Counting on science to fix the census

By Steve Chapman Copyright 1998 Chicago Tribune December 3, 1998

The United States is a large country with hundreds of millions of residents, including many who are hard for the government to find and some who make it their business not to be found--from illegal immigrants to criminals to cranky sorts who just want to be left alone. So census-takers have a formidable challenge as they strive to locate every single person within our borders.

They have been falling short in that effort. In 1990, it is estimated, they missed 8.4 million people, while mistakenly double-counting 4.4 million. The federal government says this was the first census in 40 years to be less accurate than its predecessor. Given the liberty and looseness of American life, the problem is not about to go away. If the FBI can't track down Eric Rudolph, there will always be a few million souls who will elude the clutches of the Census Bureau.

Monday, the Supreme Court heard arguments in a case centered on what should be done about this problem. The Clinton administration proposes to adopt what is known as "statistical sampling" to account for people who are somehow left out by the methods used in the past. Republicans in the House of Representatives, however, take the position that, since the Constitution calls for an "actual enumeration" of the population, the government can't count people it can't find.

Lower courts ruled in the Republicans' favor, but I will spare readers the hair-splitting arguments on legal and constitutional issues. The more pertinent question for non-lawyers is whether sampling is a sound way to conduct a national census.

From a scientific point of view, it seems eminently rational. Experts are certain that many people are passed over by the official tabulation. Most of those omitted are allegedly members of racial minorities. As a result of the omissions, some states and cities get fewer representatives in Congress or less federal money than they are entitled to.

Critics believe we can arrive at a much more precise number by the use of statistical techniques. Some of the people who don't return their census questionnaires would be visited personally by census takers, and the information gathered in these interviews would be used to make educated guesses about the numbers and socioeconomic characteristics of other households that failed to respond.

"It is fruitless to continue trying to count every last person with traditional methods of physical enumeration," says economist Charles Schultze, chairman of the Council of Economic Advisers under President Carter. He chaired a panel commissioned by the National Academy of Sciences which in 1994 concluded that sampling would yield a double blessing--a more complete enumeration of the American people and savings of at least $300 million.

But there are times when science offers the wrong answers. Using estimates to calculate the population may make sense for purely demographic purposes. But when we are talking about political representation in a democratic republic, it's an entirely different story.

Consider an analogy: Every election year, thousands of ballots are damaged or indecipherable. Using statistical methods, we could make a very plausible guess about who would have received those votes, add them to the overall tally and get a more accurate picture of public preferences than if we simply threw out the defective ballots. But no one would think of using sampling methods to decide elections.

Instead, we do things the old-fashioned way: We count only those ballots that can be read, and we resign ourselves to the impossibility of a flawless vote count. Some people get shortchanged that way. But they gain along with everyone else, because the fundamental integrity of the electoral system is preserved. The existing census may be incomplete, but it has the singular virtue of not including hypothetical people, giving it an integrity that sampling can never have.

Republicans have been accused of not wanting to use these techniques in the census because the voters who currently go uncounted are largely the sort of people who vote Democratic, as if that should resolve the matter. No doubt the GOP is acting largely out of self-interest--but then, so are Democrats, who stand to gain seats in Congress.

If it were wealthy whites who were omitted by the traditional count, Democrats would be denouncing this proposal for fiddling with the census in a way that most people can't understand and that would be highly vulnerable to political manipulation. And they would be right. Once slippery statistics replace clunky head counting, the census will be seen as little more than a creature of politics.

Turning to science to settle the issue harbors more hazards than benefits. It would be nice to have a perfect census, but wisdom suggests settling for one that is merely good enough.

Comments on this posting?
Click here to post a public comment on the Trash Talk Bulletin Board.
Click here to send a private comment to the Junkman.

Material presented on this home page constitutes opinion of Steven J. Milloy.
Copyright © 1998 Steven J. Milloy. All rights reserved on original material. Material copyrighted by others is used either with permission or under a claim of "fair use." Site developed and hosted by WestLake Solutions, Inc.