This happens when respondents may opt out of undertaking an the survey based on its content by whatever means. This distinguishes the results by over-sampling people with an interest in a certain subject or by over-polarization.