Ensaios em macroeconomia: previsão das finanças públicas e Incerteza

• Autor(a) principal: Pereira, Breno da Silva Araujo
• Data de Publicação: 2023
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UFPB
• Texto Completo: https://repositorio.ufpb.br/jspui/handle/123456789/30083
• topic: Finanças públicas; Machine learning; Economia - Brasil - Índice de incerteza; Atividade econômica - Redução - Brasil; Inflação - Brasil; Desemprego - Brasil; Public Finances; Economy - Brazil - Uncertainty index; Economic activity - Reduction - Brazil; Inflation - Brazil; Unemployment - Brazil; CNPQ::CIENCIAS SOCIAIS APLICADAS::ECONOMIA

This article analyzes the performance of univariate models to predict five Brazilian macroeconomic variables related to public finances: Federal Revenue, Net Revenue, Total Expenditure, Primary Result and Nominal Result. In addition to the Arima model, considered as the benchmark method of this research, 6 more models are applied - Prophet, XGBoost, Arima-XGboost, Prophet-XGBoost and two combinations of the forecasts - in order to identify methods that can improve the predictions of the aforementioned variables. The results indicate that for the first three variables, the Prophet-XGBoost method and the two combinations of forecasts are competing models to the Arima method, presenting more accurate results in several cases. For the other variables, the Arima method is more accurate, especially in longer time horizons.