Statistical information bulletin for Africa = Bulletin d'information statistique pour l'Afrique : No.14

Abstract

All statistical data collection systems, such as censuses and surveys are prone to error is gradually becoming acknowledged by both producers and users of statistics. For analytical purposes, we can divide the errors of a census in to two: errors of coverage and errors of content. Coverage errors comprise those errors that affect the headcounts or counts of housing units. These errors take two forms under enumeration, the more usual type, occurs when some persons or housing units that should be counted are missed. Over enumeration, on the other hand, occurs when a set of persons or housing units are enumerated more than once. Content errors are errors related to the information on the characteristics of persons and housing units report in the census. An evaluation program of a census attempts to estimate the magnitude of errors and sometimes to delineate their causes and patterns. The results will enable users of the statistics to decide, given their magnitude, whether they are suitable for use in a given circumstance. moreover, the identification of causes of certain errors will help survey statisticians to institute remedial measures in future censuses or surveys. This paper has the objective of reviewing the problems and practices of content and coverage error evaluation in African censuses, covering the period 1960 to 1980.