April 20, 2001

Experts Reach Different Conclusions Than Census Bureau On Decision Not To Adjust 2000 Count

WASHINGTON — The Presidential Members of the U.S. Census Monitoring Board convened a forum of expert statisticians and demographers to discuss their analysis of Census 2000 data and the factors that led to the recommendation not to adjust for the 6.4 million people missed and the 3.1 million people counted twice.

In their recommendation to the Secretary of Commerce, the Census Bureau cited four concerns they did not have adequate time to resolve in order to recommend a correction for the overcount and undercount: 1) inconsistencies with demographic analysis, 2) synthetic error, 3) balancing error, and 4) accuracy of geographic areas with populations below 100,000 people.

Now, more than a month after the recommendation, a group of expert statisticians and demographers analyzed the data available and have come to different conclusions regarding these concerns.

Dr. Jeffrey Passel, demographer and researcher at the Urban Institute, argued that "the inconsistency with demographic analysis should not be used as a basis for deciding against adjustment because the current estimate underestimates immigration, particularly Hispanic and undocumented immigration."

Responding to the issue of synthetic error, Dr. Joseph Kadane, Professor of Statistics and Social Sciences at Car-negie Mellon University said, "I'm puzzled as to why the Census Bureau uses synthetic error as a reason not to adjust when their materials make a case for just the opposite."

"It appears that balancing error was even less of a problem in 2000 than in the 1990 census, which leads me to believe that it shouldn't be a factor in determining adjustment," said Carnegie Mellon University Professor Dr. Steven Fienberg. "There also seems to be inconsistencies with the Bureau's methodology when determining the accuracy of populations below 100,000," Fienberg

"The science of the census did not evolve in a vacuum," said Census Monitoring Board Member and former Under-secretary of Commerce Dr. Everett Ehrlich. "If increased accuracy is the goal, then transparency, scientific discussion and the immediate release of adjusted data are part of the solution."

At a recent Senate hearing, bipartisan census experts agreed that the Bureau should release the block-level adjusted census data for research purposes. "There is a sense of urgency to release this data. Before it's too late, we need people — independent researchers and scientists — examining the data to find out if there's a problem with it and to look for solutions. The distribution of $185 billion is too important to risk not having enough time to complete a sound analysis," said Sen. John F. Kerry at the hearing.

Currently, the Census Bureau has not said when or if they plan to release adjusted census data. The bipartisan Census Monitoring Board was established in 1997 to monitor Census 2000 operations. Its findings are reported to Congress every six months. The Board's latest report was submitted to Congress earlier this week and is available at: www.cmbp.gov.

Return to the Frontpage