The 2000 census: Don't count on it

August 28, 1998

The following editorial appeared in the San Jose Mercury News on Wednesday:

Monday's decision by a federal court, the recalcitrance of Republicans in Congress and the fast-approaching deadline make it ever less likely that the national census in 2000 will employ statistical sampling to correct for people not counted directly.

Maybe a second demonstrably inaccurate census will change minds in time for 2010.

In 1990, the head-count census missed about 8 million people, mostly blacks and Hispanics, and counted 4 million people twice. About 1 million of the people missed were in California.

Statisticians widely agree that using sampling methods to estimate the number of uncounted people will increase the accuracy, and decrease the cost, of the census.

The Census Bureau has long used sampling to gather data about the American population, such as how many are renters and how many are homeowners.

But earlier this week, a federal appeals court said the census law does not permit the use of sampling in establishing the population count used to determine how many seats each state gets in the House of Representatives.

The ruling is no real legal impediment to adding sampling to the census. Congress need only change the census law. But where there is no legal impediment, there is a huge political one.

Republicans in Congress might be deliberately blind to the best way to count Americans, but they know how to count votes.

The uncounted are generally poor, transient and minority.

If they are counted -- that is, if they are added in by statistical adjustments -- states like California and New York will get more congressional districts, and the additional districts will be located in Los Angeles and New York City, which reliably elect Democrats.

So congressional Republicans, in the majority, have refused to authorize the Census Bureau to employ sampling in the 2000 census, although there has been some compromising about what the bureau can prepare to do.

Republicans have a practical objection and a legal objection; only the legal one is respectable.

The practical one is that statistical sampling is an invitation to fiddle with the numbers. In 1994, a study conducted by the National Academy of Sciences found that using proven statistical sampling methods to supplement the census would be more accurate than trying to count individuals.

The finding is endorsed by the American Statistical Association.

If the methodology used by the Census Bureau is open to analysis by professional statisticians everywhere, the danger of statistical sleight of hand is eliminated.

A census with sampling is more accurate and less expensive. Census Bureau officials, who support sampling, say it could save nearly 25 percent, about $800 million, on the 2000 census.

The GOP legal argument is that in addition to federal law, the U.S. Constitution calls for an "actual enumeration" of the population, a question which Monday's appeals court ruling avoids. The Supreme Court has not ruled on whether the Constitution locks in an 18th-century way of counting.

Congress and the courts should bear in mind the point of this once-a-decade exercise. It is to get an accurate picture of the population. Nonpartisan people who count people for a living agree on how to do that. Only partisan people-counters do not.

Pub Date: 8/28/98

Baltimore Sun Articles
Please note the green-lined linked article text has been applied commercially without any involvement from our newsroom editors, reporters or any other editorial staff.