This paper empirically identifies the attitude‐behaviour gap phenomenon, which has been witnessed frequently in environmentally friendly products, by suggesting a general and flexible Bayesian multivariate model and applying it to survey data. Since cognitive, affective and conative stages, which are closely related to the attitude‐behaviour gap, are successive stages in the consumer decision‐making process, these must be considered simultaneously in empirical analysis. For this reason, a recursive multivariate probit model is proposed. According to the empirical analysis, the affective stage (interest and preference) does not lead to the conative stage (intention to use and intention to purchase) in an analysis of eco‐friendly detergent. By contrast, in an electric vehicle analysis, the affective stage partially influences the conative stage through the relationship between preference and usage intention, and through the relationship between interest and purchase intention. It was also found that high compatibility and low complexity must be emphasized in eco‐friendly detergent, and that high relative advantage and compatibility must be emphasized for an electric vehicle in order to bridge the attitude‐behaviour gap successfully.