Oct. 26, 2001
Basis of “Revised Early Approximation” of Undercounts Released Oct. 17, 2001
John H. Thompson, Principal Associate Director for Programs
Preston J. Waite, Associate Director for Decennial Census
Robert E. Fay, Senior Mathematical Statistician
At his request, we provided Acting Director William Barron a “Revised Early Approximation” to
illustrate the effect on the A.C.E. estimates of a possible future revision. We believe that these
preliminary estimates served several public purposes: to show that evaluation data indicated that a
substantial revision of the A.C.E. estimates was required; to explain that we now had a candidate
explanation for the discrepancies between the A.C.E. estimates and demographic analysis, since
that had been the key basis of the March, 2001 ESCAP recommendation; and to show that, in spite
of this finding, we expected a revised A.C.E. to continue to show a differential undercount. We
believe that Mr. Barron appropriately explained that we had not yet revised the detailed A.C.E.
estimates, and that extensive research was required before we could do so. Because of timing, we
were unable to discuss the methods or results with the entire ESCAP.
Each of us worked directly with the staff in reviewing the findings on A.C.E. accuracy from
several evaluation studies, including the Measurement Error Reinterview (MER) and Matching
Error Study (MES). Our review indicated:
Additional work on analysis of the P-sample MER results will be important, but it
appeared that MER and MES results were offsetting and indicated little net P-sample
The MER and person duplication study now indicated the E-sample would require
substantial correction, because the A.C.E. had underestimated erroneous enumerations.
The MES results also measured some underestimation, but we were unable to include them
because of some overlap with MER.
Although there was evidence of correlation bias for Black males similar to that shown in
previous censuses, we do not anticipate that revision of the A.C.E.