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

The 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. Even in a small-sample regime, MRNN significantly outperforms VAR in forecasting out-of-sample. MRNN also fares better in interpretability by means of impulse response functions: for instance, a shock to the Fed Funds rate variable generates system dynamics that are more plausible according to conventional economic theory. Additionally, the paper shows how, due to its inherent non-linearity, MRNN can discover (in an unsupervised manner) different macroeconomic regimes. Utilizing its state dependence, MRNN may also be a useful tool for policy simulations under practically relevant economic conditions (such as Zero Lower Bound)