A Look at the Stationarity of 14 Quarterly U.S. Economic Time Series

Application of single series time series methods has become very popular, due, in particular, to the short term forecasting abilities of such models. This is particularly true of the Box-Jenkins linear models. In this paper, the relationship between the forecasting ability and the number of data points used in the estimation of Box-Jenkins models, for 14 economic time series, is considered, with the result that in a large number of cases, there is very little gain in analysing long historical records. In fact, it is possible that forecasting ability may be impaired if too many observations are considered.