A Non-Standard Approach to the Calibration of Selected Dynamic Factor Models in Macroeconomic Forecasting

In this paper, we present a comparison of the forecasting performance of selected factor models on two large monthly data panels. The first dataset contains EU variables, whereas the other contains US variables. These data panels are split into two parts: the first subsample is used to select the most performing specification for each class of models in a in-sample environment, and the second subsample is used to compare the performances of the selected models in an out-of-sample environment. In the first subsample, genetic algorithms are employed to achieve an efficient exploration of the parameter space. We find that selected dynamic factor models are globally the most performing methods on the second subsamples of both data panels.