Finnish GNP-series 1954/I–1990/IV: Small shock persistance or trend stationarity? Some evidence with variance ratio estimates

The “shock persistance” of Finnish adjusted quarterly real GNP series in logarithms from 1954/QI to 1990/QIV is analyzed using variance ratio estimators. The results indicate that the random walk component of the series is not big. The small sample properties of variance ratio estimators are studied using empirical distribution derived from simulations. The persistence measures calculated via the ARIMA modelling of the lnGNP_t series are biased upwards. The sampling properties show that the simple random walk model is not an alternative model for the lnGNP. A trend stationary alternative, an AR(2) process, gives almost the same “shock persistence” measures as the assumed unit root processes.