O presente estudo consiste em construir um índice de sentimento fiscal (FSI-BR) a partir de notícias diárias de jornais e agências de notícias, aplicando processamento de linguagem natural, e investigar os impactos desse sentimento no comportamento da estrutura a termo de juros brasileira. O trabalho organiza-se em dois capítulos: “Building a Fiscal Sentiment Index for Brazil: Insights and Implications” e “The Impact of Fiscal Sentiment on Brazil's Term Structure of Interest Rates: A Dynamic Latent Factor Approach Using Fiscal Sentiment”. Os resultados são promissores, tanto em relação à correlação do índice de sentimento fiscal com marcos históricos quanto a outros índices de risco e preços de ativos. Ademais, o modelo dinâmico de Nelson–Siegel com expectativas macroeconômicas e índice de sentimento fiscal adere bem à curva de juros, implicando fatores de nível, inclinação e curvatura que respondem de forma economicamente significativa a choques no sentimento fiscal.

This study constructs a Fiscal Sentiment Index (FSI-BR) from daily newspaper and newswire articles using natural language processing and examines how this sentiment affects Brazil’s term structure of interest rates. It includes two chapters: “Building a Fiscal Sentiment Index for Brazil: Insights and Implications” and “The Impact of Fiscal Sentiment on Brazil’s Term Structure of Interest Rates: A Dynamic Latent Factor Approach Using Fiscal Sentiment.” The results are promising: the FSI-BR aligns closely with major historical events and comoves with standard risk indicators and asset prices. Even more compelling, a dynamic Nelson–Siegel model augmented with macroeconomic expectations and FSI closely tracks the observed yield curve, with level, slope, and curvature factors exhibiting economically meaningful responses to fiscal-sentiment shocks.