Forecasting Macroeconomic Time Series: LASSO-Based Approaches and Their Forecast Combinations with Dynamic Factor Models

Abstract

In a data-rich environment, forecasting economic variables amounts to extracting and organizing useful information out of a large number of predictors. So far dynamic factor model and its variants have been the most successful models for such exercises. In this paper, we investigate a category of LASSO-based approaches and evaluate their predictive abilities in forecasting twenty important macroeconomic variables. These alternative models could handle hundreds of data series simultaneously, and extract useful information for forecasting. We also show analytically and empirically that combing forecasts from LASSO-based models and those from dynamic factor models could further reduce the mean square forecast error (MSFE). Our three main findings can be summarized as follows. First, for most of the variables under investigation, all LASSO-based models outperform dynamic factor models in the out-of-sample forecast evaluations. Second, by extracting information and formulating predictors at the economically meaningful block levels, new methods greatly enhance model interpretabilities. Third, once forecasts from a LASSO-based approach and those from a dynamic factor model are combined by forecasts combination techniques, the combined forecasts are significantly better than dynamic factor model forecasts and the naive random walk benchmark.