Summary: GAO discussed the results of the 1990 Census Post Enumeration Survey (PES), to determine whether to adjust the 1990 census counts. GAO noted that: (1) PES estimated a net undercount of about 2.1 percent of the enumerated resident population of 248.7 million, or about 5.3 million persons, and the net undercount, based on demographic analysis, was about 1.8 percent, or about 4.7 million persons; (2) an approach focusing on double-counts and missed persons indicated that the 1990 census included about 14.1 million errors, including a gross census undercount of about 9.7 million persons; (3) the Census Bureau's lack of improvement in reducing the undercount, its differential nature, and the extent of its census errors reinforce the importance of census reform; (4) accelerated time schedules limited the Census Bureau's ability to analyze the results of its 20 PES evaluations; (5) due to sampling variations, the Census Bureau could not ensure the accuracy of its evaluation of total PES error; (6) the Census Bureau's evaluation of the combined effect of sampling and nonsampling error in PES determined that the national undercount was between 1.23 percent and 2.20 percent at a 95-percent confidence level; and (7) the Secretary of Commerce will need to determine whether PES results were sufficiently precise to ensure census count accuracy at all levels of geography.
