An econometric analysis of the U.S. postwar G.N.P.

Whilst most developed countries have experienced stable economic conditions during the postwar period, the acceptance of the unit root null hypothesis implies wild fluctuations in the major economic variables over time. This paper investigates the sensitivity of the decisions to accept the unit root hypothesis to the specification of the trends underlying the U.S. postwar GNP and other macro-variables. In particular, the relationship between the nominal GNP and the resident population is found to be a non-linear one. The unit root null hypothesis can be firmly rejected when the conditional mean of the nominal GNP series is represented by a quadratic trend variable and the assumptions that changes in the price level and the resident population lead to equiproportionate changes in the GNP are not enforced on the data. The case of quarterly observations is also investigated for seasonally unadjusted and adjusted data.The author is indebted to two anonymous referees and the Managing Editor for helpful comments.