Tests for asymmetry in possibly nonstationary dynamic panel models

For partly nonstationary dynamic panel models whose component series are threshold autoregressive processes having possibly unit roots, an instrumental variable method is applied to construct a Wald test and a t-bar type test whose limiting null distributions converge to a chi-square distribution and the standard normal distribution, respectively, as component series length increases to infinity. Finite sample sizes and powers of the proposed tests are investigated by a Monte Carlo simulation.