nep-for New Economics Papers
on Forecasting
Issue of 2023‒02‒27
two papers chosen by
Rob J Hyndman
Monash University

  1. ‘Seeing’ the Future: Improving Macroeconomic Forecasts with Spatial Data Using Recurrent Convolutional Neural Networks By Jonathan Leslie
  2. Forecasting Value-at-Risk using deep neural network quantile regression By Chronopoulos, Ilias; Raftapostolos, Aristeidis; Kapetanios, George

  1. By: Jonathan Leslie (Indiana University, Department of Economics)
    Abstract: I evaluate whether incorporating sub-national trends improves macroeconomic forecasting accuracy in a deep machine learning framework. Specifically, I adopt a computer vision setting by transforming U.S. economic data into a ‘video’ series of geographic ‘images’ and utilizing a recurrent convolutional neural network to extract spatio-temporal features. This spatial forecasting model outperforms equivalent methods based on country-level data and achieves a 0.14 percentage point average error when forecasting out-of-sample monthly percentage changes in real GDP over a twelve-month horizon. The estimated model focuses on Middle America in particular when making its predictions: providing insight into the benefit of employing spatial data.
    Keywords: Macroeconomic Forecasting, Machine Learning, Deep Learning, Computer Vision, Economic Geography
    Date: 2023–02
  2. By: Chronopoulos, Ilias; Raftapostolos, Aristeidis; Kapetanios, George
    Abstract: In this paper we use a deep quantile estimator, based on neural networks and their universal approximation property to examine a non-linear association between the conditional quantiles of a dependent variable and predictors. This methodology is versatile and allows both the use of different penalty functions, as well as high dimensional covariates. We present a Monte Carlo exercise where we examine the finite sample properties of the deep quantile estimator and show that it delivers good finite sample performance. We use the deep quantile estimator to forecast Value-at-Risk and find significant gains over linear quantile regression alternatives and other models, which are supported by various testing schemes. Further, we consider also an alternative architecture that allows the use of mixed frequency data in neural networks. This paper also contributes to the interpretability of neural networks output by making comparisons between the commonly used SHAP values and an alternative method based on partial derivatives.
    Keywords: Quantile regression, machine learning, neural networks, value-at-risk, forecasting
    Date: 2023–02–07

This nep-for issue is ©2023 by Rob J Hyndman. It is provided as is without any express or implied warranty. It may be freely redistributed in whole or in part for any purpose. If distributed in part, please include this notice.
General information on the NEP project can be found at For comments please write to the director of NEP, Marco Novarese at <>. Put “NEP” in the subject, otherwise your mail may be rejected.
NEP’s infrastructure is sponsored by the School of Economics and Finance of Massey University in New Zealand.