Coelho, Pedro Miguel Pereira SimõesDamásio, Bruno Miguel PintoMiguel, Diogo Queiroz2023-05-032023-05-032023-04-14http://hdl.handle.net/10362/152363Dissertation presented as the partial requirement for obtaining a Master's degree in Information Management, specialization in Knowledge Management and Business IntelligenceWe analyze the possibility of Vector Autoregressive models being good estimators for the unemployment rate in Portugal, by studying their ability to understand the impact the COVID-19 pandemic had on the unemployment rate. We make use of Bayesen Stochastic Search Variable Selection and bootstrapping techniques for forecasting, comparing the results of these models with two benchmark techniques, ARIMA and Artificial Neural Networks. The model performance is tested through the RMSE, MSE and MAE of the estimations, and we compare the forecasting quality through a Diebold-Mariano test. We conclude that the VAR methodology can provide better forecasts than the benchmark models when combined with the Bayesian approach, both for shorter and longer forecasting horizons. We also conclude that COVID-19 did not provide the expected shock to the Portuguese unemployment rate.engCOVID-19unemploymentvector autoregressive modelsforecastingmaster thesis203275667