Inflation Threshold for Vietnam: Evidence from Threshold Regression and NARDL with Quarterly Data

Abstract

Background. Most evidence on inflation–growth nonlinearity relies on cross-country panels or short samples, leaving a gap for a single transition economy on high-frequency data. We address this gap for Vietnam over 1996–2025. Methods. We combine threshold regression with the NARDL framework on quarterly data complemented by a linear ARDL benchmark and a Granger causality pre-test. Results. The endogenous inflation threshold is estimated at 5.7% (95% CI: 4.8–7.2%). Above it, inflation exerts a sharply negative effect on real GDP growth; below it, the effect is statistically indistinguishable from zero. Long-run responses are asymmetric: positive inflation shocks are significantly contractionary (coefficient ≈ −0.26, p < 0.001), while negative shocks show no detectable effect. The threshold is stable across subsamples. Implications. The breakpoint is a conditional-mean estimate, not a welfare optimum; any operational inflation target derived from it should be framed as risk-management guidance. The results support keeping inflation below the estimated threshold band.