False consensus effect

Nobody uses the shared calendar

A manager who never opens the team’s shared calendar proposes canceling the subscription: “Nobody actually looks at that thing.” A two-question poll the next day shows that most of the team checks it daily.

The manager had no information about anyone’s habits except his own, and generalized from it to the whole team without checking. This is the clear-cut form: one person’s habit, taken as the norm, when asking would have been easy.

The mascot costume

The school fair needs a parent to spend an hour in a hot mascot costume. One parent agrees at once and says, “Most people would jump at this, it’s fun.” Another declines and says, “You’ll be lucky if anyone says yes.”

This is the shape of the original experiments: people who make a choice expect more others to make it than people who choose the opposite do. Both parents can’t be right about the same group of parents. But each one’s own reaction is some evidence about how parents feel. The false consensus effect, in the strict sense, is leaning on that single reaction more than it deserves, which you can’t tell from one estimate alone.

The rare early bird

In a book club of 40 members, five prefer meeting at 7 a.m. before work. One of them proposes moving the meetings: “I’d guess half the club would rather do mornings.”

Projection does the most damage when you’re in the minority. Someone who holds the majority view and assumes others share it will usually be roughly right; someone in a small minority who does the same will overestimate their side badly. Reviews of consensus estimates report this pattern: people in the majority slightly underestimate how common their view is, while people in the minority greatly overestimate it.

Using your own view to estimate other people’s is often reasonable:

• When you have nothing better to go on. Your own reaction is a genuine sample of one, and with no other information it is reasonable to let it nudge your estimate.
• When the people you’re estimating are similar to you in the way that matters (same situation, same background, same constraints).
• When you’re likely to be in the majority. On any two-sided question, most people are by definition on the majority side, so for most people, assuming others agree points the right way.
• When you give it up once real information arrives. A reasonable guess based on yourself should yield to a poll, a sign-up sheet or a sales figure.

The test: am I treating my own answer as one data point, or as a stand-in for everyone?

A guess, then a correction

On her first day in a new office, Dana is asked whether the team would like the thermostat raised. “I’m freezing, so I’d guess most people are too, but I only just got here.” Later a coworker tells her that the room is kept cool because several people complained it was too warm, and Dana drops the idea.

Dana used her own experience as evidence about others, which is what the effect describes. But she had no other information, she flagged how little she was going on, and she changed her mind as soon as she learned what others actually thought. This is the nothing better to go on and give it up once real information arrives conditions above.

Same conditions, same complaint

A group of volunteers spends a July morning clearing a trail with no shade. One of them, exhausted by noon, suggests starting at 6 a.m. next time: “I’m sure we’d all rather not do this in the midday heat.”

She is projecting her own preference, but onto people who were in exactly the same situation: the same heat, the same work, the same hours. When the thing driving your view is shared by the people you’re estimating, your own reaction is good evidence about theirs. This is the similar to you condition above.

Several explanations have been proposed, and they aren’t mutually exclusive:

• Your own view is the easiest one to consult. Experiments find that people’s own answers come to mind faster than their estimates of others, and that when estimating consensus they rely more on their own answer than on another individual’s.
• Selective exposure. People tend to spend time with others who are like them, so the examples they can recall are already skewed toward their own views.
• Motivation. Believing your views are normal can feel reassuring, which some researchers argue pushes estimates in the same direction.
• Reasonable induction, overdone. Using yourself as a sample of one is a sensible strategy (see Evidence); the error is in how heavily the sample is weighted, or in not looking for better data when it’s available.

In the original research, the effect also shaped how people judged others. People tended to see those who would choose differently from themselves as more unusual, with more extreme personalities. That connects it to research on how people read character into behavior, such as the Fundamental attribution error.

A few remedies have been tested:

• Knowing about the bias didn’t remove it. In a 1994 set of experiments, people who had been taught about the effect, or given feedback about actual consensus, still showed it.
• Prominent, representative information did. In incentivized experiments where people were shown what a random sample of others had chosen, the effect disappeared, and people actually gave their own choice less weight than the information. When a small effort was needed to work the same information out, the effect came back.

So the practical step is the one the test above points to: before assuming others agree, find something that tells you what they actually think, and put it where you can’t miss it.

