Commentary: Census 2000: Math, Not Politics, Please

During campaigns, politicians bet their careers on the reliability of polls. But now that the Census Bureau wants to use the same statistical sampling techniques to improve the accuracy of the 2000 Census, irate Republicans want to block the practice in a move that is both bad policy and shortsighted politics.

Getting 290 million people to respond to official surveys is impossible in an age when citizens are distrustful of government and already buried under a blizzard of junk mail. In 1990, only 61% of households bothered to answer a census questionnaire they got in the mail.

The GOP leaders are fixated on politics, not math. Census data are used to redraw congressional districts, and Republicans insist the Clinton Administration will manipulate the results to reduce the number of GOP House seats.

FEAR OF THE POOR. But the GOP's real fear is that a precise tally may more accurately count poor people and minorities, who tend to vote Democratic. Thus, the GOP demands Census only count heads, even though it knows that doing so will mean that many heads--of the homeless or families doubled up illegally in apartments--will be missed.

This could backfire for the GOP. In 1990, for example, one of the most undercounted groups was not urban minorities, but the rural poor, hardly a Democratic bloc. Also, many of the underrepresented minorities overlooked in an old-style head count would be Hispanics--who are receptive to the GOP's conservative social agenda.

An inaccurate census would also punish business. The census provides a wealth of demographic detail that companies depend on for marketing and planning. The housing industry, for instance, gets information on everything from population density to trends in the number of bathrooms in homes.

The 1990 census, which missed 1.6% of the population, showed that mailing surveys and banging on doors just wasn't working. So the Census Bureau came up with a new approach for 2000--aiming for a total response rate of 90%. It would mail brief questionnaires to 82% of households and send a more detailed survey to the rest. A sample of all those that do not respond would get a visit from a census taker. The biggest change: Census would survey a massive sample of 750,000 households to check the accuracy of the traditional count.

Not if House Republicans can help it. It has become an article of faith among some GOP members that if sampling were used, their party would lose up to 30 seats in a post-2000 redistricting. Much of this thinking emerged after GOP consultant Clark Bensen released a study on the subject in 1996. But while Bensen did predict that as many as two dozen districts would be affected, he acknowledged it was impossible to figure out whether the changes would help Democrats or Republicans.

A BOTCHED COUNT? Nonetheless, Republicans want to squelch sampling in a Census spending bill soon to reach the House floor. They would bar Census from even planning for a sample in 2000 and would prohibit the agency from spending three-fourths of the survey's $382 million budget until Congress approves its methods.

The longer Congress squabbles, the greater the chance for a botched count. The National Academy of Sciences, National Research Council, American Statistical Assn., and Congress' own watchdog, the General Accounting Office, all consider sampling credible. Republicans should back off and let the Census Bureau do its job.

Before it's here, it's on the Bloomberg Terminal.