Google Trends para previsão de variáveis macro: uso no Brasil através do algoritmo autometrics

• Ano de defesa: 2017
• Autor(a) principal: Guimarães Filho, Samuel
• Orientador(a): Marçal, Emerson Fernandes
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Mercado de trabalho; Seleção automática de modelos; CAGED
• Palavras-chave em Inglês: Google Trends; Model confidence set; Labour market; Formal job creation; Automated model selection; Autometrics
• Link de acesso: http://hdl.handle.net/10438/18025

This work aims to test if the use of Google Trends as an exogenous variable improves the prediction of the monthly data for Brazilian Formal Job Creation (CAGED) compared to a model that uses only the lags themselves. For the selection of the model was used the algorithm Autometrics and for model comparison the Model Confidence Set. In addition, the model that uses Google Trends data will be compared with some market analyst’s forecasts. The results show that the model the uses the Google data as an exogenous variable is superior to the model that only uses the lag itself. However, this model was not able to overcome the market analysts.