Words as data: quantifying Brazilian Central Bank communication

Imagem de Miniatura
Ribeiro, Marcel Bertini
Título da Revista
ISSN da Revista
Título de Volume

Ch 1: This research assesses the impact of Central Bank Communication to the term-structure of interest rates in the Brazilian Economy. It quantifies statements and minutes released by the monetary authority by employing Latent Dirichlet Allocation and Dictionary methods to the written content of these documents. Our results indicate that the term-structure of interest rates respond to the Brazilian Central Bank communication, however differently: statements seem to reach short-to-medium maturity yields while minutes also reach longer maturities; responses to statements may be as large as 8 basis points, while to minutes as large as 15 basis points. Also, the market response to communication has varied over Central Bank presidencies. Ch 2: This research employs Latent Dirichlet Allocation and Dictionary methods to construct communication series from the Central Bank of Brazil minutes. These communication series are incorporated into a Structural Vector Autoregression (SVAR) system to improve its identification and therefore contribute to the analysis of the relationship between macroeconomic variables following a monetary policy shock. Our estimations suggest that our framework mitigates the price puzzle observed in the baseline specification without the inclusion of communication series. Furthermore, our findings demonstrate robustness in identifying monetary shocks through the use of the Structural Vector Autoregression with an external instrument (SVAR-IV).

Área do Conhecimento