Essays on Social Demand Estimation: Evidence from CAPS and Steam
Author
Clayton, Andrew
Metadata
Show full item recordAbstract
This dissertation examines the role of peer influence on the demand decisions made by socially connected consumers within markets for household investment and digital video games. In both applications, novel and rich data provides visibility into demand outcomes and the evolution of social networks, allowing for the estimation of demand models which quantify the impact of socialization on individual demand. These findings contribute to our understanding of how increases in connectivity and visibility into the consumption choices of peers change market function and consumer choice. Chapter 1: Evidence suggests that the investment choices of household investors are influenced by the financial decisions made by their peers. An unresolved empirical challenge is to quantify the impact of socialization on portfolio composition relative to other factors. Rich individual data from the Motley Fool CAPS platform provides an opportunity to link asset selection decisions through an observable social network of peers. I collect data for nearly one-thousand interconnected users who collectively produce over 600,000 stock predictions between 2007 and 2013 and estimate the likelihood of asset demand conditional on a rich set of financial, macroeconomic, and individual factors. I further augment this model with measures of network socialization and peer influence to characterize how the decisions made by a user's peer group affect their own investment outcomes. The results demonstrate that the inclusion of social covariates in a model for asset demand substantially improves explanatory power and that social factors appear strongly influential in explaining household investor actions in contrast to traditional financial factors. Chapter 2: The PC video gaming industry, currently valued at over \$30 billion, features customers whose demand decisions are motivated by a variety of factors including price, perceived quality, and software attributes. Gamers are highly social and sensitive to the opinions of reviewers, peers in the gaming community, and most notably their own immediate social groups. Using novel and richly descriptive customer-level data from Steam, a leading player in the digital distribution space, this paper presents demand estimates for over 10,000 distinct game titles throughout a socially connected network of nearly 100,000 customers. I augment traditional demand models with descriptive measures of socialization and peer influence. The results suggest that peer effects are strongly influential for demand outcomes after controlling for price changes and product characteristics. Chapter 3: Recent empirical work suggests that the investment decisions made by household investors are influenced by the opinions of their peers. This paper explores whether social learning can generate beneficial externalities through aggregated crowd wisdom or whether it promotes potentially harmful herding behavior. Using individual investment data from the Motley Fool CAPS platform, I study whether individual predictions can be aggregated to improve the predictability of future asset returns. I provide formal tests of Granger causality and incorporate sentiment indices as covariates into auto-regressive conditional heteroskedasticity (GARCH) models of expected return and volatility. Portfolio simulations demonstrate that, under some circumstances, aggregate sentiment does improve predictability of asset returns in contrast to a more myopic model. Lastly, I quantify the excess portfolio returns provided using a sentiment-aware investment strategy after controlling for common market risk factors.
Collections
- Economics [179]