THE ROLE OF SEASONALITY IN ECONOMIC TIME-SERIES - REINTERPRETING MONEY-OUTPUT CAUSALITY IN US DATA

While empirical evidence on the relationship between money and income has mainly been presented using seasonally adjusted data, seasonally u nadjusted data are used in this paper to examine the time series behav iour of money, real GNP and industrial production, at both the seasona l and zero frequencies, based on tests of cointegration and seasonal c ointegration. Two important conclusions are reached in the paper. Firs t, although the univariate time series properties of M1 and real GNP a ppear to be very similar at both the seasonal and zero frequencies, se asonal comovements of MI and real GNP turn out to be different from lo ng-run comovements. Second, when seasonally unadjusted data are used, there appears to be no long-run relationship between money (M1 or M2) and output in the sense that the null of no cointegration cannot be re jected. Moreover, then is evidence of some feedback from output to mon ey so that money is not necessarily exogenous. Consequently, as we mig ht lose a possibly important chain of causation from money to income b y ignoring the information concerning seasonal fluctuations, this pape r provides further evidence that researchers should use raw data inste ad of seasonally adjusted data for inference and forecasting purposes. (C) 1997 Elsevier Science B.V.