What Kind of Trust Do We Measure Using the Generalized Trust Question? : An Approach Based on Latent Class Model

Despite the importance in social capital research, empirical studies on generalized trust show inconsistent results. For example, trust studies have not shown clear results with respect to the relationship between group participation and respondents’ level of generalized trust (Nannestad 2008). This study shows that a part of inconsistency is due to the measurement of generalized trust. Generalized trust is measured by such questions as ”Generally speaking, would you say that most people can be trusted, or that you can’t be too careful in dealing with people?” This kind of question is employed in major social surveys such as World Value Survey (WVS). This study hypothesizes that respondents think some different kinds of trust when they answer the generalize trust question. This study extracts heterogeneity of trust respondents think by latent class analysis (McCutcheon 1987), using Japanese dataset (Social Survey of Residents’ Networks and Health). As a result, this study extracts four types of trustors; (a) active trustors (27.6%), who show higher levels of generalized trust and participate all kinds of social groups, (b) inactive trustors (31.3%), who show higher levels of generalized trust but do not participate any kinds of social groups, (c) parochial trustors (19.6%), who show higher levels of generalized trust but participate local groups only, (d) distrustor (21.4%), who show lower levels of generalize trust and do not participate any kind of social group. This result shows that respondents think four different kinds of trust when they answer the generalize trust question. Inconsistent results in previous studies may be caused by the difference of group participation between high trustors.