Predicting digital currency market with social data: Implications of network structure and incentive hierarchy

Abstract

Following the recent discovery of social media’s predictive power for the financial markets, we try to advance the literature by finding ways to distinguish between value-relevant information and noises by investigating the implication of social media network structures and its incentive hierarchy system. In the first chapter, we use data from the Bitcoin market and empirically show that highly-connected social media discussion networks are less accurate in predicting future returns due to information free riding and highly correlated information. However, social media information linkages nevertheless serve as landmarks for identifying informed social media actors: value-relevant information is more likely to be shared by users who stimulate active discussions among their peers. In the second chapter, we examine how social media incentive hierarchy systems shape users’ posting motivation and thus influence the information quality. We find that on average, the active social media users holding higher badges provide less accurate prediction compared to inactive users with lower badges. The reduced motivations after obtaining the higher level badges and the more frequent use of social media for socialization purposes imply a higher proportion of noises in their posts.