Enumerate-gate
BARONE, MICHAEL
Enumerate-Gate by Michael Barone The fight over the 2000 Census is a fight between the Republicans and the Democrats over statistics. It reached the high level of national politics when the...
...Not as much as some Republicans fear, but probably enough to cost them the House if the vote in 2002 is the same as in 1996...
...The one number that Riche and the other statisticians have been reluctant to say out loud is the error margin for adjusted results at the block level—the smallest geographic unit for which results are reported...
...They assigned each state a certain number of representatives for the first Congress in what they thought was rough proportion to population...
...They knew that estimates could and would be politically manipulated and that an enumeration, though it would not be perfectly accurate, would anchor would-be manipulators more closely to verifiable facts...
...We get the best response to someone in the community," Riche says...
...The framers were familiar with population estimates, and they required "actual enumeration...
...It is this White House's standard operating procedure to dismantle, at the INS or the Democratic National Committee, verification procedures that get in its political way...
...One will be an enumeration, much like that in 1990, in which census workers will attempt to get responses from mailed questionnaires and in-person follow-ups...
...they are aggregations of blocks, often with grotesque shapes, sometimes not even contiguous, stitched together to suit the political needs of redis-tricters...
...Today's statisticians, including the Clinton administration's director of the Census Bureau, Martha Farnsworth Riche, don't think much of actual enumerations...
...While Riche's professional credentials may well entitle her to a presumption of good faith, they provide no assurance that she has the political shrewdness and clout to keep White House political operatives from getting their way...
...Riche emphasizes that the independent survey of 750,000 requires many fewer—and thus higher-quality—interviewers...
...Riche and the statisticians insist that block-level error doesn't matter because it becomes much smaller when blocks are aggregated into Census tracts, counties, and states, and because nobody uses block-level data for anything anyway...
...many of them will not be professional interviewers and will therefore be prone to mistakes...
...In the 1990 Census, five different adjustments done by sampling would have produced five different apportionments, with 11 states either gaining or losing a representative...
...but no other country had ever had a regular census every so-many years, and no other country had thought to apportion political representation according to population enumerated in a census...
...At least one statistician understands the political risk...
...Riche's plan thus offers Clinton-Gore operatives a tempting opportunity: get politically savvy people hired as local enumerators, especially for the second, independent survey...
...The Census Bureau has to hire more than 1 million people to complete the enumeration...
...Dennis Hastert, whose subcommittee has oversight of the Census, is right when he calls the bureau's current plan "a risky scheme ripe for political tampering...
...The framers called for "the actual enumeration" of the nation's population every ten years—literally every body in America—to be used to apportion the members of the House among the states...
...The architecture and animating spirit as well as the words of the Census clause are very much on the Republicans' side and against census adjustment...
...But block-level data won't even out if error is systematic—if the independent survey has been manipulated to overcount Hispanics, for instance—and anyone who says that block-level data are not used for anything has never practiced political redistricting...
...Statisticians can't give you one way to make an adjustment that all other statisticians will agree is the best way...
...This is a White House that had no scruples about getting the Immigration and Naturalization Service to drop criminal checks on applicants for citizenship so that more Democrats could be naturalized in time for the 1996 election...
...Give me the ability to overcount Hispanics, and I can create two new heavily Democratic congressional districts in Los Angeles County and take those seats away from voters somewhere else...
...Michael Barone is a senior staff editor at Reader's Digest and co-author of The Almanac of American Politics 1998 (National Journal/Times Books...
...The ability to 'create' or 'eliminate' millions of strategically placed citizens with the stroke of a pen introduces a potent and disturbing new political weapon...
...but INS commissioner Doris Meissner said the same thing, and probably meant it, too...
...I have, and I can tell you that districts are not enduring entities, made up of randomly chosen blocks...
...The Census Bureau has only just come out with a strategy for adjusting the 2000 Census...
...Sampling, says David Murray, head of the Statistical Assessment Service, "creates a powerful temptation for the party in power to skew the adjustment its way...
...But they wanted all subsequent Congresses apportioned by "actual enumeration...
