Economic growth dynamics: a machine learning-augmented nonlinear autoregressive distributed lag model of asymmetric effect

Breaking away from traditional methods, this study investigates the asymmetric effects of macroeconomic shocks on economic growth in Nigeria using a novel Machine Learning-Augmented Nonlinear Autoregressive Distributed Lag (NARDL) model. We utilized a Random Forest algorithm to data-driven feature selection, thus model optimization and its enhanced robustness against random selection of lags. The results confirm the presence of a long-run cointegrating relationship and show that positive and negative oil price shocks statistically differ in their effects on GDP. We also find inflation to have a strong negative long-run effect, and government capital expenditure is a significant driver of growth. Such embedding of machine learning in the NARDL model is a more empirically valid policy analysis tool that supplies key findings for policymakers in resource-dependent economies.