The integrity of the United States Census may be at risk. That’s certainly what some Republicans are concerned about in the Obama era.
For years, liberals have rightly pointed out that the decennial census undercounts certain groups — minorities, illegal immigrants, transients. An internal Census Bureau analysis from 2003 also determined that the last census overcounted whites and Asians. The liberals’ solution? Statistical sampling, in which census workers would study closely the populations of selected sample areas and use the resulting numbers to correct the numbers obtained by direct enumeration.
Statistical sampling was a hot political issue in the late 1990s, and it may be making a comeback. We got a whiff of census controversy during the brief period earlier this year when Republican senator Judd Gregg was President Obama’s choice for secretary of commerce. Liberals demanded that the White House remove the Census Bureau from the commerce secretary’s authority, where it has traditionally resided. This appears to have played at least some part in Gregg’s decision to withdraw.
For many Republicans, it looked like political gamesmanship — and possibly the prelude to a new fight over a practice they thought they had defeated a decade ago.
Why are the Republicans so opposed to sampling? For one thing, they say that using it in the census is unconstitutional. Article I lays out the reason for the census — the apportionment among the states of taxation and of members of the House of Representatives — and then indicates how it is to be carried out: “The actual Enumeration shall be made within three years after the first meeting of the Congress of the United States, and within every subsequent term of ten years.” Republicans, adhering to their strict-constructionist tradition, say the word “enumeration” shows that the Founders wanted an actual count. Furthermore, since sampling is likely to target minority and urban populations — traditional Democratic voters — conservatives worry that apportionment of representatives would be unfairly adjusted to favor Democrats.
Democrats, meanwhile, claim to support sampling for its scientific precision. “There are two ways to get an improved census: a national registration program or sampling,” says Dr. Kenneth Prewitt, who was director of the Census Bureau from 1998 to 2001. Currently a consultant for the bureau, Prewitt is one of the most vocal supporters of census sampling. He says sampling “takes on a mystique for some people,” but he argues that it is much more precise than the current “headcount” policy.
Republicans like Rep. Patrick McHenry from North Carolina disagree. McHenry, the ranking member of the subcommittee that oversees the census, calls sampling “census adjustment” and says it actually makes the count less accurate. He points to a 2002 article by two UC Berkeley professors, which notes, “adjustment can easily put in more error than it takes out.”
“We’ve come a long way,” McHenry says about improved technology since the census began in 1790. “The 2000 Census was the most accurate in our nation’s history. We’ve got a lot to be proud of.”
Prewitt thinks census-taking is getting harder, particularly for those groups historically undercounted. “It’s going to be difficult in 2010,” he says. “It won’t get easier to do censuses.”
Prewitt offers a colorful challenge to those who say sampling doesn’t get the job done. “Next time you go to do a blood test, have the doctor take it all out,” he says. Then he gets serious. “Anyone who doesn’t believe in sampling doesn’t believe in any numbers” the government uses, he says. He notes that the federal government uses statistical data all the time for developing policy in areas such as labor and crime.
Prewitt’s approval of statistical sampling is not new. He strongly advocated it in his official capacity a decade ago. Under Prewitt’s direction, the 2000 census planned to use a “dual-track” system that would have produced two sets of census data: one from sampling, one from direct counting. That was a compromise in response to congressional opposition to the use of sampling alone. Still, the Republican-controlled House of Representatives filed suit against the Commerce Department in 1998 to challenge the constitutionality of the use of the sampling data for congressional apportionment. The Supreme Court ruled in favor of the House but declined to rule on the constitutional issue. As a result, sampling is still allowed for non-apportionment use, including redistricting. Rep. Lynn Westmoreland of Georgia, another Republican member of the census subcommittee, is wary of political manipulation of the census. “We don’t want that to be a back door for the Democrats,” he says.
Prewitt admits that the dual-track system wasn’t the best idea. He notes that sampling won’t be used in 2010 since there isn’t enough time to begin the process. But would Prewitt like to see census sampling in the future? “I still believe,” he says proudly. He stresses that he’s only one of about 100 consultants to the Census Bureau, but his status as a former director may give him a leg up.
So are Obama and the Democrats itching to use sampling in future censuses for political gain? The folks in charge of the Census Bureau demur. Commerce Secretary Gary Locke told the senators at his confirmation hearing that there were “no plans in the Department of Commerce or the Census Bureau to use any type of statistical sampling with respect to population count.” The Census Bureau’s director, Robert Groves, seemed to confirm this during his own hearing in May. “I agree fully with Secretary Locke’s testimony that statistical adjustment of the census is eliminated as an option for reapportionment and further that statistical adjustment will not be used for redistricting,” Groves said.
Rep. Westmoreland isn’t convinced. “The people that are in charge right now,” he says, “I wouldn’t put anything past them.” Sampling may not be used for reapportionment or redistricting, but will there be new claims of undercounting, with new demands for adjustments in the distribution of federal aid? Will there be a push to institute sampling in the 2020 census? The liberals aren’t going to forget this issue. We mustn’t forget it either.
– Michael Warren, a Collegiate Network intern at National Review, studies economics and history at Vanderbilt University.