WallStreetBets on Wall Street An Empirical Analysis of the Market Power of WallStreetBets

Abstract

In this thesis, we apply sentiment analysis techniques to test whether sentiment on WallStreetBets has had an impact on stock returns, trading volume, option volume, and implied volatility from January 01, 2020, to March 15, 2021. We analyze each submission and comment posted on WallStreetBets during this time interval that can be linked to discussion of a selected sample of stocks, and apply sentiment analysis techniques to identify whether each post displays positive, neutral, or negative sentiment. We then analyze stocks on an individual and aggregated basis to test the following hypotheses: whether sentiment on WallStreetBets has had an impact on (i) stock returns; (ii) stock volume; and (iii) option volumes and implied volatility. First, there are large variations in the results for sentiment’s impact on return on an individual basis, and while reverse causality can be attributed to explain much of the results we observe for some individual stocks, we find indicative evidence of WallStreetBets sentiment having had a statistically significant impact on the return of other stocks. On an aggregated basis, sentiment is shown to explain returns better the day after sentiment is recorded, suggesting an ability to influence future stock returns. Second, by looking at sentiment against volume we find a statistically significant relationship on most stocks in our sample, suggesting forum sentiment drives stock activity. This relationship on an aggregated basis is stronger without lagged effects, meaning same-day sentiment drives stock volumes. Finally, we find the strongest relationship in our study when looking at option-related metrics, showing a clear effect on both call and put volume as well as implied volatility both on an individual and aggregated basis. The results from our minute-by-minute model during the January 2021 rallies suggest that forum activity was a statistically significant driving force behind volume in the affected stocks. However, on the same data we could not find a statistical relationship on return, suggesting there were other influences behind the price increases than comments on the forum alone. We also develop trading strategies based on sentiment on WallStreetBets, and find that these would have yielded remarkable returns in the time interval we explore.