Essays on Network Effects on Online Reviews

Abstract

This dissertation studies online reviews and focuses on the network effects on individuals' incentives to contribute to an important form of online word of mouth—online reviews. Provided by either consumers or third-party professionals, online reviews are closely correlated with consumer purchasing decisions and hence sales. Individuals have incentives of free riding and maximizing social capital when providing feedbacks online. The first essay, "Free Riders versus Social Capital: An Empirical Analysis of an Exogenous Shock on Online Reviews," examines the average peer effects. We leverage a "natural experiment," which led to an exogenous expansion in the user population of a major online review platform to better understand the trade-off between the two conflicting incentives. We find that a larger population of audience and peer review writers, an immediate consequence of the exogenous shock, causally led to more reviews posted, higher and more diverse ratings assigned, and reviews of higher quality by the users. We continue our exploration of the peer impacts on individuals' contributions to online reviews in the second essay, "Impact of Network Size on Contributions: The Moderating Roles of User Characteristics," by studying the moderating roles of several user characteristics. We find that, first of all, the increases in an individual's contribution volume, valence (the review ratings), and helpfulness of reviews are all smaller for a more active user. Also interesting enough, we find that a user who focuses more on book reviews responded more positively to the exogenous shock, while in sharp contrast a user with preferences in writing other reviews was less affected. Last but not least, although not statistically significant, the popularity plays a negative role in moderating the group size effects. The third essay, "Heterogeneity in Peer Effects: An Application of Finite Mixture Models," further extends our study of peer influence to consider the impacts of unobserved heterogeneity. An example of latent variables under investigation is individuals' intrinsic motivation. We use a Finite Mixture Model to identify the segments of users. Our statistical process recommends that a three-segment model performs the best. We interpret the three groups as users of low motivation, mediocre motivation, and high motivation. Interestingly, we find that the increases in contribution volume, valence, and quality of reviews are all bigger for users of high motivations than the other two segments. We also examine how the moderating roles of user characteristics vary across segments. Our findings have important theoretical as well as managerial implications.