Nonresponse bias in trust surveys

Social trust is typically measured using surveys that ask people if they agree that most people can be trusted. A potential problem is that falling response rates plague these surveys. If nonresponses are systematic, comparisons of social trust over time will be biased. We examine social and legal trust among non-respondents by conducting a classroom survey where the first part included questions on social and institutional trust and is answered during class, whereas a second part of the survey is handed in by respondents later. Surveys from 300 Danish and Swedish university students suggest that if anything, social trust among survey responders is somewhat lower than among non-responders. Using two waves of the Swedish National Election Study, we also show that conditional on education, social trust is uncorrelated to dropping out of the panel survey.