This study investigates the effect of volatility spillover on nonlinear causality tests. We employ three nonlinear causality tests as logistic smooth transition, exponential smooth transition and time-varying models. These three causality tests include usual asymptotic test, heteroskedasticity-robust tests using the heteroskedasticity-consistent covariance matrix estimator, and wild bootstrap tests. Our Monte Carlo simulation results show that nonlinear causality tests find spurious nonlinear causality when errors have volatility spillover effect. In particular, the size distortions of nonlinear causality tests increase when persistence of volatility is high and/or volatility spillover effect is large. These properties are observed for heteroskedasticity-robust tests and wild bootstrap tests in addition to asymptotic tests. Furthermore, the power of nonlinear causality tests decreases in the presence of volatility spillover effect.

nonlinear causality tests, volatility spillover effect.