It's no small matter. Bias can unknowingly creep into the most professional of surveys. Its impact on data quality can be significant and destroy any faith you have in the data collected.
Let's first put a spotlight on eight examples of survey bias:
1. Formatting Issues
Respondents who react or express an attitude based on the way the survey is structured.
2. Wording Choices
Negatively or ungrammatically worded sentences can adversely influence responses.
3. Pleasing the Researchers
Respondents who modify their opinions to support what they believe are the objectives of the survey. (They might even have prior knowledge of the survey.)
4. Maximum Responders
Respondents who believe the survey requires only extreme responses. Strongly positive or negative replies may occur when there are too many response scale questions, i.e. those requiring an answer on a scale, for example, of 1 to 5 where "1 is Very Important and 5 is Not Important at All".
5. Non-committal Responders
Respondents who believe all response scale questions should be answered as neutral, middle-of-the-road due to perhaps a lack of interest or non-engagement with the survey.
6. Truth Deniers
Respondents who feel they must reply to questions of a sensitive nature (i.e., concerning alcohol consumption, cigarette smoking, sex, etc.) with how they would prefer to be perceived rather the offering the truth.
Respondents who are either trying to please the researchers or truly have no opinion so they agree or disagree with every question, even to the point of contradicting themselves.
8. Sour Pusses
Respondents who do just the opposite and disagree with every question.
The most common and effective approach addressing many of the above issues is via a technique called survey randomization. This allows you to alter the order and flow of unrelated questions to capture true and honest opinions, remove discrepancies in data collection, and, in turn, reduce bias. Survey randomization ensures authentic, high quality insights needed to move forward with the results of any survey.
About the author: