Using sentiment analysis on Reddit to predict stock returns

Abstract

This thesis explores if sentiment analysis can be utilized to predict meme stock returns by analyzing social media activity on the Reddit forum WallStreetBets. We further look at how meme stocks differ from non-meme stocks in their return predictability on this forum. To test this, we run OLS regressions and panel regressions on aggregated daily data collected from Reddit.

We find that the sentiment of WallStreetBets does have significant correlation with meme stock returns within our sample. Our results show significant differences between the examined meme stocks and non-meme stocks. Despite not being able to prove a causal relationship between the WallStreetBets activity and meme stock returns, we consider our findings promising for further research.