Does Twitter Strengthen Volatility Forecasts? Evidence from the S&P 500, DJIA and Twitter Sentiment Analysis

Abstract

The role of public sentiment in stock market volatility has recently become increasingly relevant. Twitter, in theory, offers an inexpensive way to measure real-time public sentiment. We take advantage of a natural experiment to assess the potential improvement that social media adds to forecast performance of ARIMA and ARFIMA models of realized volatility using E-mini S&P 500 (ES) and E-mini DJIA (YM) futures contracts. Comparing models over time, we find that accounting for Twitter sentiment strengthens out-of-sample volatility forecasts across all time periods. While statistical significance exists, economic significance is harder to quantify, and it is unclear if a first-mover advantage exists from continuously monitoring real time Twitter sentiment.