Forecasting Time Series with a Mixture of Stationary and Nonstationary Factors

Accurate forecasting of macroeconomic variables, such as economic growth and inflation, is central to making economic policy decisions. This thesis contributes to the literature by developing improved methods for forecasting such variables. The central theme of this thesis is to extend the method based on the Factor Augmented Regression model. The new methods developed in the thesis may use a mixture of stationary and nonstationary unobserved factors and observed variables as predictors, time-varying parameters, and two-level factors. These advances constitute a significant contribution to improve econometric methods in this area.