Bitcoin Returns – a Mood Killer?

Abstract

The thesis examines whether cryptocurrency returns may be used to explain sentiment expressed on social media. To do this, we have constructed a unique dataset consisting of posts on Twitter and Reddit mentioning Bitcoin and market data. The average sentiment scores on Twitter and Reddit are predicted with a linear regression model using Bitcoin returns. Our analysis finds that Twitter users react positively to price increases and negatively to price falls. However, the explanatory power of the model is somewhat limited. In addition, we discover disparities in predicting power within the sample. When examining the proportion of positive and negative posts, we find that price movements impact the proportion of negative posts more than positive. On the other hand, we find no meaningful relationship between Bitcoin prices and sentiment expressed on Reddit. This thesis explores whether differences in risk preferences between Twitter and Reddit users might explain these findings.