The application of time series forecasting methods to an estimation problem using provisional mortality statistics

Provisional estimates of mortality for selected causes of death are published each month by the National Center for Health Statistics. These estimates are based upon a ten per cent sample of death certificates in the United States. Final mortality results, based upon all the death certificates for a calendar year, are available one to two years after publication of the provisional estimates. This paper explores the potential of time series forecasting techniques for improving mortality estimates by using the correlation structure between the provisional and final series to obtain mortality estimates that are expected to be closer to final values than currently used provisional estimates.