Countdown to nation's head count Census: Plans call for an unprecedented use of scientific sampling to supplement the actual 2000 count, but Congress is deeply skeptical that sampling will work.

Sun Journal

April 07, 1997|By James Bock | James Bock,SUN STAFF

It is the government's biggest peacetime mobilization: Open hundreds of offices, hire an army of temporary employees and mail forms to 119 million housing units. The object: a statistical snapshot of who and where an estimated 274 million U.S. residents are on April 1, 2000.

The census, the national head count that has been taken every decade since 1790, is still three years away, but a sense of urgency is already developing on Capitol Hill and at U.S. Census Bureau headquarters in Suitland.

The General Accounting Office recently labeled the 2000 census "at high risk" of mismanagement.

The basic problem is this: The Census Bureau plans unprecedented use of scientific sampling to supplement the actual head count. Congress, which has authority over the census, is deeply skeptical that sampling will work.

In the past, the census has tried to count every household, first by mail, then door to door for those who don't respond to the questionnaire. Paying census takers to make repeated visits to elusive households has been expensive.

In 2000, the Census Bureau plans to count 90 percent of each census tract (areas of about 4,000 persons) by mail and in person. But for the final 10 percent, census takers would survey one of every 10 remaining households, then estimate the other nine households based on that sample.

For example, if the survey of a city census tract showed average household size of 2.6 members and 63 percent black population, then the statisticians would figure the remaining, uncounted households had the same characteristics.

Brookings Institution economist Charles Schultz, who chaired a National Academy of Sciences census advisory panel, warns that if Congress forces the Census Bureau to drop sampling, "two things would happen: The costs would be tremendous, and you could have a statistical fiasco."

The stakes are high: The census is used to apportion congressional seats among states, to draw state and local district boundaries, and to allocate more than $100 billion in federal funds.

The controversy over how to count the nation in 2000 began shortly after the flawed 1990 census ended. The 1990 head count cost more than ever ($2.6 billion, or about $25 per household) but still missed 1.8 percent of the U.S. population, half again as much as in 1980, the Census Bureau says.

For minority populations, it was even less accurate: It missed 5 percent of Hispanics and 4.6 percent of blacks. And the net undercount masks greater flaws: While 10 million people were missed, 6 million were counted twice, the GAO found. That left a net undercount of 4 million -- but a total of 16 million errors.

Based on a post-census survey, the Census Bureau proposed to adjust the 1990 count. The correction would have boosted Baltimore's population, for instance, from 736,000 to 772,000. But the Bush administration rejected adjustment, a position upheld last year by the Supreme Court.

Even the adjustment was flawed: It first proposed to strip one congressional seat each from Wisconsin and Pennsylvania and to give them to California and Arizona. But a computer programming error was discovered, and the Pennsylvania seat restored.

A Wisconsin Republican, Rep. Tom Petri, is leading the charge on Capitol Hill to oppose statistical sampling and adjustment.

"Adjusting the census is a little bit like saying, 'Let's stop having elections and just do a poll and say that's the winner,' " Petri has said.

Critics of sampling note that the Constitution requires an "actual enumeration" of the nation's population. Advocates say that the Census Bureau does count people but that it has always made last-resort estimates of households it could not reach.

"You can't conduct a really accurate census in 2000 if you rely on traditional procedures" of trying to count every household, says Robert B. Hill, director of Morgan State University's Institute of Urban Research and chair of a census advisory committee on the black population.

If only Americans would mail back their census forms, sampling would not be an issue. But participation has declined. In 1970, 85 percent of the public responded to the mailed census questionnaire without prompting. In 1990, only 63 percent did.

Two-income families, single-parent households, language barriers and distrust of government all make reaching and counting people more difficult than ever.

Finding capable census takers is challenging; women who once took such jobs now often work full time. Some census takers won't venture into tough, hard-to-count neighborhoods. Statisticians suspect some "curbstone": make up data without knocking on doors.

The Census Bureau plans to rectify 1990's failings by compiling better address lists with local government help; spending $100 million on advertising to entice people to return questionnaires; making "user-friendly" forms widely available; and sampling the population that has been missed.

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.