Applied Dynamic Factor Analysis for Macroeconomic Forecasting

The use of dynamic factor analysis in statistical modeling has broad utility across an array of applications. This paper presents a novel hierachical structure suited to a particular class of predictive problems - those which necessitate the aggregation of numerous forecasts in the presence of substantial data missingness and a need for systematic dimensionality reduction. The model hierarchy is presented in the context of the prediction of U.S. Nonfarm Payrolls, a well-known economic statistic, though can be generally applied for any context exhibiting an analagous data structure.