Social Learning and Sentiment Contagion in the Bitcoin Market

Abstract

Using novel data on investors' social interactions, we document sentiment contagion as a direct consequence of social learning in the Bitcoin market. Our findings suggest that the social learning process is inefficient, as investors respond positively to social sentiment, but social sentiment does not positively predict returns. Echo chamber effect and selective interpretation of information may simultaneously contribute to such inefficiency. We further confirm that sentiment contagion is not a sideshow: at the individual level, sentiment contagion predicts the direction of trading conditional on its occurrence; at the Bitcoin market level, the aggregated sentiment contagion index positively predicts market trading volume and volatility. Moreover, we highlight the error-prone nature of social learning, wherein the socially constructed optimism predicts Bitcoin crashes. Finally, we establish a link between social learning and bubbles: the elevated propagation of optimism is highly correlated with the trading volume.

Keywords: Social Finance, Sentiment Contagion, Bubbles, Bitcoin

JEL Classification: G11, G12, G41, G53