The Forecasting performance of the Factor model with Martingale Difference errors

TL;DR

This paper evaluates a new factor model with martingale difference errors (FMMDE), proposing a novel factor selection method and demonstrating its superior forecasting performance on US macroeconomic data.

Contribution

It introduces a new methodology for selecting the number of factors in FMMDE and compares its forecasting accuracy with other models using extensive empirical analysis.

Findings

FMMDE outperforms alternative models in predicting US macroeconomic variables.

The proposed factor selection method shows good finite-sample performance.

FMMDE provides advantages in forecasting key economic aggregates.

Abstract

This paper analyses the forecasting performance of a new class of factor models with martingale difference errors (FMMDE) recently introduced by Lee and Shao (2018). The FMMDE makes it possible to retrieve a transformation of the original series so that the resulting variables can be partitioned according to whether they are conditionally mean-independent with respect to past information. We contribute to the literature in two respects. First, we propose a novel methodology for selecting the number of factors in FMMDE. Through simulation experiments, we show the good performance of our approach for finite samples for various panel data specifications. Second, we compare the forecasting performance of FMMDE with alternative factor model specifications by conducting an extensive forecasting exercise using FRED-MD, a comprehensive monthly macroeconomic database for the US economy. Our empirical findings indicate that FMMDE provides an advantage in predicting the evolution of the real sector of the economy when the novel methodology for factor selection is adopted. These results are confirmed for key aggregates such as Production and Income, the Labor Market, and Consumption.