Leif Anders Thorsrud

NLP in Applied Macroeconomics and Forecasting

Abstract

Expectations, and expectation formation, play a crucial role in all close to all economic models. Despite this, text, narratives, and written news, have been used relatively little when modelling agents’ information set, i.e., what they condition on when forming expectations. In this presentation, I present evidence showing that news coverage and stories; i) Contain high frequency and narrative sound information about important macroeconomic variables such as inflation, consumption, and the business cycle; ii) Potentially play an important role in terms of understanding information rigidities; iii) Allow us to measure important concepts, such as climate change transition risk, to a larger extent than what is possible using more traditional data.