Finding Alphas, 2nd Edition

22The Impact of News and Social Media on Stock Returns

By Wancheng Zhang

INTRODUCTION

Stock prices naturally respond to news. But in recent years, news and sentiment seen on social media have grown increasingly significant as potential predictors of stock prices. However, it is challenging to make alphas using news. As unstructured data that often includes text and multimedia content, news cannot be understood directly by a computer. We can use natural language processing (NLP) and machine learning methods to classify and score raw news content, and we can measure additional properties of the news, such as novelty, relevance, and category, to better describe the sentiment of the news. Similar techniques can be applied to social media data to generate alphas, though we should bear in mind that social media has much more volume and is much noisier than conventional news media. This chapter gives an overview of ways to find alphas using news and social media.

NEWS IN ALPHAS

It is not easy for machines to accurately parse and interpret the meaning of news. As in other areas in statistical arbitrage, an algorithm has the advantages of fast response time and broad coverage, but at the expense of weaker accuracy than a human analyst. Nowadays, trading firms can analyze news within 1 millisecond and make trading decisions instantly. Big news usually causes large price movements instantly, often with a subsequent overshoot and reversal.

Since 2007, the application of sophisticated ...