2000 Ad

`Most people understand that a poll taken before an election is a statistical sample," said President Clinton last month in Houston. "And sometimes it's wrong, but more often than not it's right." With that, the poll-driven President reiterated his support for a poll-driven census.

If President Clinton has his way, the 2000 Census will be the first in our history in which the census-takers do not attempt to count directly every person living in this country. Instead, they will count 90 per cent of the population and attempt to "find" the remaining 10 per cent (about 27 million people) through sampling, a complicated technique that involves computer models, statistical projections--and a lot of guesswork.

This isn't just a complex technical problem. It's a political matter with potentially enormous consequences. Sampling could cost the Republicans their majority in Congress by turning more than twenty congressional seats now held by Republicans over to Democrats in the next decade. "It amounts to the largest gerrymander in American political history," warns a GOP memo. If Republicans hope to control Congress for the next generation, this is a fight they cannot lose.

But they certainly aren't winning. Last year, President Clinton vetoed a flood-relief bill when Congress attached to it a rider that would have prohibited the Census Bureau from sampling in 2000. Republicans thought Clinton wouldn't put sampling ahead of disaster victims. Their plan backfired, and they were left with their worst public-relations disaster since the 1995-96 shutdown of the Federal Government.

Republicans in Congress intend to take another shot at sampling this fall, when the Census Bureau will seek approval of its 1999 budget. Meanwhile, other opponents of sampling are working on the judicial front--two law-suits challenging the Bureau's plan are currently under way--but few observers expect the courts to rule the technique illegal.

The census is always controversial because the stakes are so high. The Constitution requires the Federal Government to count the total population every ten years for the purpose of distributing seats in the House of Representatives. Congressional apportionment is a zero-sum game, and every census has its clear winners and losers. In recent decades, states in the Midwest and Northeast have watched their influence in Washington shrink as fast-growing states in the South and West have acquired a larger share of the 435 seats in the House. Experts predict the 2000 Census will shift another nine seats from the Rust Belt to the Sun Belt. Furthermore, states and localities receive roughly $170 billion per year from the Federal Government on the basis of their population.

No census has ever been 100 per cent accurate. People living in remote areas and the homeless are hard to count. Illegal aliens and others hiding from the law don't want to be counted at all. Despite these challenges, every census since the first one in 1790 improved upon the accuracy of its predecessor--until 1990. That year, the Census Bureau estimates that it neglected to count 8.4 million Americans and double-counted 4.4 million others, resulting in a net undercount of about 1.6 per cent of the population.

Identifying 98.4 per cent of a population of roughly 250 million people may not seem bad, especially considering that the public's rate of participation in the census hit an all-time low that year. In 1980, 75 per cent of households receiving mailed census forms returned them; in 1990, this rate dropped to 65 per cent. Also, in recent decades the Census Bureau has been contending with an increasingly mobile population, and a higher percentage of people living in unstable family structures. It's remarkable that the 1990 census performed as well as it did.

But not everybody agrees--especially because of the allegation that a disproportionate share of the undercount was among blacks and Hispanics. "The 1990 census failed both the public and Congress, and we cannot let that happen again," said Rep. Carolyn Maloney (D., N.Y.), ranking minority member on the newly created House Census Subcommittee. The Census Bureau and many Democrats claim that the way to prevent a similar undercount next time is by using a complicated method of statistical sampling. The plan basically calls for counting 90 per cent of the population in each of the nation's 65,000 census tracts by traditional methods. The final 10 per cent will be sampled. If a tract produces a response rate of 60 per cent, for example, census enumerators will contact three out of every four households in the remaining 40 per cent, in order to hit their 90 per cent threshold. The characteristics of the one out of four households deliberately ignored in the follow-up will be extrapolated using information from those that were actually reached.

Meanwhile, the Census Bureau will conduct an overlapping sample of 750,000 households and the results will be compared to the main census. When enumerators locate people who were missed by direct-counting methods--urban black women, for instance--federal statisticians will assume an undercount in the overall census and attempt a correction. The bureau will weigh more heavily the census forms received from urban black women in order to offset this group's alleged under-representation. For balance, other groups, such as white suburban homeowners, may find the weighing of their responses adjusted downward. Sampling essentially creates a population of fractional people, with some census forms being more equal than others.