[Back]

E. Bura, A. Barbarino:
"Near-equivalence of Dimension Reduction Methods in Large Panels of Macro-variables";
Talk: RSS 2018 Conference, Cardiff; 2018-09-03 - 2018-09-06.

In an extensive pseudo out-of-sample horserace, classical estimators (dynamic factor models, RIDGE and partial least squares regression) and the novel to forecasting Sliced Inverse Regression and Recurrent Neural Networks exhibit almost near-equivalent forecasting accuracy in a large panel of macroeconomic variables across targets, horizons and subsamples. This finding motivates our theoretical contributions in this paper.
We show that most widely used linear dimension reduction methods solve closely related maximization problems with solutions that can be decomposed in signal and scaling components. We organize them under a common scheme that sheds light on their commonalities and differences as well as on their functionality.