Charles L. Schultze is correct when he warns that the country is headed toward a failed census in 2000 ["Which Census for 2000?" op-ed, June 7]. But he is wrong to point to congressional reluctance to adopt the Census Bureau's unproven statistical estimation methodology as the reason. The reason is the Clinton administration's stubborn refusal to face the fact that its sampling plan won't work.
Congress, charged by the Constitution with responsibility for the census, is simply doing its job when it casts a skeptical eye on the president's unprecedented plan. The administration wants to alter 200 years of census-taking and replace the traditional head count with a statistical estimation method that is basically nothing more than a poll -- a fact confirmed by the president during his recent trip to Houston. Congress, on the other hand, wants an honest, reliable census in which Americans can have confidence. The burden of proof is on the Clinton administration to prove that its plan is motivated by more than Democratic Party "political" science. So far that proof is lacking.
Given all the controversy in which the bureau is embroiled, it comes as no surprise to see Schultze return to defend the beleaguered sampling plan. He is a well-known partisan Democrat who served in the Johnson and Carter administrations, and, as the co-editor of a pro-sampling 1995 National Academy of Sciences (NAS) report, his personal reputation depends in part on whether the method is used in the next census.
While the NAS report makes for interesting reading, it is unfortunately long on rhetoric and short on science. At a recent congressional hearing on the census, Berkeley statistics Prof. Philip B. Stark said: "I read the National Academy report; I found the evidence quite weak. There are no data on which the conclusions are based. There is no mathematical theory on which the conclusions are based." In conclusion, Dr. Stark suggested that those following the sampling debate "place more weight on the evidence than on the letterhead that the evidence is on."
What is even more surprising to me, as a former statistics professor, is that this panel could recommend an even greater dependence on sampling in the 2000 census despite significant evidence that a test of sampling in the 1990 census showed the method doesn't work.
The academy's report simply ignores evidence that shows sampling doesn't work. It is silent on a Census Bureau study that found that sampling is less accurate than a head count for cities and towns with populations of fewer than 100,000 -- the types of localities that make up the vast majority of American communities.
Nor does the report mention the fact that if sampling had been used in the 1990 census, Pennsylvania would have mistakenly lost a seat in Congress to Arizona. This error was not even caught until early 1992. How can Congress have faith in a method that can't even accurately allocate seats among the states?
Neither the academy nor Schultze mentions that in the bureau's 1990 sampling test, the adjusted numbers were so inaccurate that they couldn't even get the ratio of males to females right in most cities and counties -- despite the fact that demographers have known for decades that this ratio is a universally constant 51 to 49 percent across every age, culture or geopolitical group boundary.
Perhaps without fully realizing it, Schultze betrays what sampling is really all about. He writes that "without the use of sampling to complete the count, a range of special programs to track down the hard-to-find must be designed and integrated into the census process."
Let me get this right. If the bureau can't take the easy way out and use the crutch of statistical estimation in the census, it might actually have to go out and do the hard work the American taxpayer is paying it to do? The only conclusion I can reach is that the Clinton administration is purposely ignoring ways to reach every American in the census in favor of doing a statistical estimate based on partisan political science.
Census Day 2000 is less than two years away, and we are on a course toward failure. Unfortunately, Schultze, the National Academy of Sciences and other administration allies continue to ignore the evidence that sampling doesn't work. They continue to ignore warnings from the General Accounting Office and the Commerce Department's inspector general that the 2000 census is at risk.
The census still can be saved, but only if the Clinton administration decides it wants to save it. A good first step is a move toward a census process with more counting and less calculating.
The writer, a Republican representative from Florida, is chairman of the subcommittee on the census.
Comments on this posting?
Click here to post a public comment on the Trash Talk Bulletin Board.
Click here to send a private comment to the Junkman.
Copyright © 1998 Steven J. Milloy. All rights reserved on original material. Material copyrighted by others is used either with permission or under a claim of "fair use." Site developed and hosted by WestLake Solutions, Inc.
Material presented on this home page constitutes opinion of Steven J. Milloy.