Financial Conditions and Economic Activity in Brazil: Supervised Machine Learning

Financial Conditions and Economic Activity in Brazil: Supervised Machine Learning

• Autor
• FRANCISCO ALVES DE OLIVEIRA FILHO
• Co-autores
• Rogério Sobreira , Allisson David de Oliveira Martins , Marcos Falcão Gonçalves , Sávio Medeiro Viana
• Resumo

• This paper constructs a Financial Conditions Index (FCI) for Brazil using a supervised factor approach and evaluates its predictive power for economic activity. The index is built using grouped Principal Component Analysis and economically informed weights derived from predictive regressions of future economic growth. Monthly data from March 2014 to December 2024 are employed. The predictive performance of the FCI is evaluated using Random Forest and XGBoost algorithms under a rolling-origin out-of-sample framework. Results indicate that financial conditions contain persistent and economically meaningful information about future economic activity, particularly for the IBC-Br index. Random Forest systematically outperforms XGBoost across forecast horizons. The findings highlight the importance of non-linear transmission channels and reinforce the relevance of composite financial indicators for macroeconomic monitoring in emerging markets.

• Palavras-chave
• Financial Conditions Index; Economic Forecasting; Machine Learning; Brazil; Nonlinear Dynamics; Growth-at-Risk
• Modalidade
• Comunicação oral
• Área Temática
• Macrofinanças

Voltar Download