For the 2000 census the most recent revision of the Accuracy and Coverage Evaluation (A.C.E.)—Revision II—estimated a slight national net overcount of about 0.5 percent of the household population, or 1.3 million extra people, which is the first estimated net overcount in census history. The A.C.E. Revision II also estimated a difference of 2.9 percentage points between the net undercount rate for blacks (1.8 percent) and the net overcount rate for non-Hispanic whites (1.1 percent) and a difference of 1.8 percentage points between the net undercount rate for Hispanics (0.7 percent) and the net overcount rate for non-Hispanic whites (U.S. Census Bureau, 2003c:Table 1).

The final Revision II A.C.E. estimates appear to tell a story of success in reducing the net undercount and differences in net undercount rates in 2000 compared with 1990. Despite the extensive, imaginative, and high-quality work by Census Bureau staff to develop the Revision II estimates, however, it is difficult to draw clear-cut conclusions from them about undercounts and overcounts in 2000 because of limitations in the available data for reestimation. The evaluations that fed into the Revision II estimates provided more information than previously available about the numbers and sources of erroneous census enumerations and, similarly, more information with which to determine the residency status of the independent A.C.E. sample. They provided little new information, however, about the numbers and sources of census omissions. Because of changes in estimation methodology, the revised 2000 estimates are not comparable with the 1990 PES results.

To set the stage for our assessment of population coverage in the two censuses in Chapter 6, we first describe the two major coverage evaluation methods—dual-systems estimation (5-A) and demographic analysis (5-B). We then review the history of coverage evaluation for 1990 and 2000 (5-C and 5-D, respectively), including decisions about the possible use of coverage estimates to adjust census counts for measured net undercount. In June 1991 the Census Bureau director recommended that PES-based population estimates be used to adjust the 1990 census counts, but the secretary of commerce decided against adjustment. In December 1992, the Bureau director decided that revised PES estimates would not be used to adjust intercensal population estimates. In 2000 the Census Bureau planned to use A.C.E.-based population estimates to adjust the census counts, but on three separate occasions—March 2001, Octo-