Third-Person Effect: Why Ads Seem to Work on Everyone but You

The effect is a comparison, and the two halves of it come from different places. The judgment about yourself draws on what you can observe from the inside: you watched the ad, you did not feel persuaded, you cannot remember buying anything because of it. The judgment about other people draws on a theory of how media work on people in general, which for most people is a theory of strong effects. The two judgments are made with different evidence, and the second is almost always the larger.

Three explanations have been offered for why the gap runs in this direction. The first is self-enhancement: being swayed by an ad is a weakness, and people are reluctant to attribute weaknesses to themselves. The strongest evidence for this account is that the gap reverses for messages it would be creditable to be influenced by. For public service announcements, health advice and other socially approved messages, people tend to rate themselves as more influenced than others, a first-person effect, and Sun, Pan and Shen's 2008 meta-analysis found message desirability to be the strongest moderator in the literature. The second explanation is that people have no access to their own persuasion. Attitude change is slow and mostly invisible from the inside, so introspection reports no effect while the theory of other people reports a large one. The third is that the people being judged are not real individuals but a vague mass, and a vague mass is easy to imagine as gullible.

The gap grows with social distance. Davison noted in 1983 that the effect was larger the more different the others were from the respondent, and later studies confirmed that people rate a message as having more effect on the general public than on their friends, and more on their friends than on themselves. It also grows when the message is one the respondent disapproves of or thinks is poorly made, and it is larger among people who regard themselves as well informed on the topic.

The effect has a behavioral half that is harder to measure and more consequential. Davison's original argument was that people act on their beliefs about what a message does to others: officers withdraw troops they think have been demoralized by leaflets, voters turn out to counter propaganda they think is working on their neighbors, and citizens support restrictions on media they believe are harming other people. Support for censorship of pornography, violent entertainment and misinformation correlates with the size of the perceived gap in study after study, though whether the gap causes the support or both come from a third source is still argued over.

W. Phillips Davison, a sociologist at Columbia University, published "The third-person effect in communication" in Public Opinion Quarterly in 1983. The paper opens with an anecdote from the Second World War. Japanese planes dropped leaflets over a US service unit on Iwo Jima made up of Black troops with white officers, urging the troops not to fight a white man's war. By the account that reached Davison, the leaflets had no visible effect on the troops but a large one on the officers, who had the unit withdrawn. The message had worked on the people who thought it would work on others.

Davison tested the idea with four small informal surveys of 25 to 35 people each, and he presents them as suggestive rather than conclusive. The most direct was a questionnaire given at Columbia in 1978, just after the New York State election and a strike that had shut down the city's three main newspapers. Respondents rated how much the governor's campaign theme about his opponent's unreleased tax returns had influenced the way New Yorkers in general voted, and how much it would have influenced their own vote, each on a scale from 0 to 7. The mean for New Yorkers in general was 3.4; the mean for the respondents themselves was 2.26. Scored person by person, 48 percent saw more influence on others than on themselves, 36 percent saw the same influence on both, and 6 percent saw more on themselves. The other surveys asked about television advertising and children, and about the 1980 presidential primaries, and found the same direction. Davison flagged that his questions were not always worded identically for self and others and that the samples were tiny.

The hypothesis was taken up quickly. Richard Perloff's 1993 review covered the first decade of research and found the perceptual gap in nearly every study that looked for it, with the size depending on the message and on who the others were. Paul, Salwen and Dupagne's 2000 meta-analysis of 32 studies and 121 effect sizes reported an average correlation of .50 between target and perceived effect, though Sun, Pan and Shen later argued that the formula used had overstated it. Their own 2008 meta-analysis of 60 papers, 106 studies and 372 effect sizes, published in the Journal of Communication, estimated the average self-other gap at d = 0.65, equivalent to a correlation of about .31, with a confidence interval from 0.56 to 0.73, and found that it did not depend on the research setting, the population, the study design or how the effect was measured. The perceptual gap is one of the better replicated findings in communication research; its behavioral consequences are much less settled.

The effect appears wherever people estimate what a message is doing to an audience that includes themselves. The rows below are illustrations, not findings from the studies cited on this page.

The gap between the two judgments is the thing to watch, and it can be narrowed from either side.

• Ask the self question honestly. Introspection reports no persuasion because persuasion is not felt; look instead at what you bought, shared or believed after exposure, compared with before.
• Make the others specific. The effect shrinks as the other person becomes more like you. Instead of the public, think of a named friend, and ask whether the message would really move them.
• Notice when the message is one you dislike. The gap is widest for messages people disapprove of, which is exactly when the estimate of harm to others is least reliable.
• Before supporting a restriction on some kind of content, separate two questions: what evidence shows it actually does to audiences, and what you believe it does to other people. The second is where the effect lives.
• When you create persuasive material, test it on your audience rather than predicting its effect. Your prediction will be inflated for the people you think of as others and deflated for the people you think of as like you.

The third-person effect is a specific self-other gap about the influence of messages. Several broader biases produce similar gaps about other things.

• It is not the bias blind spot, the general tendency to see cognitive biases in others more than in oneself. The third-person effect is one instance of that pattern, restricted to persuasion by media and messages.
• It is not illusory superiority, the finding that most people rate themselves above average on desirable traits. Being hard to persuade is one such trait, and self-enhancement is one of the explanations offered for the third-person effect, but the effect also appears in forms that self-enhancement does not predict, such as the reversal for desirable messages.
• It is not naive realism, the conviction that one sees the world as it is and that those who disagree are biased or uninformed. Naive realism is about the truth of one's views; the third-person effect is about how those views were formed.