Can you do the wrong thing and still be right? Hypothesis Testing in I(2) and near-I(2) cointegrated VARs

We review the I(2) model in the perspective of its application to near-I(2) data, and report the results of some Monte Carlo simulations on the small sample performance of asymptotic tests on the long-run coefficients in both I(2) and near-I(2) systems. Our findings suggest that these tests suffer from some finite-sample issues, such as size bias. However, the behaviour of these statistics is not markedly different in the I(2) and near-I(2) case at ordinary sample sizes, so the usage of the I(2) model with near-I(2) data is perfectly defensible in finite samples.