Participation Bias

Participation bias, also called non-response bias, is the distortion that occurs when the people who take part in a study or survey systematically differ, in ways that affect the outcome being measured, from the people who decline to take part, so the resulting sample no longer represents the population it was meant to describe. Researchers have documented concrete examples of who tends to opt out: people who refuse to answer surveys about AIDS, for instance, have been found to be older, to attend church more often, to be less likely to believe their answers will stay confidential, and to disclose less about their own sexual history than those who respond. The most cited historical case is the Literary Digest's 1936 prediction for the United States presidential election: the magazine mailed 10 million questionnaires, received 2.38 million responses, and forecast a decisive win for the Republican candidate Alf Landon with 57.08 percent of the popular vote, but Franklin D. Roosevelt won instead, with Landon taking only 37.54 percent of the vote and eight electoral votes; later analyses in 1976 and 1988 concluded the error was mainly caused by non-response bias, with problems in the original sampling frame as a secondary factor. The same underlying pattern can also emerge across a longitudinal study through attrition, as participants who are younger, economically disadvantaged, or dissatisfied with the study drop out at disproportionate rates as it continues.