Response bias

From RationalWiki
Jump to navigation Jump to search
Tell me about
your mother

Icon psychology.svg
For our next session...
Popping into your mind

Response bias is the name of a group of cognitive biases which commonly influence the results of surveys, questionnaires, interviews, and polls. Like statistics, it's a great way to lie while still being honest, although honest mistakes are also common.


Acquiescence bias[edit]

Acquiescence bias is the tendency for people to respond positively ("yeah, sure") to survey questions when in doubt.

For example, 55% of respondents agree and 42% disagree with the statement, "The best way to ensure peace is through military strength", but when given the choice between the former and "DiplomacyWikipedia is the best way to ensure peace", only 33% agree with the first statement, and 55% percent agree with the second.[1]

The Pew Research Center recommends that it is better practice to offer respondents a choice between alternative statements, than to simply ask whether they agree with something.[1]

Social desirability bias[edit]

Respondents eschew answers that are considered socially undesirable. This is especially common on topics such as drugs, immigration, personal finances, health, religion, sex, and personal achievements.

Some examples:

  • About 15% of US males report visiting a prostitute at least once in their life, but other surveys also show that men report a lot more sexual partners than women do, so something doesn't add up here. Once you account for the prostitution, the numbers do seem to add up, though, and you end up with about 23% of males who've paid for sex.[2]
  • Responses will differ on topics such as feminism or sex depending on the gender of the researcher (especially when the participant is male).

Demand characteristics[edit]

This is the tendency of participants to (subconsciously) fit their behaviour to what is perceived to be the expected or desired behaviour.

Some examples:

  • Rape myths were more accepted when the researcher was dressed conservatively than when the researcher was dressed provocatively.[3]
  • When informed that the survey's interest was menstrual cycle symptoms, women reported significantly more negative symptoms than women who were not told the survey's interest.[4]

Extreme responding[edit]

The tendency to only select the most extreme responses in a survey where one is expected to choose from a scale (eg. to choose "Strongly agree" over "somewhat agree").