Census "count" numbers to be official census 2000 numbers

On October 17, 2001, the Census Bureau announced it will release the remaining Census 2000 data products, post-census estimates, and survey controls using unadjusted data. On March 1, 2001 it made a similar announcement concerning redistricting data.

These announcements essentially bring to rest several years of debate between supporters and opponents of a census count adjusted by statistical analysis. In the end, the Bureau declined to use adjusted numbers because of doubts about the reliability of the Accuracy and Coverage Evaluation (ACE) survey. In the opinion of a panel of Census Bureau scientists, this post-census test against the accuracy of the original census counts was inadequate to be used to adjust them.

For a more complete analysis of the testing and the factors involved in the decision, you can review the Census Bureau panel documents at:

            http://www.census.gov/dmd/www/EscapRep2.html

In short, ACE was a survey, separate from the census, of approximately 314,000 households, designed to be used to make comparative corrections to the 2000 Census. The Bureau, however, discovered a high number of discrepancies in expected values, especially when compared against another test method, Demographic Analysis. Mostly, ACE did not compensate sufficiently for overcounts, where people such as college students might have been accounted for in more than one place. 

A significant reason the Bureau felt more confident with the count than with any adjustment is because of the general opinion that the 2000 Census count was a decided improvement over 1990. Because of expanded partnership programs, simplified forms and a national, directed advertising campaign, there was a higher than expected return rate for census forms. Congress also appropriated large sums to provide a sufficient workforce to do a complete follow-up of unanswered forms. Due to the above programs and an expansion of “imputed” returns, the Bureau anticipates that it has also worked toward reducing the differential undercount. Past tests show clearly that a census generally undercounts minorities at a greater rate than the rest of the population.

The following table from the Bureau’s, “Report of the Executive Steering Committee for Accuracy and Coverage Evaluation Policy on Adjustment for Non-Redistricting Uses,” describes the improvements in the undercount from 1990 to 2000. The Bureau estimates that net undercount (the difference between duplicate counts and uncounted persons) in 2000 was actually not statistically significant, further limiting the need for adjustment.

Estimates of Percent Net Undercount, by Race, Sex and Age: 1990 and 2000

(a minus sign denotes a net overcount)

At this point, all we have is national information. We have no specific information of rates for the State of Minnesota or for any parts of the state. There has been no announcement from the Bureau of any possible release of those numbers in the future. Historically, Minnesota’s undercount rate has been lower than the national average. Any indicators we have to date suggest that trend should continue.

The Bureau plans to leave open the possibility that further study will improve ACE, allowing for the use of adjustments in future estimates programs and survey controls. However, there is no timeframe or deadline for this continuing research.

Although the decision means the 2000 Census “head count” will stand, political disagreement is likely to continue. Rep. Dan Miller (R-FL), chairman of the House Government Reform Committee's census panel, issued a statement saying that “it is time to put adjustment, for political purposes, to rest.” On the other hand, Rep. Carolyn Maloney (D-NY), former ranking member of that panel, claimed the announcement “gave us more questions than answers.” Several large city mayors are planning to sue over the decision, as unadjusted counts will likely mean fewer people and, therefore, fewer federal dollars.