This paper develops three text-based indicators of financial stress for Brazil using Google Trends search data and financial news from major Brazilian outlets between 2008 and 2025. The first indicator is based on a curated dictionary of Portuguese-language search terms, the second uses a machine-learning selection procedure, and the third is built from 33,881 news articles. We assess their empirical relevance for Brazil’s 5-year sovereign CDS spread using local projections, Granger-causality tests, and regime-switching analysis. The dictionary-based index shows the most stable association with subsequent CDS widening. The news-based measure also contains useful information, especially in the Granger exercises, while the machine-learning specification is less consistent in this application. The results suggest that search-based and news-based indicators can complement conventional tools for monitoring sovereign financial stress in Brazil.
Comissão Organizadora
Comissão Científica
Ferramenta completa de submissão, avaliação e publicação de anais para eventos científicos.