Status: contested. The basic pattern, people’s estimates of others shifting toward their own choices, meets this site’s bar for “robust”: it has replicated in a large preregistered multi-lab study with large effects. But the entry’s claim is that this is an overestimate, and whether the pattern is an error, rather than sensible reasoning from a sample of one, is seriously disputed, with credible evidence on both sides about when a truly excessive projection appears. Where the evidence falls between two labels, the site chooses the less confident one.

The original studies. Ross, Greene and House (1977) gave 320 undergraduates one of four short scenarios, each ending in a choice between two actions, and asked them to estimate what percentage of their peers would choose each option, and then which they would choose themselves. Across the scenarios, people who picked an option thought it more common than people who picked the alternative did. In most of the scenarios, they also rated the kind of person who would make the opposite choice as having more extreme traits. In another study from the same paper, as it is commonly described, students were asked whether they would walk around campus wearing a sandwich board reading “Repent”: those who agreed estimated that roughly 64% of their peers would too, and those who refused estimated roughly 23%.

Reviews. Mullen and colleagues (1985) meta-analyzed 115 tests of the effect; the Many Labs 2 team summarized it as robust and moderate in size across paradigms (r = 0.31). Marks and Miller (1987) reviewed ten years of research and the competing explanations, including selective exposure and cognitive availability.

The replication. Many Labs 2 (Klein et al., 2018), a preregistered project run in dozens of labs in many countries, re-ran two of the 1977 scenarios, one about a supermarket and one about a traffic ticket, with more than 7,000 participants each. In the supermarket scenario, people who chose the first option estimated that 69% of others would too, against 43% among those who chose the second (d = 1.18); for the traffic ticket, 72% against 49% (d = 0.95). Every sample in the first scenario showed a significant effect in the predicted direction, and nearly all in the second. The sandwich-board study, which involved a real choice rather than a hypothetical one, was not part of the project, and no direct replication of it was found in reviewing this entry. Independent economics experiments have also found the pattern (Dohmen and colleagues, 2023).

Is it false? What those studies measure is the gap between two groups’ estimates. Dawes (1989) argued that such a gap doesn’t show an error: your own choice is information about the population, so a rational person should estimate their own choice as somewhat more common than someone who chose the opposite would. On this view the effect is false only if people weight their own choice more heavily than they would weight a randomly chosen other person’s. Researchers then designed tests of that stricter, “truly false” version:

• Krueger and Zeiger (1993) and Krueger and Clement (1994) compared each person’s estimates with the actual consensus across dozens of personality-questionnaire statements, and checked whether the errors leaned toward the person’s own answers. They did. The bias persisted in people who had been taught about it, given feedback about actual consensus, or shown the unanimous answers of 20 randomly chosen others, and people’s estimates were more affected by their own answer than by another person’s.
• Engelmann and Strobel (2000) ran incentivized experiments in which people made simple choices and then estimated others’ choices. When they were shown the choices of a random subset of other participants, there was no false consensus effect: on average, people gave their own choice less weight than the information. In a later study (2012), the effect reappeared when retrieving the same information took a small effort, and the authors concluded that the cause was a failure to use information that isn’t prominent, rather than egocentrism.
• Real-world estimates. Furnas and LaPira (2024) surveyed lobbyists, civil servants, journalists and other political professionals, and found their misperceptions of actual public opinion systematically matched their own policy preferences.

Projection as a strategy. Krueger, whose experiments found the “truly false” effect, has also argued that social projection, the more neutral term for using yourself to estimate others, is often useful. A meta-analysis (Robbins and Krueger, 2005) found that people project more onto groups they belong to than onto groups they don’t, and described projection as an egocentric shortcut for inductive reasoning (reasoning from cases to a general conclusion). A 2007 review describes it as allowing “quick and reasonably accurate predictions about others.” Krueger and Clement (1997) showed that one assumption, that most people believe they’re in the majority, explains why majority members slightly underestimate their group’s size and minority members strongly overestimate theirs.

Popular version. The effect is often described as “everyone thinks everyone agrees with them.” The research claim is narrower: estimates of others lean toward your own position. Leaning somewhat is expected even from a perfectly rational estimator, and the dispute is over how much more than that people lean, and when.

What remains uncertain. How large the truly excessive part is in everyday life, where representative information is rarely handed over but is often available with some effort. Most tests of the strict version found in reviewing this entry are older, smaller and from a few labs, and the large replication measured only the traditional version, which cannot separate reasonable projection from overprojection.