Summary

Pursuant to a congressional request, GAO: (1) provided information on the range of total errors in 1990 census; and (2) compared error levels in the 1990 and 1980 censuses.

GAO found that: (1) for the 1990 census, both the Post Enumeration Survey (PES) and the Bureau of the Census demographic analysis showed a net census undercount; (2) the 1990 census included a minimum of 14.1 million gross errors, or 5.7 percent of the census count, and perhaps as many as 25.7 million errors, depending on how broadly the census error was defined; (3) the census included such errors as persons who were counted more than once, nonexistent persons included because census enumerators falsified data, and persons included in the census but assigned to the wrong location; (4) the estimate was based on preliminary PES data that indicated that approximately 4.4 million persons were either double-counted in their census block or in a nearby block, or were fictitious; (5) a more comprehensive approach to defining census errors revealed that there were about 10.2 million erroneous inclusions in the 1990 census; and (6) the minimal percentage of gross error in the 1990 census was 5.7 percent of the census count, while the minimal percentage of gross error in 1980 was 3.4 percent of the count.