Modeling Multivariate Time Series in Economics: from Auto-Regressions to Recurrent Neural Networks

28 Pages Posted: 20 Dec 2018

Multiple version iconThere are 4 versions of this paper

Date Written: November 30, 2018

Abstract

Modeling of multivariate time series in an agnostic manner without assumptions about underlying theoretical structure is traditionally conducted using Vector Auto-Regressions. They are well suited for linear and state-independent evolution. A more general methodology of Multivariate Recurrent Neural Networks allows to capture non-linear and state-dependent dynamics. This paper takes a range of small- to large-scale Long Short-Term Memory MRNNs and pits them against VARs in an application to US data on GDP growth, inflation, commodity prices, Fed Funds rate and bank reserves. In our exercise, MRNN outperforms VAR in forecasting: its out-of-sample predictions are about 20% more accurate. MRNN also fares better in interpretability by means of impulse response functions: for instance, a temporary shock to the Fed Funds rate variable generates system dynamics that are more plausible according to conventional economic theory. Lastly, it is reassuring to see good performance of an RNN in a small-sample regime (but our preferred specification is fairly minimalistic nevertheless).

Suggested Citation

Verstyuk, Sergiy, Modeling Multivariate Time Series in Economics: from Auto-Regressions to Recurrent Neural Networks (November 30, 2018). Available at SSRN: https://ssrn.com/abstract=3294796 or http://dx.doi.org/10.2139/ssrn.3294796

Sergiy Verstyuk (Contact Author)

Harvard University ( email )

1875 Cambridge Street
Cambridge, MA 02138
United States

Do you have negative results from your research you’d like to share?

Paper statistics

Downloads
56
Abstract Views
417
Rank
79,979
PlumX Metrics