Assessing the validity of generalized trust questions: what kind of trust are we measuring?

In the social capital literature a distinction is made between trust expressed in people in general, and trust in people who are known to us personally. In this article we investigate the frames of reference respondents make use of when answering two commonly used interpersonal trust questions. Half of our sample was administered a version which asks respondents whether "most people" can be trusted. The other half of the sample was administered an alternative version of the question, in which the object of trust is restricted to "people in your local area." Immediately after answering the trust question all respondents were asked to report, in their own words, who came to mind when formulating their response. Counter to the widespread assumption that these questions measure generalized trust, we find that a substantial number of respondents report having thought about people who are known to them personally. Furthermore, respondents who report having thought about individuals who are known to them also report substantially higher levels of trust than people who say they thought about abstract categories such as "people in general." Our results suggest that apparent differences in trust across question formats and groups within the general public derive, at least in part, from heterogeneity in question interpretation