Court-ordered release of adjusted census data spawns debate over accuracy of original count

From the Winter 2003 issue of The News Media & The Law, page 27.

By Jennifer LaFleur

When the 2000 census figures were released for St. Louis, Mo., in March 2001, they put the city’s population at 348,189. Advocates for minorities and the poor were skeptical, concerned that those communities were not accurately counted. When new figures based on statistical sampling were released in December 2002, the city’s population jumped 1.4 percent to 353,112.

Similar increases appeared in other cities around the country with large minority and low-income populations. Data for the city of Los Angeles showed an “undercount” of more than 75,000 people in the original count.

After it conducted its initial 2000 count, the Census Bureau used statistical sampling to get a more accurate count where census takers were unable to contact residents.

But the Census Bureau had decided, long before any numbers were released that it was going to use the actual population count, not the estimate derived from the statistical sample, as the official count that is used to divvy up federal funds and to create political districts. The Census Bureau also said that the adjusted count would not be released at all.

Concerned that some citizens might be ignored by the mail and door-to-door tallies, Oregon state senators Margaret Carter (D-Portland) and Susan Castillo (D-Eugene) challenged the bureau’s withholding of the adjusted figures, saying that because the census tally is used to determine redistricting and funds allocation, the public has a strong interest in knowing what the second count could reveal. They won in federal district and appeals courts. (Carter v. Department of Commerce)

A federal appeals court panel in Portland, Ore. (9th Cir.), Oct. 8, 2002 ordered the data’s release, ruling that the adjusted counts were not exempt from the Freedom of Information Act.

So on Dec. 5, the Census Bureau provided journalists and other requesters with CD-ROMs containing the data from its statistically adjusted count.

U.S. Congressmen in the House Government Reform Committee filed a similar case in a federal district court in California. The court ruled in March 2002 that the plaintiffs could have the adjusted census figures. (Waxman v. Evans)

The newly released adjusted data showed that the 2000 census count missed about 3.3 million U.S. residents.

That difference worries lawmakers from high minority cities such as St. Louis.

“The more things change, the more they remain the same,” said former ranking Census Subcommittee member Rep. William Lacy Clay (D-Mo.) in a statement. “We are in the process of analyzing the data and will offer further comments when the process is completed.”

And although the Census Bureau released the numbers, it did not stand behind them. Along with the data files, the CD-ROM contained a disclaimer saying that “the estimates dramatically overstate the level of undercoverage in Census 2000.”

Some advocates for poor and minority communities called the Census Bureau’s retraction a political move, but there was some statistical reasoning behind it.

“Imagine the big pile of census forms,” said USA Today database editor Paul Overberg. “You know there should be more. Whatever that upward growth of the pile is would be your undercount.”

“The problem is on the other end — people who got counted twice,” said Overberg, who analyzed census data for USA Today. “The machinery to catch those people, did not catch millions of dupes.”

Figuring out the net result can be confusing. Here’s the math: The U.S. House of Representatives Committee on Government Reform found that more than 6 million people were missed in the initial count, and more than 3 million were counted twice. The net “shortage” in population comes out to be about 3.3 million.

Such duplication was more of a problem this time around because of a “radical increase in the number of people with second homes — vacation homes or second homes people have in commuter marriages,” said Eugene Erickson, a professor of sociology and statistics at Temple University who conducted data analysis for the Census Monitoring Board.

For Ericksen, Overberg and others, the data released in December provides a new opportunity to examine how the census was conducted.

“It’s certainly something we can’t ignore — 3.3 million Americans. That’s a big number even for a big country,” said Frank Bass, director of computer-assisted reporting for The Associated Press.

The potential price tag on those missed populations is what concerns some local communities. An August 2001 study of the undercount for the Census Monitoring Board estimated that a 3.35 million undercount would mean a loss of $4.1 billion in federal funds for 31 states.

Rep. Henry Waxman (D-Calif.) called the error rates in the 2000 census count unacceptable.

“They conflict fundamentally with the principle of ‘one person, one vote,'” Waxman said in a statement published for a recent conference on the undercount. “Ultimately, the effect of the undercount is to systematically deny full federal funding and equal political influence to states and urban areas, particularly those with large minority populations.”

The U.S. Supreme Court already has drawn the line on which data are to be used for creating federal political districts, ruling that the adjusted numbers cannot be used for redistricting. States and communities may, however, use the adjusted data to realign political boundaries or redistribute funding.

“The Census Bureau didn’t want to release any of the data,” Temple’s Ericksen said. “This whole procedure has been shrouded by secrecy. It’s a good thing the data are out and for people to start asking questions.”

“The Census Bureau never showed what the evidence was that we shouldn’t take these numbers seriously and never pursued the research to tell us what the numbers should be. I’m hopeful that this will facilitate that debate,” Ericksen said.

Whether or not the actual count, the overcount and the undercount have been accounted for is still undecided. The panel that reviewed the bureau’s decision to go with the unadjusted data supported the agency’s decision but called for further study.

The Census Bureau had no official statement about its future plans for the data other than that it “continues to study the data” to see if it is possible to correct the numbers using a post-census survey that was conducted in late 2000.

For more information or to access census data, go to the U.S. Bureau of the Census Web site at www.census.gov or to the University of California Los Angeles Institute of Social Science Research at www.sscnet.ucla.edu/issr/.