Estimation and Testing for Cointegration for Seasonal Time Series

Yılmaz AKDİ (Ankara University)
Nuri UÇAR (Bilkent University)

This paper proposes a new unit root test and estimator based on the periodogram analysis for the seasonal time series. We have observed that our estimator is better than OLS estimator for the cointegrating vector under the assumption that long-run equilibrium exists. It is found that the asymptotic distribution of the normalized periodogram ordinate is not affected by the seasonality factor, which implies that the test statistic can be used for testing seasonal unit root as well as zero frequency unit roots. Another interesting feature of this new test is to have an exact (known) distribution whereas the other tests such as DHF(1984), HEGY (1990) and so on suggested until now have the nonnormal distribution under the assumption of nonstationarity.