Can FAVAR improve Swedish inflation forecasting?

Abstract: The purpose of this thesis is to investigate whether factor augmented vectorautoregression (FAVAR) models estimated using principal component analysis are able to improve monthly inflation rate forecasts for Sweden. We produce 42 forecasts for the period January 2012 to June 2015 and evaluate the forecasts by their root mean square errors as well as their ability to correctly predict the sign of the inflation rate. The models forecasting performances are compared using the Diebold-Mariano test of equal predictive accuracy. Our results show that the investigated FAVAR models cannot significantly improve forecasts relative to a univariate model and that the FAVAR models perform worse with twelve lags than with only one lag.