| |
| author | Richa Patel |
| title | Twitter Sentiment Implications on the Russell 1000 Universe |
| abstract | The purpose of this thesis is to evaluate the effectiveness of using Twitter sentiment as a factor to inform
portfolio constructing decisions. This approach was used instead of a standard linear regression because it is widely used in finance today to look
for a subtle signal in very noisy data. Moreover, due to the excessive noise in Twitter data, the nature of this thesis for the most part is
exploratory and tests only market neutral trading strategies constructed by stocks from the Russell 1000 universe. In my results, I find that
the overall strength of the sentiment signal in predicting the direction of returns of the stocks in the Russell 1000 is weak and decreasing
over time. Though some of my results suggest that the strength is likely non-zero. I find that under certain restrictions without transaction
costs, using Twitter sentiment to inform stock decisions does have some statistical significance at the 5% level. Incorporating a sector-neutral
portfolio helps to further improve results for the signal, but incorporating a beta-neutral portfolio in addition to the sector-neutral portfolio
does not improve the results dramatically. I also find that across all portfolios, there were large differences in the individual sector performance,
but the best performing year remained unchanged at 2013. |
| school | The College of Liberal Arts, Drew University |
| degree | B.A. (2017) |
| advisors | Dr. Giandomenico Sarolli
Dr. Steve Surace |
| committee | Dr. Giandomenico Sarolli
Dr. Jon Kettenring
Joseph Noto |
| full text | RPatel.pdf |
| |
