Making text count: Economic forecasting using newspaper text

Abstract

Summary This paper examines several ways to extract timely economic signals from newspaper text and shows that such information can materially improve forecasts of macroeconomic variables including GDP, inflation and unemployment. Our text is drawn from three popular UK newspapers that collectively represent UK newspaper readership in terms of political perspective and editorial style. Exploiting newspaper text can improve economic forecasts both unconditionally and when conditioning on other relevant information, but the performance of the latter varies according to the method used. Incorporating text into forecasts by combining counts of terms with supervised machine learning delivers the highest forecast improvements relative to existing text‐based methods. These improvements are most pronounced during periods of economic stress when, arguably, forecasts matter most.

The power of text-based indicators in forecasting Italian economic activity

“…Sentiment indicators and topic analysis are becoming popular tools for forecasting and structural analysis. Newspapers-based indicators can be a timely source of information to track the business cycle, as already documented by recent papers (Thorsrud, 2018;Kalamara et al, 2020;Bybee et al, 2019;Shapiro et al, forthcoming 2020;Rambaccussing and Kwiatkowski, 2020).…”

“…Our database is extracted from Dow Jones Factiva one of the largest archives used in the forecasting literature using textual based indicators (Thorsrud, 2018;Fraiberger, 2016;Bybee et al, 2019;Kelly et al, 2018;Shapiro et al, forthcoming 2020). Kalamara et al (2020) have recently analyzed around half a million articles from three main British newspapers showing that simple text-based indicators can improve economic forecasts, in particular during downturns. Rogers and Xu (2019) have also recently analyzed the predictive performance of different measures of economic uncertainty in the US to forecast real and financial outcome variables, showing some additional predictive content of such measures.…”

Text-Based Recession Probabilities

“…Ardia et al (2019) use machine learning techniques to build a weighted index based on topical analysis-based sentiment subindices, that is used to forecast US industrial production. Kalamara et al (2020) also apply machine learning tools, combined with data extracted from major UK newspapers, to improve the prediction of an empirical business cycle model. Shapiro et al (2020) and Burri and Kaufmann (2020) have also recently proposed sentiment-scoring model based on lexicons, for the USA and the Swiss economies, that improve the accuracy of real activity forecast models.…”

Fine-Grained Implicit Sentiment in Financial News: Uncovering Hidden Bulls and Bears

“…Due to the direct economic incentives and value of information acquisition in financial markets, sentiment analysis has attracted a vast amount attention in the field of economics and finance. SA has been applied in various applications such as stock prediction [56][57][58][59][60][61], financial market analysis [62], impression management of brands or people [2,[63][64][65], macro-economic policy metrics [66,67], and forecasting macro-economic trends and risk [68][69][70]. Many studies focused on social media text, especially financial microblog StockTwits and tweets [18,[71][72][73].…”