Abstract

I investigate the impact of unscheduled, firm-specific news on high frequency stock market reactions using a five-year sample of intraday news releases and their corresponding Sentiment, Relevance, and Novelty scores generated by the Thomson Reuters News Analytics (TRNA) algorithm. My analysis features one-second interval price and volume data as well as matching trade and quote (TAQ) data for 55 Nasdaq stocks listed on the index between 2011 and 2015 inclusively. I examine cumulative abnormal returns, volumes, and trades, and further employ a quantile regression model that includes measures of news traffic to determine whether machine- readable news can effectively flag short-term trading opportunities. In line with related studies, I find significant increases in abnormal trading activity in the first few minutes surrounding a news release, with volume proving to be more sensitive to the TRNA metrics than returns. Positive news is traded more aggressively than negative news on the knee-jerk, but also experiences sharper reversals in abnormal returns in the hours following. Furthermore, results from the quantile regression analysis appear to confirm that news traffic in the run-up to and at the release time significantly impact abnormal returns. Although my results appear consistent with the hypothesis that trading activity should increase as TRNA thresholds become stricter, simulated holding period returns remain negative, highlighting the many complexities involved in algorithmically trading the news.