Numbering the people

It has all the makings of a massive offensive: years of planning and training, high-tech logistics, the need for speed. Officials will recruit 3 million people to put 300,000 on the ground - six times the number of troops headed for Kosovo.

But these men and women aren't headed for war. Their only weapons are paper and pencil. Their objective: to count the people and paint the social and economic portrait of the United States.

Ever since 1790, when US marshals fanned out on horseback with tally sheets, America has counted its people once a decade. It's the world's most regular population survey and the nation's largest peacetime mobilization. In scale, Census 2000 will not disappoint.

Overseen here at the Census Bureau's World War II-era headquarters outside Washington, it will be the largest, most complex, and most expensive mobilization yet. Ironically, it may also prove disturbingly inaccurate.

Census 2000 packs such political dynamite - congressional seats and tens of billions of dollars in federal aid are at stake - that most groups can't resist poking a finger at the nation's population calculator. As the squabbling escalates, the bureau is moving closer to a statistical monster with two "official" population counts and a string of dissatisfied constituents.

"The census distributes money and power," sighs John Thompson, associate director for the decennial census. It's always been controversial, he adds.

Even in the best of times, counting people represents a daunting task. To get its figure, the Census Bureau sweeps through the country twice. First, it picks up all the nation's addresses - some 120 million in all - and checks them. Most of the bureau's regions are finishing up that process. Next spring, the bureau will send out those familiar surveys to find out how many people live in each household, what they earn, and a slew of other questions.

Households that don't mail back the survey - one out of four did not in 1990 - will be visited by census-takers who try to get the data in person. These temporary workers are the census's front-line troops. They walk through the ghetto and slip into ritzy gated communities. Some rural addresses they track don't have a street name, much less a number. Some people they track don't want to be found.

In northwest Arkansas last fall, for example, census workers were met by white supremacists toting shotguns. "They were told that they were not going to cooperate ... and they were told to get off their property," says Sara Duke, field manager for the Census 2000 office for Arkansas, based in Little Rock. So the workers got the data from nearby county officials, instead.

Such armed confrontations are rare. More common is the plight of Steve Choi. As founder of the Asian American Coalition in Atlanta, he's eager for the Census Bureau to fully count the area's burgeoning Asian population. But he faces an uphill struggle persuading his own constituents to participate.

Leave me alone

"They don't want to be involved in anything to do with government," he says. Thanks to their experience with dictatorial regimes back home in places such as South Korea and Cambodia, "they try to hide" when government workers come knocking.

Then there's the language barrier. Since many of the area's Asian Americans are first-generation immigrants, their English is limited or nonexistent, Mr. Choi says. So even if census-takers manage to find them, they're unlikely to draw out much information.

The Census Bureau is trying to help. It is printing census forms in four Asian languages (Chinese, Vietnamese, Korean, and Tagalog) and census guides in dozens of other languages. It's also hiring outreach specialists for various ethnic groups. And it promises for the first time to place paid ads to get the word out.

But some minority groups complain such efforts aren't enough. For example, the ethnic specialists are spread thinly around the country. Of the 65 such specialists assigned to the three-state region overseen by the bureau's Atlanta office, only one is Asian American. When Choi asked last month for a second to be hired to cover Miami's burgeoning Asian population, the office said no.

And though the bureau is putting out forms and guides in many languages, "they haven't come up with any way to tell people how to find the guide," complains Karen Narasaki, executive director of the National Asian Pacific American Legal Consortium in Washington, D.C. "They're relying completely on the community to do it.... And it's a daunting task."

Other minority groups also shy away from census-takers. Hispanics who are in the US illegally worry the Census Bureau will share its information with the Immigration and Naturalization Service.

Privacy concerns have dogged the census since its inception, but officials vehemently insist the data they collect about an individual remain strictly confidential. The bureau is waiting for clarification from the Justice Department that a new immigration law won't compromise that confidentiality.

African-Americans, meanwhile, are fighting hard in Congress to allow the Census Bureau to use sampling techniques to minimize the undercount of minorities.

