Title: Identifying firm-level news shocks from financial news media
Authors: Julian Ashwin - University of Oxford (United Kingdom) [presenting]
Abstract: A text mining approach is investigated which can be used to identify firm level news shocks from news media, and whether these are of macroeconomic relevance. We use a variety of approaches, including Named-entity recognition, to match articles from the Financial Times newspaper to firms listed on the London Stock Exchange. We find that being mentioned in that day's edition has a robust, statistically significant and substantial effect on both absolute return and trading volume of an individual firm's stock price. Both supervised and unsupervised topic modelling approaches are used to extract richer information from news media which can predict firm level stock returns and separate ``good'' and ``bad'' news. We argue that different topics plausibly identify not only shocks to sentiment, but also new information about a firm's future profitability. This also allows the comparison of the effects of different types of news on an individual firm's stock price. These identified news shocks are then used to investigate how new information propagates across the stock market, and whether this process is related to the structure of the production network. This evidence is used to assess the plausibility of the claim that firm or sector specific news shocks could have effects similar to those of a negative technology shock by resulting in a misallocation of production factors.