Semiparametric Threshold Cointegration with Endogenous Regime Switching: A Penalized Spline Augmented Error-Correction Framework for Nonlinear Macroeconomic Adjustment

Abstract

This study develops a semiparametric threshold error-correction model (STECM) that integrates penalized spline smoothing with endogenously determined regime-switching boundaries to capture asymmetric long-run adjustment dynamics in macroeconomic panel data. Conventional linear cointegration frameworks systematically misattribute regime-dependent heteroscedasticity and structural instability, yielding inconsistent parameter estimates under nonlinear equilibrium reversion. Employing a generalized cross-validation bandwidth selector and a sup-Wald threshold identification procedure, the proposed estimator achieves uniform consistency under mild mixing conditions. Monte Carlo simulations confirm substantial finite-sample efficiency gains over Hansen–Seo and Enders–Granger benchmarks. Empirical application to OECD current-account–output-gap nexuses (2000–2023) reveals significant asymmetric half-life differentials across expansion and contraction regimes, with meaningful implications for fiscal consolidation policy design and external-balance surveillance mechanisms.