^^ be used to apportion congressional seats, which is the Constitution's justification for having a census, but may be used for other purposes, including redistricting and the distribution of federal grants.

Such a process is inherently political, Peter Skerry argues in a forthcoming book, Counting on the Census: Race, Group Identity, and the Evasion of Politics.

Advocates of adjusting the census, Skerry says, should specify what degree of accuracy is acceptable.

"We might begin," he writes in a policy brief for the Brookings Institution, "by asking whether there is such a thing as a structural undercount, comparable to structural unemployment: a level of undercount below which we cannot expect a free society to go." Skerry is an associate professor of government at Claremont McKenna

College, and a senior fellow at the Brookings Institution, which will publish the book April 14. The policy brief is available on the Web at Brookings.edu (follow the Press Room link).

Two issues drive the debate over adjusting the census; one is that minorities are undercounted at several times the rate of whites, and the other is that the federal government hands out $185 billion, based, in part, on census figures.

But the fiscal stakes are actually much lower than they appear, Skerry says. The congressional General Accounting Office found that only $449 million of $136 billion (0.33 percent) would have been distributed differently among the states if adjusted figures from the 1990 census had been used for 15 big federal programs.

That's real money, of course, but compared with the total of all state budgets, it's barely round-off error.

The fraught question of minority undercounts also has less practical effect than the frenzied debate over it would suggest. The political effects of adding more minorities to the census are not easy to predict. A state or district may seem to increase its population by correcting an undercount, and still lose ground to some other jurisdiction where the correction was larger.

Redistricting after 1990 created a number of "minority-majority" districts that were safely Democratic, while enhancing the election prospects of Republicans nearby. In many districts where the undercount is high, the voter turnout is low; correcting the former does nothing to change the latter.

Republicans' reluctance to gamble on the consequences of adjusting the census, Skerry notes, "hardly justifies the utter certainty with which virtually all commentators assert that adjustments would help Democrats and hurt Republicans."

Even defining race and ethnicity will grow more difficult. The listing of "racial" categories on the 2000 census manages to be both nonsensical and offensive. Maybe by the time 2010 rolls around, we'll all have had our DNA mapped and we'll be able to tell the Census Bureau exactly where on the planet each of our chromosomes originated.

Failing that, however, group boundaries increasingly blur. People do not always identify themselves the same way; one Census Bureau study found a significant percentage of people who changed their response between the 1990 census and a subsequent survey. I've written before in favor of adjusting the census, but Skerry's description of the problems certainly persuades me I should take another look.

A small measure of inaccuracy may be better than the alternatives.

(660 words)