Disconfirmation Effect on Online Rating Behavior: A Structural Model
Information Systems Research, 28(3), pp. 626-642, 2017
38 Pages Posted: 31 Oct 2014 Last revised: 30 Mar 2018
Date Written: 2017
This research studies the effect of “disconfirmation” — the discrepancy between the expected and experienced assessment of the same product — on the behavior of consumers leaving online product reviews. We propose a modeling framework in which an individual’s pre-purchase expectation is shaped by (1) the product ratings she observes and (2) the perception of the review system she has at the time of the purchase. Upon product consumption, the individual obtains the post-purchase evaluation and encounters a certain level of disconfirmation. Drawing on Bayesian learning framework, we model individual perception of the review system as a subjective attitude underlying how well the aggregate ratings match one’s own usage experience. A hierarchical Bayesian model is developed and estimated using a rich data set comprising complete purchasing and rating activities on an e-commerce website.
Our results suggest that an individual’s decisions of whether to post a rating and what rating to post are affected by disconfirmation in two distinct manners. Specifically, an individual is more likely to leave a review when the magnitude of disconfirmation she encounters is larger. In addition, when the individual decides to rate a product, the rating she chooses may not neutrally reflect her post-purchase evaluation and the direction of such a bias is in accordance with the sign of disconfirmation. We also observe several moderating effects: the disconfirmation effect on posting is attenuated by the time gap between purchase and receipt of the same product but accentuated by the dissention in product evaluations among peer consumers. A more granular examination reveals that infrequent raters are systematically more susceptible to disconfirmation than frequent posters. The insights from this research lead to actionable strategies for marketers and designers of recommender systems.
Keywords: disconfirmation, online reviews, rating bias, user-generated content, word of mouth, learning models, hierarchical Bayes, Bayesian estimation
Suggested Citation: Suggested Citation