THE use of the moving average has become widespread because of the simplicity of its application in the computation of the trend of statistical time series. This method, however, has two essential disadvantages, which make it applicable only in a limited degree. The first of these is that cxtrapolation is impossible with a moving average; that is to say, it is not possible to estimate from the moving average the probable ordinate of trend even for the immediate future, unless the estimate is based upon other considerations quite unrelated to the moving average. Even for the last time-interval of the series, for which an actual item is displayed, it is not possible with a moving average to determine exactly the ordinate of trend, because the item pertaining to the next consecutive time-interval, which is the minimum necessary for the formation of an average, is not given. With this question of extrapolation I propose to deal fully in a subsequent article. In the present article I shall consider a second disadvantage of the moving average, which lies in the fact that