The goal of this paper is to evaluate the informational content of sentiment extracted from news articles about the state of the economy. We propose a fine-grained aspect-based sentiment analysis that has two main characteristics: 1) we consider only the text in the article that is semantically dependent on a term of interest (aspect-based) and, 2) assign a sentiment score to each word based on a dictionary that we develop for applications in economics and finance (fine-grained). Our data set includes six large US newspapers, for a total of over 6.6 million articles and 4.2 billion words. Our findings suggest that several measures of economic sentiment track closely business cycle fluctuations and that they are relevant predictors for four major macroeconomic variables. We find that there are significant improvements in forecasting when sentiment is considered along with macroeconomic factors. In addition, we also find that sentiment matters to explains the tails of the probability distribution across several macroeconomic variables.
We use a dataset of 12 million residential mortgages to investigate the loan default behavior in several European countries. We model the default occurrence as a function of borrower characteristics, loan-specific variables, and local economic conditions. We compare the performance of a set of machine learning algorithms relative to the logistic regression, finding that they perform significantly better in providing predictions. The most important variables in explaining loan default are the interest rate and the local economic characteristics. The existence of relevant geographical heterogeneity in the variable importance points at the need for regionally tailored risk-assessment policies in Europe.
4 November 2022 in International Journal of Forecasting
During the Covid-19 pandemic, economists have struggled to obtain reliable economic predictions, with standard models becoming outdated and their forecasting performance deteriorating rapidly. This paper presents two novelties that could be adopted by forecasting institutions in unconventional times. The first innovation is the construction of an extensive data set for macroeconomic forecasting in Europe. We collect more than a thousand time series from conventional and unconventional sources, complementing traditional macroeconomic variables with timely big data indicators and assessing their added value at nowcasting. The second novelty consists of a methodology to merge an enormous amount of non-encompassing data with a large battery of classical and more sophisticated forecasting methods in a seamlessly dynamic Bayesian framework. Specifically, we introduce an innovative “selection prior” that is used not as a way to influence model outcomes, but as a selection device among competing models. By applying this methodology to the Covid-19 crisis, we show which variables are good predictors for nowcasting gross domestic product and draw lessons for dealing with possible future crises.