Seasonal Adjustment of Vital Statistics by Electronic Computer

Abstract

T HE MONTHLY variation in the frequency of vital events has long been the subject of scientific study and popular c,onjecture (1-4). In this country there has been remarkable stability during the past decade in the rhythmic yearly variation of the birth rate, the marriage rate, and the death rate. These seasonal patterns, while of great intrinsic interest, pose a problem for the analyst who is interested in determining the underlying trend of a time series. Knowledge of the trend on a monthly basis, free from distortion due to large seasonal fluctuations, facilitates comparisons between different types of series to study; for example, the influence of business cycles, on marriage, and birth rates (5), the relation between excess mortality associated with influenza epidemics and the risk of fetal deaths (16, 7), and the epidemiology of heat waves (8). Death rat-es and deaths due to influenza and pneumonia, free from seasonal factors, serve as quantitative indices of the impact of pneumonia-influenza epidemics (9, 10). In general, information about the current trend in birth, marriage, and death rates provides a foundation on which to base estimates of future developments in time series of vital statistics. To establish the level of trend from monthly data, the task of the analyst is to estimate the seasonal factor and to eliminate it from the original observations. Several methods have been developed ito do this mostly in connection with economic data (11). Same-month-year-ago comparisons are perhaps the most familiar method used. This involves eliminating the seasonal factor by dividing the figure for a given month by the figure for the same month in the previous year, and expressing the reisult as some percentage change. This method, discussed later, has serious limitations. Two oither techniques, the monthly means and the link relatives, were among the first refinements; but they too give approximate results. The ratio-to-movingaverage method, with its advantages olf greater flexibility and precision, has been adopted by many groups engaged in large-scale seasonal adjustment work despite the fact that it is quite laborious. In 1954 the Bureau of the Census, and more recently the Bureau of Labor Statistics, developed electronic computer programs for seasonally adjusting time series, using an adaptation of the standard ratio-to-moving-average method (12-14). The application of these programs to time series of vital statistics yields useful information about their characteristic seasonal patterns and underlying trend. This paper describes tfhe method developed by the Bureau of the Census (version X-9 of Census Method II) and its application to national vital statistics collected by the Division of Vital Statistics, National Center for Health Statistics, Public Health Service.