Ensaios em macroeconomia: previsão das finanças públicas e Incerteza

• Ano de defesa: 2023
• Autor(a) principal: Pereira, Breno da Silva Araujo
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal da Paraíba; Brasil; Economia; Programa de Pós-Graduação em Economia; UFPB
• Palavras-chave em Português: 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
• Link de acesso: https://repositorio.ufpb.br/jspui/handle/123456789/30083

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.