The Demeny-Shorter and three-census methods for correcting age data

This paper presents, tests, applies, and compares methods that utilize age data collected at consecutive censuses to examine and adjust for age and coverage errors. The Demeny-Shorter method, for example, was devised for this purpose, and its flexibility in regard to census coverage errors is examined. The Demeny-Shorter method is found difficult to apply directly, so a method based on the same idea as the Demeny-Shorter method but utilizing age data from three, instead of two, successive censuses is presented and discussed as a possible alternative. This three-census method is applied to data from Turkey's censuses and, in some cases, found to be better than the Demeny-Shorter method, because the former allows for and estimates the likely changes in census coverage and different patterns of age errors in successive censuses. Unfortunately, the three-census method cannot be applied to data from most developing countries on account of a lack of the requisite series of censuses.