T HE PROBLEM of the analysis of economic time series has occupied the attention of a great number of economists, mathematicians and statisticians, since the time when a considerable amount of statistical data became available in this field. It is probably one of the most unfortunate circumstances for our science that most economic data come in the shape of time series. Hence they cannot easily be dealt with by the ordinary statistical methods which have proven useful in other statistical applications, especially in physics and biology. The most important reason these methods are inapplicable is very simple to understand. Items of data which are ordered in time are in general not mutually independent or random in time. They violate one of the most significant assumptions required in the application of probability theory on which all modern statistical tools are founded. Possible methods of analyzing interrelated statistical events have not yet been very fully explored, in spite of the fact that some progress has been made especially by the Russian school of probability. (Markoff chains.') In practical statistical problems as distinguished from theoretical probability questions, we are far from satisfactory solutions. Other features may also be present, which make the application of statistical methods difficult. The variance of the population, for instance, may not be constant in time.2 There are two problems to be distinguished in the analysis of economic time series. The first is one which this material has in common with all other statistical data. This first problem is an investigation of the random character of the data, e.g., the separation of the random and the non-random parts of time series. There are still many workers in the field of economic statistics who fail to realize fully the great importance of this problem. It arises from the fact that only an answer to the question of what is and what is not random in a time series can