False Consensus: Public (Mis)Estimation of Public Opinion

By Greg M. Shaw

Pollsters have made much over the past seventy years about how important it is for citizens to understand the contours of public opinion on political issues. George Gallup and Saul Rae elaborated this in their 1940 classic, The Pulse of Democracy, and Frank Newport, editor-in-chief of the Gallup poll, revisited the theme in his 2004 book, Polling Matters: Why Leaders Must Listen to the Wisdom of the People. Part of this argument hinges on the benefits of helping citizens correctly estimate what the public thinks.

As it turns out, most people aren't very good at doing this on their own. The average person struggles to guess what percentage of the public as a whole thinks one way or another about any given topic. Asking survey respondents to estimate public opinion tends to elicit systematically biased answers. Specifically, most people overestimate the degree of popular consensus surrounding their own positions. This is the false consensus effect, and it carries implications about the ability of errors in individuals' judgments to cancel each other out in large publics. Lacking this self-correcting quality, public opinion may not be as wise as some would have it.

The degree to which people project their own beliefs onto liked objects, such as the U.S. Constitution, is astounding. National surveys over the past two decades have produced such findings as 29 percent of Americans believing the Constitution guarantees everyone a job and 75 percent who thought Americans are constitutionally guaranteed a high school education, according to a Hearst Corporation survey conducted by Research and Forecasts in 1986, and 30 percent believing it guarantees health care, reported in a 1991 survey by the Los Angeles Times. False consensus results when people project their own beliefs in this manner onto public opinion.

False consensus effects vary in size. They appear more powerfully when survey respondents sense a personal stake in question, which suggests that more intimately held beliefs, such as fundamental elements of one's world view, are more subject to this distortion than estimates of who will win an upcoming election in a neighboring state. Respondent sophistication also plays a role, suggesting that one's ability to perceive popular consensus accurately may be partly a function of education. More highly educated people should be able to estimate more closely the president's job approval rating than those with less education.

The causes of false consensus remain somewhat unclear. Psychologist Bernard Whitley has suggested perceptual bias-specifically, salience-rather than such explanations as the motivation to conform, the ability to process information logically, or the strength of respondents' own beliefs. In Whitley's 1998 study, the more sexually active friends college women could name, the more sexually active they thought college women generally were, regardless of the status of their own virginity.

We know a lot about how dramatic false consensus can be, but we know less about what kinds of issues tend to elicit it and the role of education in immunizing respondents from it. To shed some light on these two questions, my students and I conducted a telephone survey of 513 randomly selected adults in central Illinois last spring. Respondents were asked to estimate the percentage of Americans who smoke, attend church or synagogue weekly, approve of President George W. Bush's job performance, approve of homosexual couples adopting children, and believe the Bible to be literally true. Respondents were then asked if they themselves maintained these habits and opinions. The false consensus questions were embedded toward the end of a larger questionnaire that asked about several local issues, and the results were compared to findings from the same questions on national surveys.

These subjects were selected to represent a variety of issues spanning behaviors and opinions, more- versus less-familiar topics, and symbolically important as well as more mundane themes. None of the positions polled represented a universally reprehensible status, so there was no clear "out-group" on any of the issues. (This last point was particularly important because the existence of out-groups has been shown to minimize false consensus effects by members of those groups. People don't brag about being child abusers.) Finally, the selection represented subjects on which reasonable people disagree.

While there is certainly no objective way to rank the intimacy of the survey items for respondents, biblical literalism is arguably more of an intimate belief (importantly, among biblical literalists) than approval of the president's job performance. Similarly, cigarette smoking, a common habit, is probably less likely to trigger false consensus-building than belief in the rights of gay couples to adopt children, a hot-button issue. Though it is difficult to rank these items precisely in terms of intensity of belief, certainly these five questions provided some diversity along that dimension.

Due to local demographics, the respondents to the Illinois survey were somewhat more educated (53 percent were college graduates), of higher income (41 percent lived in households earning between $51,000 and $100,000 annually), and more Republican (43 percent, including leaners) than the nation. The sample was 61 percent female, the median age forty-three years. With the exception of polling too many women, the sample closely reflected the known demographics of the area.