False Consensus Effect: Why Most People Seem to Agree With You

Your own position is the most available piece of evidence you have about what people think. When a question arrives, the answer you would give comes to mind first and fastest, and an estimate about everyone else gets anchored on it. Nothing deliberate is required: the estimate is built from the sample you have, and that sample starts with one person who is certainly you.

Three ordinary mechanisms do most of the work. The first is selective exposure. The people you know are not a random draw from the population; they share your city, income, work, interests and often your views, so the evidence actually available to you really does show high agreement. The second is construal. Two people who answer the same question differently are often answering different questions, because each has filled in the missing detail in their own way, and the version you have in mind seems obviously the one most people would act on. The third is the resolution of ambiguity: when you have no information, your own answer is a better guess than nothing, so using it is a reasonable heuristic that happens to be miscalibrated in a predictable direction.

The effect is relative rather than absolute, a point Ross, Greene and House were careful about. Someone who holds an unusual view usually knows it is unusual; what the research predicts is that they will still estimate more agreement than people on the other side do. Joachim Krueger and Russell Clement found in 1994 that the gap persisted in their experiments when people were shown the actual consensus figures and when they had been taught about the effect beforehand, which is why they called it truly false. That is a result about one set of designs rather than a law: how the information is presented matters, and it does not show that no form of correction works.

There is a corollary in the same 1977 paper that gets less attention. Behavior that differs from your own looks more revealing of the person doing it. If you would decline the sandwich board, the student who agrees to wear it looks like a particular kind of person; if you would wear it, the decision tells you nothing much about anyone. The bias therefore produces both an error about numbers and an error about character.

Lee Ross, David Greene and Pamela House published "The false consensus effect: An egocentric bias in social perception and attribution processes" in the Journal of Experimental Social Psychology in 1977. It reports four studies.

Study 1 gave 320 Stanford undergraduates, 80 per story, one of four written scenarios: signing a release for a supermarket commercial, voting for group or individual term papers, paying a traffic fine or contesting it, and a referendum on funding space exploration. Each reader estimated what share of their peers would choose each option, then said which they would choose themselves and rated the traits of the typical person choosing each. On every story, people who said they would pick an option estimated it as more common than those who said they would pick the alternative. Averaged across the four stories, those choosing the first option put its popularity at 66 percent, while those choosing the second put it at 49 percent.

Study 2 took the effect beyond single scenarios. Eighty students sorted themselves into one of two categories on 35 person-description items covering traits, habits, problems and political expectations, and estimated the share of college students in general in each category. Thirty-four items were usable, and the difference ran in the predicted direction on 32 of them.

Studies 3 and 4 were the sandwich board. In Study 3, 104 students read the request as a hypothetical and made estimates; in Study 4, 80 students faced it for real, choosing privately after the request had been made in the room. Combining the two signboard messages used in Study 4, "Eat at Joe's" and "Repent," the 48 students who agreed estimated that 62 percent of peers would agree, and the 32 who refused estimated 33 percent. The effect survived the move from a questionnaire to a decision with consequences, which was the point of running it. Brian Mullen and colleagues collected 115 hypothesis tests of the effect for a meta-analysis eight years later.

The effect appears wherever a decision depends on an estimate of what other people want or do. The rows below are illustrations, not findings from the studies cited on this page.

The bias comes from using your own answer as data, so the counters are all ways of getting other data.

• Write your estimate down before you look it up, then look it up. A prediction on paper can be scored; an impression cannot.
• Ask people who are not like you, and ask them in a way that does not reveal your own answer first. The order of those two things decides what you learn.
• Count rather than canvass. Five conversations feel like evidence and are not a sample; a short anonymous poll of everyone affected usually is.
• State the question precisely. Much apparent disagreement is two people answering different versions of a vague question, so write down the version you have in mind and check that others have the same one.
• Expect your estimate to be too high and adjust toward the base rate, especially on questions where your position is strongly held.
• When you discover you are in the minority, resist the second half of the bias: a person who chose differently is usually not a different kind of person, just someone who weighed a close call the other way.

Several biases involve projecting from yourself outward, and they differ in what is being projected.

• It is not projection bias, which is about projecting your current tastes and state onto your own future self, as when a full shopper buys too little food for next week. False consensus projects onto other people instead.
• It is not naive realism, the conviction that you see the world as it is and that reasonable people therefore must agree with you. Naive realism is the belief about objectivity underneath the estimate; false consensus is the numerical error in the estimate.
• It is not the spotlight effect, which is about overestimating how much others notice you. That one is about attention directed at you, not about agreement with you.