An Assessment of the Marginal Predictive Content of Business Conditions and Economic Uncertainty Indexes

In this paper, we evaluate the marginal predictive content of a variety of new business conditions and economic uncertainty indexes. Our indexes are defined as latent factors extracted from a high dimensional macroeconomic dataset (business conditions indexes) and as functions of predictive errors from models that incorporate these latent factors (economic uncertainty indexes). Estimation of the indexes is based on a number of extant and novel machine learning methods that combine variable selection and shrinkage methods including principal components analysis and the least absolute shrinkage operator. When predicting 14 monthly series selected from 8 different groups of economic variables, our new indexes are shown to result in significant improvements in predictive accuracy, relative to predictions made using benchmark models. Moreover, while inclusion of either business conditions indexes or economic uncertainty indexes yields forecast accuracy improvements, greater predictive gains accrue when both types of indexes are included in predictive regressions. Additionally, our business conditions indexes are found to be particularly useful for short-term forecasting, while our economic uncertainty indexes exhibit their best predictive content for long-term forecasting. Finally, our mean-square forecast error ``best'' indexes are almost always constructed by utilizing a novel factor-lasso estimation method that accounts for the high level of collinearity across the variables in our macroeconomic dataset