Conferences

The analyses of items addressing the measure of individual trust in institutions often result in factor solutions that underline the multidimensionality of the concept. However, little attention is paid to the nature of the items and to their distributions (e.g. dichotomous items, or four point-scales with skewed distributions), characteristics that hardly fulfill the requirements of factor analysis. The paper addresses this problem explicitly and shows how the cumulative nature of the data on institutional trust can be dealt with in a more efficient and convincing way through a non parametric IRT technique, namely Mokken scaling. The dominance model developed by Mokken (1971), considering explicitly the order of items in terms of absolute frequencies of positive answers, helps also to take into account an aspect often removed from the analysis of trust, that is the trustworthiness of specific institutions. The application of Mokken analysis also have conceptual consequences, as it is shown that once the order of the items is taken into account, the measure of trust in institutions can be reduced to one single dimension. The problem of cross-country comparisons is addressed and a systematic procedure to test invariance of item order across country is proposed. The data used to carry out the comparison come from international datasets, namely Eurobarometer and Eurorpean Value Study.