Are Polls Accurate?

Abstract

Polls can be and frequently are extremely accurate. They also can be and frequently are grossly inaccurate. Surveys that measure facts, (such as habits or ownership,) tend to be more accurate than surveys that measure opinion, and surveys that measure opinion are, in turn, more accurate than surveys that attempt to measure future intentions. While sampling error is the most commonly given reason for polling error, it is usually the smallest error in a survey. Other reasons for error in a survey include poor timing of the survey, misinterpretation of survey results, and improper context in which a question or question series is asked. The greatest error of all, however, results from question wording. “Yeasay” questions—questions that implicitly suggest that yes is the right answer—produce a consistent bias in results. But many other kinds of error, often of greater magnitude, result from other aspects of question wording. It is far more important in assessing the accuracy of a survey to know the wording of the questions asked than the magnitude of the statistical sampling error.