Stability of Long-run Relationships for Countries in Transition: A Hansen Test Study

The results of a Monte Carlo research for the Hansen Lc, MeanF and SupF stability tests for long-run relationships are reported. The tests are related to the Phillips-Hansen and Hansen semiparametric methods, which involve kernel density estimation of the long-run covariance matrix. We compare the effect of the choice of kernel on the performance of the tests, and also check the effect of misspecification of the model (lags for the disturbances) on the behaviour of test statistics, and behaviour of percentiles. The results indicate that the best – both in terms of efficiency and robustness to misspecification error – is the Parzen kernel.