The impact of response format on relations among intentions, attitudes, and social norms

Abstract

This research investigates the response consistency error which may be induced by certain common methodological practices in tests of predictive models such as the theory of reasoned action (Ajzen and Fishbein 1980). Specifically, the physical presence of a questionnaire was expected to provide respondents an inducement and basis for maintaining consistent responses. In addition, the practice of presenting all measures of a given construct together seemed likely to heighten correlations among constructs, resulting in heightened predictions. Two experiments were conducted in which respondents completed measures of the theory of reasoned action for a variety of behaviors using either a paper or computer-administered questionnaire with items in standard or random order. The results of these studies indicate that, contrary to expectations, the degree of attitude-intention consistency across behaviors may be attentuated in standard paper administrations.