Economists and statisticians are gradually amassing new knowledge of economic processes and enlarging the range of data which provide the underpinnings of their craft. Since the end of the Second World War, a number of new statistical tools have been developed for measuring the performance of the economy, and a good deal of new quantitative information on the structure and behaviour of the economic system has been collected. Among these new tools have been developments in the field of social accounting, and new and improved techniques for isolating and studying fluctuations in the business cycle. It is with the latter question that this paper is concerned-the decomposition of time series into their trend-cycle, seasonal, and residual elements. While the work in this field in Canada in the early 1950's was initially focused on measuring cyclical fluctuations in the quarterly national accounts, it has long since evolved a full-fledged life of its own, and a general programme of time series analysis has been under way at the Dominion Bureau of Statistics since 1955. The recent application of electronic computers to the problem is rapidly transforming the entire field.
The present paper has a three-fold aim: to review the rationale of time series analysis; to give an account of the background and status of the work at the Dominion Bureau of Statistics; and to provide a brief outline of the electronic computer programme now in use in this area.