An empirical investigation into the drivers of re-subscription in massively multiplayer online games : a commitment trust theory approach

Page iii Abstract This is a relationship marketing PhD which is examining, using Commitment Trust Theory, the customer decision to maintain subscribing to a massively multiplayer online game. This PhD is not an examination of initial purchase decision, but of the ongoing, post-purchase, customer retention. In keeping with the contextual nature of Commitment Trust Theory, this study examines the antecedents of the re-subscription decision and their effect on the key mediating variables of Commitment and Trust and modifies the framework to model the subscription based nature of the business situation and the context. The key contribution of this research to the literature is the application of the Commitment Trust framework to a customer’s ongoing relationship with a massively multiplayer online game entertainment product; a situation and context which has not been examined in the literature. An online questionnaire survey was used to collect a sample of data from 2226 massively multiplayer online game customers. This sample data was then analysed using Structural Equation Modelling to test the relationship hypotheses between the constructs proposed by Commitment Trust Theory. Furthermore, hypotheses examining the effect of relevant demographic and categorical variables upon the constructs of Commitment Trust Theory were also tested and analysed using appropriate statistical techniques. Evidence was found to support the Commitment Trust Theory framework in a massively multiplayer online game subscription situation, with the study’s model explaining 85.7% of the variance of the sample data, with evidence presented to support the key mediating variable approach to modelling the circumstances. The study, based on examining the effect size of the construct relationships using standardised regression weights then gives evidence that a more parsimonious model which reduces the number of constructs from 16 to six (a 70% reduction in complexity) would still produce a model explaining 85.3% of the variance of the sample data (a 0.4% loss in explanatory power).