We analyze methods for selecting topics in news articles to explain stock returns. We find, through empirical and theoretical results, that supervised Latent Dirichlet Allocation (sLDA) implemented through Gibbs sampling in a stochastic EM algorithm will often overfit returns to the detriment of...

I examine how financial markets interact with news about the COVID-19 pandemic. A twelve topic model optimizes the trade-off between number of topics and topic coherence. Using this model, I show that before mid-March 2020 markets react more to the same quantum of news when volatility is higher...

We find that an increase in the ``unusualness'' of news with negative sentiment predicts an increase in stock market volatility. Similarly, unusual positive news forecasts lower volatility. Our analysis is based on more than 360,000 articles on 50 large financial companies, published in...

We introduce FDIF, a measure of Fed communication surprise based on the text of FOMC statements. FDIF measures the difference between text-implied and actual values of key market variables. Positive FDIF of countercyclical variables (e.g., credit spreads) is associated with negative...

We introduce FDIF, a measure of Fed communication surprise based on the text of FOMC statements. FDIF measures the difference between text-implied and actual values of key market variables. Positive FDIF of countercyclical variables (e.g., credit spreads) is associated with negative...