Modeling the Formation of Dyadic Relationships between Consumers in Online Communities

Members of online communities access consumption related information posted by several other members. Over time, members are influenced by some members of the community, and develop informal relationships with them. We propose a model for the timing of the establishment of relationships between two community members. We explicitly model the salient properties of dyadic data (data on the timing of relationships formed between a pair of community members) arising due to their inherent interdependence: the correlation between the dyadic observations due to reciprocity, the correlation between sender and receiver specific observations due to the "duality" of roles of each member, and the interaction of the latent characteristics of dyad members due to "homophily". The proposed model outperforms benchmark models in terms of in sample fit and predictive power. Moreover, ignoring the properties of dyadic data leads to biased inferences about the effects of covariates. We explore the factors that influence relationship formation between community members who post product reviews in an existing online community. The amount and valence of the information that the dyad members share, the similarity between the information that they post and their network characteristics affect relationship formation. We demonstrate how the model can be applied to improve the targeting effectiveness of marketing campaigns through agents in online communities