ON THE STATIONARITY OF SAVINGS IN THE UNITED-STATES

The question of stationarity of saving as a percent of disposable pers onal income is examined in detail. NIPA data is taken annually from 19 29-1988, and quarterly from 1970-1988, The initial approach uses Dicke y-Fuller tests. Their results indicate that, depending on the time per iod chosen, one is led to believe that both stationarity and non-stati onarity exists in the data. ARIMA models are then fitted to both annua l and quarterly data. Relatively large non-differenced models are nece ssary to fit the long range annual data (1929-1988), while relatively small differenced models fit quarterly and short range annual (1946-19 88) models. It is concluded that at least for saving as a percent of d isposable personal income in the United States, it is reasonable to be lieve that saving is not a stationary process.