The bandwagon effect

Grain of Salt teaches the test rather than the conclusion. The bandwagon effect is one of the few biases where the crowd is sometimes right for good reasons, and this page spends as much space on those cases as on the failures.

What the bandwagon effect is

The bandwagon effect describes adoption that feeds on adoption. Each new participant makes the option more attractive to the next, and the attraction operates whether or not the option is any good, because the signal being copied is the count rather than the quality. The bias and the fallacy are two views of the same thing: the effect is what happens in the behavior, and the appeal to popularity is what happens when someone writes the behavior down as a premise.

The psychologist Solomon Asch demonstrated the strength of the pull in the 1950s with a task that had an obviously correct answer. Participants judged which of three lines matched a reference line, after a group of other people, working for the experimenter, had given the same wrong answer aloud. A substantial share of participants went with the group against the evidence of their own eyes on at least one trial. The lines were unambiguous. The pressure was social, and it was enough.

The bias in one sentence, and the two parts it always has

Every bandwagon argument has two parts: an adoption count that is usually accurate, and a conclusion about quality or truth that the count does not license. The count is rarely the lie. Best selling often means exactly what it says.

A hardware chain advertises the country's best selling extension ladder. That is a fact about sales. The conclusion a shopper draws, that it is the ladder least likely to fold under them, is a fact about load ratings, and nothing in the sales figure touches it. The ladder may be the best selling one because it is the cheapest, because it fits in a car, or because it is the one stocked at eye level in nine hundred stores.

The four forms the effect takes

The effect takes 4 forms, and only the last is a genuine argument rather than an impression.

• The visible queue. People join what is already busy, from a line outside a restaurant to a crowded lane at a checkout, using the crowd as a proxy for quality.
• The ratings cascade. Early reviews shape later ones, so a product that opens well accumulates further praise from people who read the praise first.
• The category default. One option becomes the thing everybody uses, and choosing it stops requiring a reason while choosing anything else starts requiring one.
• The stated appeal. Popularity is offered explicitly as evidence: eight million customers, the market leader for a decade, the choice of most professionals.

Form two has a technical name worth knowing: an informational cascade. Once people can see each other's choices, each person rationally weighs the crowd alongside their own information, and after a few rounds the crowd is copying the crowd. Nothing new is being added, and the count keeps rising anyway.

A worked example, and the repair

A worked example shows what the repair costs and what it buys. Here is the raw version, from a school committee: "9 of the 12 schools in the district have moved to this timetable, so we should move to it too." The count is checkable and probably right. What the sentence does not say is why the other nine moved, and the answer may be that the first two moved and the remaining seven copied.

The repair replaces the count with the reasons behind it: "9 of the 12 schools moved to this timetable. 2 of them evaluated it against attendance and lesson time and published what they found, and 7 adopted it because neighboring schools had. We should read the 2 evaluations and ignore the 7." That version uses the same fact and extracts the part of it that carries information. A crowd of 9 that contains 2 independent judgments is a crowd of 2.

What the bandwagon effect is confused with

The bandwagon effect is confused with 6 things, and 3 of them are not errors at all. Each entry below carries the one question that separates it from the bias.

• The appeal to popularity is the same phenomenon written down as an argument. The bias is what a person does; the fallacy is what a person says. Ask: is the crowd being reported as a cause of my choice, or offered to me as a reason for it?
• Network effects look identical from outside and are not a mistake. A messaging app, a rail gauge or a file format genuinely improves as adoption rises. Ask: would this thing still work as well for me if I were the only user?
• Social proof is popularity used as legitimate weak evidence, which is what it is when nothing better is available. A queue outside an unfamiliar kitchen is worth something. Ask: is this the best evidence available, or the easiest, with a load rating or a test report one search away?
• Appeal to authority offers 1 name rather than many. The two are combined constantly in advertising, where the market leader is also the expert's choice. Ask: is the weight coming from a headcount or from a credential?
• Groupthink happens inside a decision making group, where the desire for agreement suppresses dissent. Strangers who never meet can produce a bandwagon. Ask: did these people deliberate together, or merely see each other's choices?
• Confirmation bias takes over once you have joined, making the choice look better afterwards by steering attention to the evidence that supports it. Ask: did the crowd persuade me before the decision, or reassure me after it?

Non sequitur is the category above all of this. The appeal to popularity is a non sequitur with a headcount attached, in the same way that ad hominem is a non sequitur with a person attached, and naming either one identifies where to look rather than settling anything.

When the bandwagon effect is not a fault

The bandwagon effect is not a fault when each person in the crowd checked the thing for themselves, because the same move that fails as an argument then carries real evidence. Popularity is information rather than noise under 3 conditions, and treating those as the fallacy is as much of an error as the fallacy itself.

1. Aggregate the verdicts, if each person judged before seeing what the others chose. Independent estimates average out individual error, which is why 40 unconnected reviews can be worth more than 400 written after reading each other.
2. Count the users, if the value of the thing depends on how many people use it. A messaging app, a language, a rail gauge and a file format are better for you in proportion to their adoption, so the crowd is not a proxy for the benefit, it is the benefit.
3. Read the record, if the thing has been in wide use long enough for failures to surface. A ladder design sold for 30 years has been load tested by everyone who stood on it, and a record across many users is a sample rather than a bandwagon.

One question separates all 3 of those from the fallacy, and it is the question to carry away from this page: did each person test it, or did each person see that the others had chosen it? Testing adds information to the count and copying adds only weight, and the two are indistinguishable from outside unless you ask.

The honest limit is that independence is fragile and usually invisible. A crowd adds information only while its members are deciding for themselves, so a visible count can destroy the very quality that made it worth reading, and by the time a number is large enough to be reassuring it has usually stopped being independent.

The test to run on the next popular claim

Ask 4 questions of any claim that rests on how many people agree.

1. Ask how many of the adopters decided independently, and how many could see what the others had chosen.
2. Ask what the crowd knows that you do not, and whether that knowledge is available to check directly.
3. Ask whether the popularity is the benefit, as with a file format, or a proxy for one, as with a ladder.
4. Ask what you would think of the claim if the number were reversed, and whether anything except the number would change.

The bandwagon effect in economics, and the effects that sit beside it

Economics has treated the bandwagon effect as a demand curve problem since 1950, when Harvey Leibenstein set out three interdependent effects in the Quarterly Journal of Economics. They are usually taught together because they pull in different directions, and knowing all three stops the bandwagon from being read as the whole of social influence.

Overcoming the effect in your own decisions takes one procedural habit rather than any strength of character: write down what would make the option right for you before you look at what anyone else chose. Specify the load, the budget, the two features you actually need. A criterion set in advance is the only thing that survives contact with a crowd, because it lets you notice the difference between a product that meets your list and a product that meets everyone else's.

Where to go next

FALLACIESLogical fallaciesWhat they are, with examplesFALLACIESAd hominemAttacking the person, not the argumentFALLACIESNon sequiturThe conclusion that does not followBIASESConfirmation biasDefinition and examples