[영문]The U. S. Census Bureau used administrative records to examine the quality of the estimates of duplicate enumerations in Census 2000. The estimated number of duplicates in the Census 2000 count of 281,421,906 people was 5.8 million, based on census information only. The identification of duplicates was possible because Census 2000 was the first census to use optical character recognition technology that permitted converting all the names on the census questionnaires into electronic format. Although additional field work to evaluate the duplicates was not practical by the time they were discovered, an administrative records database derived from seven government record files was available for evaluating the estimates of census duplicates. The database had census-like records and was formed using new methodology as part of the research associated with Census 2000. The evaluation using administrative records validated that there were a large number of duplicate enumerations in Census 2000. A clerical review also offered a validation of the estimates of duplicate enumerations based on only census information but raised questions about the accuracy of some types of duplicates identified only with administrative records. The results pointed to areas for refinements in evaluating census duplicates with administrative records. The validation of the estimates of duplicate enumerations through the use of administrative records demonstrated further potential for contributions from administrative records in the evaluation of census duplication