Stochastic and deterministic trends in Finnish macroeconomic time series

The appropriateness of the Dickey-Fuller unit root test is studied using two alternative unit root test models. The segmented trend model is strongly supported and the second-order trend model is favoured for some series. The sample set consists of observations of nine basic macroeconomic time series describing the fundamentals of the Finnish economy between 1860 and 1989. The results clearly show that care is required in interpreting unit-root tests since failure to reject does not entail that the null is true. Structural breaks in the data generating process, in this case wars starting in 1917 and 1939, support models of the deterministic trends class. However, it is argued that the univariate testing procedures laid down in the unit root literature do not provide information to macroeconomic controversies.