...They are not just trying, as many Republicans think, to put a pretty face on a Carvillean strategy to rob Republicans of representation...
...How much effect would political adjustment of the Census have...
...The statisticians are, I think, sincere...
...It reached the high level of national politics when the Republicans added a provision to the disaster-relief bill in June barring the use of "sampling"—estimating the total population from an actual count of a statistically random sample—and the Democratic president cited the provision as one of his reasons for vetoing the bill...
...as he told the National Council of La Raza convention, "legal immigrants should not be discriminated against, and told they are not first-class citizens of the United States...
...That is one reason statisticians hate enumerations...
...But different adjustments will inevitably produce very different political results...
...It's easy to see how the bureau's results can be manipulated politically...
...John Morgan, Newt Gingrich's favorite district-by-district political analyst, has said that Democrats could pick up 20 seats by Census adjustment...
...If the independent survey determines (as the Census Bureau survey did after the 1990 Census) that young black married women in cities like Chicago and Detroit were undercounted, it will add numbers of young black married women in similar cities all over the country...
...And there will be a natural tendency for statisticians to prefer a method of adjustment that seems likely to favor their own political party...
...That means that a block whose reported population was 100 could actually contain 72 people or 128, nearly twice as many...
...Even the 1.8 percent figure is itself an estimate, since no one knows or can know precisely what the total population really is...
...But the statisticians think that they can produce a more accurate figure through sampling...
...The political consequences are obvious: It makes a difference whether a seat is lost by Kansas (whose representatives today are all Republicans) or by Massachusetts (all Democrats...
...The reason is that it is very high: In a 1995 Census Bureau test, the error margin at the block level, at the 95 percent confidence limit (at which results of political polls are usually reported) was plus or minus 28 percent...
...each will give you results that are defensible on statistical grounds...
...then watch as the computers magically adjust the population of Democratic blocks up higher and higher...
...Essentially the bureau plans to conduct two surveys and "weave them together," as Franklin Roosevelt once told subordinates to do with two inconsistent programs...
...This was an extraordinary proposal...
...why would it suddenly develop scruples about adjusting Census numbers for political purposes...
...I think that's too high, but the number is well above zero...
...The 1990 Census, they argue, produced a 1.8 percent undercount, with significantly higher undercounts of Indian-reservation residents, Hispan-ics, and African Americans...
...The Census is used for lots of other things today—for allocating federal aid funds, for guiding commercial marketers— but it is in the Constitution for only one reason, and that is to apportion the House of Representatives...
...There are any number of methods of adjustment...
...Who will conduct these surveys...
...The other will be an independent, statistically valid survey of 750,000 respondents at 25,000 locations across the country...
...encourage them to count as many people as possible in high-undercount (and high-Democrat) segments of the population...
...Riche says that she will insist on the integrity of the process...
...But she also says that in high-undercount areas—Indian reservations and black and Latino neighborhoods—the bureau has hired local people to do the enumerations...
...it will conduct a "dress rehearsal" of the strategy in three places— Sacramento, rural South Carolina, and a Wisconsin Indian reservation—in April 1998...
...When I worked for pollster Peter Hart, he refused to print any numbers on which the error margin was more than plus or minus 8 percent, on grounds that their spurious precision would mislead our clients...
...It's not hard to see why House Democratic leader Dick Gephardt strongly supports census sampling and is quite happy to see Democratic ranks swelled by imputed numbers of even non-citizens...
...Even in the best of faith, we all (as responses to Clarence Thomas and Anita Hill showed) tend to break ties in favor of the home team...
...but unlike in 1990, attempts will stop after census-takers reach an estimated 90 percent of the total population...
...There had been censuses before, in the time of Caesar Augustus even...
...Which is another way of saying that I think the framers of the Constitution had it right and the statisticians have it wrong...
...But I think the Republicans are right to oppose statistical adjustment of the Census...
...The results of the enumeration will then be adjusted by the results of the independent survey...
...But the Census battle is just as much a clash between the framers of the Constitution and today's professional statisticians...
...And she adds, "We will keep the best of these for the second survey...
Vol. 2 • August 1997 • No. 47
