Small sample properties of CIPS panel unit root test under conditional and unconditional heteroscedasticity

This paper used Monte Carlo simulations to analyze the small sample properties of cross-sectionally augmented panel unit root test (CIPS test). We considered situations involving two types of time-series heteroskedasticity (unconditional and ARCH) in the unobserved common factor and idiosyncratic error term. We found that the CIPS test could be extremely robust versus the two types of heteroskedasticity in the unobserved common factor. However, we found under-size distortion in the case of unconditional heteroskedasticity in the idiosyncratic error term, and conversely, over-size distortion in the case of ARCH. Furthermore, we observed a tendency for its over-size distortion to moderate with low volatility persistence in the ARCH process and exaggerate with high volatility persistence