Can we trust the general trust question? A survey experiment
The extent to which citizens trust one another is a key concern around the world. Robert Putnam's infl uential theory of social capital has led to academic and policy interest in comparative and longitudinal variation in citizen trust. In the United States, Britain and other advanced democracies, the long-term trend appears to be downward, leading to pessimistic conclusions regarding the social atomisation and fragmentation of modern societies. The vast majority of the empirical evidence for this decline in social trust comes from what we refer to as the General Trust Question (GTQ). This question, which has been fi elded in a great many national and international surveys, asks respondents to choose whether they think 'most people can be trusted', or 'you can't be too sure'. Clearly, this question is not without its problems methodologically. First, it is worded in a very general way. There is good evidence to suggest that questions of a very general nature tend to be interpreted in diverse ways by respondents and are particularly sensitive to context and quesation ordering effects (Bishop 2005). Second, in some surveys the GTQ is offered with two response alternatives ("can be trusted" or "can't be too careful") and in others an "it depends" option is is offered. It is not at present clear what the effect is of offering this additional response alternative and whether it is potential 'trusters' or 'non-trusters' from the binary alternative, who are more likely to choose it. In this paper we present results from a split ballot experiment to evaluate the sensitivity of the GTQ to these varying response formats and to context and framing effects arising from question ordering. Results indicate that sometimes large differences in levels of trust can emerge as a function of response format and question order.