder that an early lead can be luck.
• Separate the two pressures. If you would choose differently alone, the pull is social, not informational, and Asch's participants who said so afterwards were often right about themselves.
• Look for independent judgments. Ten reviewers who could not see each other's reviews tell you more than a thousand who could.
• Notice when popularity is the argument. "Everyone is switching" is a fact about everyone, not about the thing.

The bandwagon effect belongs to a cluster of social influence effects that are often used interchangeably.

• It is not authority bias, which is deference to one source because of its status. The bandwagon runs on numbers rather than rank: a thousand strangers, not one expert.
• It is not the false consensus effect, in which people overestimate how many others share their own view. The bandwagon effect is the reverse direction of influence, from the perceived majority to the individual.
• It is not the mere exposure effect, where repeated contact with something makes it more liked. A popular song is also a heavily played one, so the two effects travel together, but exposure works without any knowledge of what other people chose.