Summary: As part of its Accuracy and Coverage Evaluation (ACE), the U.S. Census Bureau interviewed people across the country to develop an estimate of the number of persons missed, counted more than once, or otherwise improperly counted in the 2000 census. In conducting the interviews, which took place in person or over the phone, Census faced several challenges, including (1) completing the operation on schedule, (2) ensuring data quality, (3) overcoming unexpected computer problems, (4) obtaining a quality address list, and (5) keeping the interviews independent of census follow-up operations to ensure unbiased estimates of census errors. The Bureau completed the interviews largely ahead of schedule. On the basis of the results of its quality assurance program, the Bureau assumes that about one-tenth of one percent of all cases nationally would have failed the program because they were believed to have been falsified. Early on, the Bureau dealt with an unexpected problem with its automated work management system, which allows supervisors to selectively reassign work among interviewers. According to the Bureau officials, the Bureau addressed the underlying programming error within two weeks, and the operations proceeded on schedule. The address list used for interviews had fewer nonexistent listings than did the lists used by the major census questionnaire delivery operations. An accurate address list is important to prevent unnecessary and costly efforts to locate nonexistent addresses. Although the Bureau implemented controls to keep the nonresponse operation separate from the interviews, the assumed independence of the census and ACE was put at risk because another follow-up operation intended to improve census coverage overlapped with the interviews.