"No one was pleased with the 1990 census," says Hilary Shelton, Washington bureau director for the National Association for the Advancement of Colored People. The Census Bureau admits it missed 8 million people and double-counted 4 million others - a net undercount of 4.8 million Americans. That marked a major change in the bureau's progress. Until 1990, the census's undercount had declined steadily since at least 1940.

Undercounting minorities

Minorities were affected the most from the 1990 undercount: 4.4 percent of blacks were missed, 5 percent of Hispanics, 12.2 percent of at least one tribe of native Americans. It's less clear who was double-counted. They were typically older Americans - and, as one theory puts it, wealthy people who could afford two homes.

Thus, for all the money spent on the 1990 Census, there's a distinct possibility it seriously overcounted the wealthiest Americans, who need government help least, and undercounted the poorest Americans who need it most.

"People wonder why the schools are falling apart in the inner cities?" Mr. Shelton says. One reason, he says, is that they don't get their fair share of federal tax dollars. For example, the 1990 Census missed some 11,000 poor students in Richmond, Va., meaning that the district lost more than $7 million a year in federal revenues.

And, since the census only occurs once a decade, "that's $70 million that didn't go into the Richmond school district because of inaccurate statistics," he adds. Adults got missed, too. So inner-city commuter rail stations and bus stops didn't get built because planners didn't know the people were there. And "those are the people who need a bus stop because they're the ones who don't have cars," Shelton adds That is why census officials have hit upon a curious idea: by taking a representative sample of the people who don't turn in their forms, the bureau will get a more accurate count than if it tried to conduct a fuller traditional count.

Conservatives, however, say statistical sampling is unconstitutional. The Supreme Court agreed. In a ruling earlier this year, it said the Census Bureau could not use the technique for reapportionment. Thus, when Congress doles out House seats next year, it will rely on a traditional count from the Census Bureau.

But the bureau still intends to use statistical sampling for everything else. So, while Congress will use one census figure to determine how many congressional seats Pennsylvania will lose, the Pennsylvania legislature may well use another census figure to determine how to carve up the state into legislative districts.

The battle over methods isn't over. The Republican-controlled Congress holds the purse strings and the bureau announced earlier this month it needs an extra $1.7 billion to complete the job. That would push the price tag up to $6.3 billion - more than twice what the 1990 census cost.

Such political battles over numbers aren't new. George Washington complained about the undercount of the first census in 1790. The 1920 census showed such a huge population shift to the cities and swelling immigrant population that Congress decided not to reapportion seats during the decade. But today's fight over sampling marks the most raucous debate yet over the bureau's statistical methods.

Although the Census Bureau puts out other population estimates and economic surveys, the decennial census is the only one that offers specific details down to the neighborhood level. "It's the local story that's really big in the census," says Larry Long, a demographer with the bureau's housing and household economic statistics division. "There will be lots of studies about income differences between urban neighborhoods." Another big area of focus, he adds, will be the flow of ethnic groups into rural and small-town America. "We are going to be surprised at the extent to which this is taking place," he predicts.

Another first: Census 2000 will reveal a fuller spectrum of Americans' ethnic heritage. Instead of forcing respondents to check only one ethnic background, now they can choose as many as apply: black, Chinese, Filipino, native Hawaiian, and so on, or write in any race.

Next year's census will also be the most technologically advanced. Rather than hiring operators to input the paper forms into computers, a scanning and character-reading system put together by Lockheed Martin will do most of the inputting. When the machines don't understand something - they will rarely read an entire form perfectly - human operators will stand by to read and input the data in question.

The system will have to be fast, because there will be days next spring when the bureau will get more than 6 million forms mailed back. Dick Taylor, the system's designer, says that during the bureau's dress rehearsal in three US cities last year, the systems scored better than human operators would have, with only a 1.3 percent error rate. The bureau has often spurred innovation. The 1890 census first used mechanical punch cards on a large scale. The 1950 census was tabulated by one of the first electronic computers.

"The big leap for 2010 will be the Internet" and a paperless form, predicts Mr. Taylor. "It may well be that this is the high-water mark in terms of paper."

But not in terms of controversy, if history holds true in the next millennium.

(c) Copyright 1999. The Christian Science Publishing Society