Title: Time series analysis using the radial basis functions: Application to the US economy
Authors: Nobuyuki Kanazawa - Hitotsubashi University (Japan) [presenting]
Abstract: A flexible nonlinear method is proposed for studying the time series properties of macroeconomic variables. We focus on a class of Artificial Neural Networks (ANN) called the Radial Basis Functions (RBF), which is capable of producing history- and shock-dependent impulse responses without imposing a strong functional form assumption. To assess the validity of the RBF approach in a macroeconomics time series analysis, we conduct a series of Monte Carlo experiments using a data generated from a nonlinear New-Keynesian (NK) DSGE model. We find that the RBF time series can uncover the nonlinear NK economic structure from the simulated data whose length is as small as 300 (quarters). Finally, we apply this RBF time series method to the US macroeconomic data from 1948-2015 and show that the response of real GDP growth to utilization-adjusted TFP shocks is negative after 2008 when the nominal interest rates hit zero. The finding is consistent with the prediction of the New Keynesian model.