Two Essays On Econometric Forecasting With An Econometric Model PDF Download

This dissertation comprises two essays in macroeconomic forecasting. The first essay empirically examines approaches to combining factor models and robust estimation, and presents the results of a "horse-race" in which mean-square-forecast-error (MSFE) "best" models are selected, in the context of a variety of forecast horizons, estimation window schemes and sample periods. For the majority of the target variables that we forecast, it is found that various of these shrinkage methods, when combined with simple factors formed using principal component analysis (e.g. component-wise boosting), perform better than all other models. It is also found that model averaging methods perform surprisingly poorly, given our prior that they would "win" in most cases. The second essays outlines and discusses a number of interesting new forecasting methods that have recently been developed in the statistics and econometrics literature. It focuses in particular on the examination of a variety of factor modeling methods, including principal components, independent component analysis (ICA) and sparse principal component analysis (SPCA). Further, it outlines a number of approaches for creating hybrid forecasting models that use these factor modeling approaches in conjunction with various type of shrinkage methods. The results show that pure factor modeling approaches alone are not enough to lead to our overall finding that simple linear econometric models as well as models based on various forecast combination strategies are dominated by more complicated (factor/shrinkage) type models.