Trading on Social Impact

How can social media sentiment help us understand the future direction of corporate equity prices? The goal of this project was to find some signal from public sentiment from social media that can lead us to better explain some variance in stock price, and design a trading method which can profit from the results. More specifically, what might happen when that sentiment is related to a company’s social impact, or whether it is positively or negatively influencing the world. Generally speaking, there are many more factors than social media which contribute to a company’s performance, and an analysis interpreting these factors is still the goal of top hedge funds today. In the few months during which this project took place, the goal was to determine whether social media sentiment, namely from Twitter, would allow us to perform significantly better than if we had market data alone, and to determine whether certain companies’ stock prices are more or less affected by this sentiment. Some companies had equity prices which appeared to be much more correlated with the Twitter sentiment than others. These were usually consumer facing companies as well as chemical and energy companies: In the case of consumer facing companies, it is possible that negative sentiment would deter other consumers from using their products in the future. For energy and chemical companies, it seemed likely that negative sentiment, especially with regard to social impact, would be emphasized during such an accident as an oil or chemical spill. These events would also impact the stock price, explaining any correlation. For these most affected companies, a reinforcement learning framework was constructed and tested against a greedy trader which only had access to stock price. In only one case out of twelve did the greedy trader outperform the reinforcement trader, which trained on sentiment data as well as pricing data.