Sociodemographic Differences in Respondent Preferences for Survey Formats: Sampling Bias and Potential Threats to External Validity