Research

Abstract: Recent technology breakthroughs on social media platforms and smart connected devices have provided academic researchers and practitioners with unprecedented access to data on social interactions between consumers in the process of adoption and consumption of products, services, and ideas. However, in many cases, the social connections that can be recovered from such data are fundamentally endogenous since they emerge or get measured while intermediated by peers' interactions with products and therefore are not independent from the outcomes an analyst might be interested in, such as product adoption, consumption intensity, or customer churn. This paper addresses the endogeneity in social connections in the empirical measurement of peer effect in customer churn. Understanding peer effect in churn is of paramount importance in many industries where goods are consumed in a social manner, both online and offline, as churn could be contagious among peer consumers. Equipped with the knowledge of causality in peer influenced churn behavior, managers can craft effective retention campaigns. To tackle the peer network endogeneity issue, we first model the process of social network formation, which is based on observed consumer interactions during product use, and then model the consumers' choice to churn under peer influence. Modeling peer network formation allows us to recover the unobserved individual-specific parameters that might affect both peer network formation and the decision to churn. The recovered latent individual specific parameters correct for peer ties endogeneity in the peer effect model of churn. We apply the model on data from the popular massively multiplayer online game World of Warcraft, where gamers form social groups to progress through the game content. We find significant evidence of peer ties being explained by the latent characteristics of gamers. Based on our estimation results, we run counterfactual simulation studies to investigate how churn dynamics is affected by the structure of the generated peer network and the connection between generated peer network structure and gamer characteristics compositions.

Consumer Information Asymmetry in Online Product Reviews

Manuscript available upon request

Abstract: Firms usually know the distribution of tastes and price sensitivities across the consumer population, or do their best to learn it. At the same time, individual consumers are unaware of this distribution. This constitutes an information asymmetry. This paper shows that when an individual makes a purchase decision using online consumer product reviews as her major source of information about the product quality, the lack of information about consumer heterogeneity does not allow her to correctly infer the true product quality from the reviews. The inferred product quality is systematically biased, and the bias depends on the consumer’s characteristics. In the market with such information asymmetry, a strategic forward-looking firm is able to maximize its profits by setting the price to attract consumers who would give the product a high rating. Using a structural dynamic model and simulations, this paper shows that the firm still earns higher profits if consumers are informed about the consumer heterogeneity distribution, i.e. in the market with no information asymmetry. The findings are in line with the empirical evidence that firms selling their products/services over the Internet (such as hotel booking web-sites) try to reduce the information asymmetry, for example, by disclosing the average product ratings by consumer types to their prospective customers.