Replacing ARDL? Introducing the NSB-ARDL Model for Structural and Asymmetric Forecasting

This paper introduces the NSB-ARDL (Nonlinear Structural Break Autoregressive Distributed Lag) model, a novel econometric framework designed to capture asymmetric and nonlinear dynamics in macroeconomic time series. Traditional ARDL models, while widely used for estimating short- and long-run relationships, rely on assumptions of linearity and symmetry that may overlook critical structural features in real-world data. The NSB-ARDL model overcomes these limitations by decomposing explanatory variables into cumulative positive and negative partial sums, enabling the identification of both short- and long-term asymmetries. Monte Carlo simulations show that NSB-ARDL consistently outperforms conventional ARDL models in terms of forecasting accuracy when asymmetric responses are present in the data-generating process. An empirical application to South Korea's CO2 emissions demonstrates the model's practical advantages, yielding a better in-sample fit and more interpretable long-run coefficients. These findings highlight the NSB-ARDL model as a structurally robust and forecasting-efficient alternative for analyzing nonlinear macroeconomic phenomena.