Text-Based Factor Models
Lin William Cong, Assistant Professor of Finance
We aim to construct an econometric framework to extract information and systematic factors from various sources of textual data, exploiting recent development in machine learning such as deep learning and topic modeling in analyzing unstructured data. In one application, we hope to contract and combine the new text-based factors with non-text based factor models of Fama and French (1992, 1993), Carhart (1997), and Fama and French (2016), and further to investigate the statistical power and economic forces captured by the new textual factors. In another application, we hope to analyze how FOMC announcements and meeting discussions translate into market reactions.