Time-Varying Dictionary and the Predictive Power of FED Minutes
- PDF / 1,368,824 Bytes
- 33 Pages / 439.37 x 666.142 pts Page_size
- 5 Downloads / 145 Views
Time‑Varying Dictionary and the Predictive Power of FED Minutes Luiz Renato Lima1,2 · Lucas Lúcio Godeiro3 · Mohammed Mohsin1 Accepted: 10 August 2020 © Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract This paper proposes a method to extract the most predictive information from FED minutes that is specifically adapted to the problem of forecasting. Instead of considering a dictionary (set of words) with a fixed content, we construct a dictionary whose content is allowed to change over time. Specifically, we utilize machine learning to identify the most predictive words (the most predictive content) of a given minute and use them to derive new predictors. We show that the new predictors improve real time forecasts of output growth by a statistically significant margin, suggesting that the combination of supervised machine learning and text regression can be interpreted as a powerful device for out-of-sample macroeconomic forecasting. Keywords Text regression · Supervised machine learning · Elastic net · Central bank communication · Forecasting, real time JEL Classification C53 · C55 · E37 · E47
1 Introduction According to Gentzkow et al. (2017), the information encoded in text is a rich complement to the more structured kinds of data traditionally used in empirical research. Indeed, in recent years, we have seen an intense use of textual data in different areas of research. The idea consist of transforming strings into numeric variables, and then use it as predictors in different models. Several studies have already explored this additional source of information. In the finance literature, Garcia (2013) studies the effect of sentiment on asset prices during the twentieth century (1905 to 2005). * Luiz Renato Lima [email protected] 1
Department of Economics, The University of Tennessee, Knoxville, USA
2
Federal University of Paraiba, Joao Pessoa, PB, Brazil
3
Federal University of the Semi-Arid Region (UFERSA), Mossoro, RN, Brazil
13
Vol.:(0123456789)
L. R. Lima et al.
The author uses texts from the New York Times to construct sentiment variables using textual analysis. He concludes that the predictability of stock returns using news content is particularly strong during recessions. Engelberg and Parsons (2011) compare the behavior of investors with access to different media coverage of the same information event. They focus on the relation between local media coverage and local stock portfolio trading volume and find that local media coverage is a strong predictor of local trading returns. In the economics literature, Dossani (2018) analyzes how the tone of the Central Bank press conferences impacts risk premia in the currency market. He measures the tone as the difference between the number of hawkish and dovish phrases made during a press conference. He used four currency future contracts traded on the Chicago Mercantile Exchange (CME) and found that implied risk aversion increases when Central Banks are hawkish and decreases when Central Banks are dovish. Other examples
Data Loading...
