Judd Gregg At Commerce Could Keep Democrats From Cooking 2010 Census Numbers

By Michael Barone, Thomas Jefferson Street blog

Others have written on the Gregg appointment. I want to make one observation: the Commerce Department includes the Bureau of the Census. Ten years ago the Clinton administration attempted to use sampling instead of an actual headcount for the Census enumeration which is used for reapportioning the House of Representatives and for redistricting of all kinds. It was resisted by the Republican Congress, notably by Florida Congressman Dan Miller, who chaired the relevant House subcommittee. And it was finally abandoned after career Census Bureau statisticians, who like most professional statisticians prefer sampling to headcounts, conceded that they could not guarantee that the sampling procedures they proposed to use would be more accurate than a headcount.

I have expected Democrats to attempt to use sampling again, and Republicans no longer have majorities in Congress. But Gregg could exercise an important supervisory role here, by at least insisting on the kind of statistical rigor that Census Bureau career statisticians showed they had when the Clinton administration (in my view) contemplated cooking the numbers.

The idea of course was to draw on the fact that headcounts tend to undercount minorities (though the 2000 Census did so less than in the past) and to impute huge numbers of black and Hispanic residents, thus giving Democrats more congressional and state legislative seats.

An important thing to keep in mind when you consider the possibility for mischief is this. The margin of error for a sample of a large population unit is quite small, as the professional statisticians will tell you. But that margin of error increases as you measure smaller and smaller units, as the professional statisticians who favor sampling will admit when you press them. Redistricters build districts out of Census tracts (which may have 5,000 or so people) and Census blocks (which may have anywhere from a dozen to a couple of hundred). If you have a sampling procedure that systematically overestimates the presence of one particular kind of groups (let's say Hispanics), then the sum of the errors in large numbers of heavily Hispanic blocks could be very big--and could be worth another Democratic state legislative district (one to be sure with very few voters). I've done redistricting myself and have been intimately familiar with congressional redistricting over the last 45 years, and the possibilities of distorting the process by sampling are great. If Judd Gregg can head that off, maybe it's worth his leaving the Senate.

By the way, the last time the Census was in the hands of one political party--that is, when one party held the White House and majorities in Congress in the year when the Census was conducted (and thus in the year or so before it was conducted), was in 1980, when sampling was not a serious issue. The last time the Census was, in this sense, in the hands of the Republicans was in 1930. The reapportionment following that Census (the first in 20 years, since after the 1920 Census the Congress in defiance of the Constitution refused to reapportion House districts) combined with the economic upheaval of 1929-32, produced a huge gain for Democrats--and the only House since the one elected in 1896 in which most of the members were freshmen.

-- Read more by Michael Barone.

-- Read more from the Thomas Jefferson Street blog.

-- Read more about Judd Gregg.

By Michael Barone