Unit roots versus trend stationarity in growth rate estimation

This paper estimates growth rates of quantity and trade variables of 32 primary commodities for the periods 1970-1980, 1980-1990, and 1990-1999 through log-linear trend regressions. The properties of the models were assessed through conventional and stationarity statistics. Based on Monte Carlo experiments, of the 105 regressions examined, in only 35 cases the error term was found to be stationary at the 5% level of significance, implying that in many cases the stochastic component is far more important in explaining the behaviour of the series under consideration, in turn implying that the validity of growth rates should be interpreted with some degree of caution.