Everyone concerned with the already controversial 2000 census seems to prize scientific accuracy—but for the most part, strangely enough, only selectively. In the name of greater accuracy, liberals and the Census Bureau favor making a statistical adjustment to the headcount to correct for the predictable failure to reach all black Americans and other minorities, while conservatives object to this departure from past practice (which, as it happens, would aid Democrats and hurt Republicans). But their positions are reversed when it comes to certain other ways of enhancing the accuracy of the census: namely, adding a "multiracial" category to more accurately reflect the condition of persons of mixed ancestry, or— more radically—getting rid of all the classifications based on the unscientific concept of "race."

This last proposal, however rational, would indeed be a sharp break with the past. "Asking about race or color is an old American tradition," notes Brandeis University historian Lawrence H. Fuchs, author of The American Kaleidoscope (1991), writing in a Society (Sept.–Oct. 1997) symposium. "It goes back to 1820, when the census counted ‘free colored persons.’ In 1850, it subdivided color into ‘white, black or mulatto’ for free inhabitants. Slaves were not counted and were assumed to be black." Twenty years later, "the word ‘race’ was used for the first time," with the racial categories being "white," "colored," "Chinese," and "Indian." The categories have changed over the decades. The current ones—"white," "black or Negro," "Asian or Pacific Islander," and "American Indian," as well as "other race" and "Spanish/Hispanic origin"—were adopted by the Census Bureau in 1980. Despite the fact that race has been thoroughly discredited as a biological concept by scientists, racial categories remain in the census, Fuchs notes, for a simple reason: "to implement affirmative action programs."

Yet the existing racial categories have offended many of the increasing number of interracial families in the country. The number of children in families with one parent white and the other black, Asian, or American Indian more than tripled between 1970 and 1990, rising from fewer than 400,000 to 1.5 million. Six percent of black householders in 1990 had nonblack spouses.

From organized groups of these families came the push in recent years to add a "multiracial" category to the 2000 census. Census Bureau surveys indicated that only a tiny percentage of people who classify themselves as black would shift to the multiracial category. Even so, civil rights leaders viewed the proposed classification "as a wrecking ball aimed at affirmative action," writes Lawrence Wright in the New Yorker (July 25, 1994). In early July, a federal government task force nixed the multiracial category, asserting that it would stoke "racial tensions and further fragmentation of our population." Instead, the task force suggested that people of mixed ancestry be allowed to select more than one of the existing racial categories on the census form. This approach, Judith Lichtenberg and three University of Maryland colleagues approvingly write in Report from the Institute for Philosophy & Public Policy (Summer 1997), "would allow lighter-skinned blacks, or people with one black parent, to opt out of an exclusive identification as black if they wished, but it would not give official status to a new, competing affiliation."

The government is expected to adopt this recommendation this fall, but Orlando Patterson, a sociologist at Harvard University, argues in a New York Times (July 11, 1997) op-ed piece that the "more than one race" approach is no better than the "multiracial" one and "will only intensify and reinforce our misguided obsession with racial categories." In his view, the Census Bureau should simply cease classifying Americans by race. "After all, why do we need a ‘race’ category when we already have an ‘ethnic’ one on the forms? . . . Distinguishing between race and ethnicity is an ingrained part of America’s racial ideology. The racial categories maintained by the Census Bureau can only perpetuate the idea that there is such a thing as racial purity and that people in the United States have essential biological differences."

Patterson’s fellow Harvard sociologist Nathan Glazer wonders in the New Republic (Oct. 7, 1996) why the Census Bureau "[has] gotten so deeply into this business of trying to make ever more refined racial and ethnic classifications at all," devoting more than two-fifths of the most recent census short form to racial and ethnic questions. He thinks there is far more such "counting" than necessary.

"Race in America means blacks, as a result of our long, sad, history, and it is of course blacks for whom the numbers are important, to rate our progress, or lack of it, in achieving equality," Glazer writes. "That is the only race that counts. There is no need to count Asian groups to the nth generation. In any case, intermarriage will make the effort futile in a few decades. There is no need to count American Indians, either. The figures have been inflated wildly in the last two counts as people with only a fraction of Indian ancestry call themselves American Indians. And there is no need to record ‘Hispanicity,’ as if it were an indelible mark impervious to American assimilation."

Ideally, in Glazer’s view, the census would ask only three questions about race or ethnicity: "Are you black? Where were you born? Where were your parents born? That would tell us all we really need to know or can know with any degree of accuracy."

If an accurate count of black Americans is of signal importance, then so is the undercount. The Census Bureau and others contend that a statistical adjustment can be reliably made to correct for the large number of blacks—an estimated 5.7 percent in 1990, compared with 1.3 percent of whites—not counted in the census. The uncounted blacks are mostly in urban neighborhoods with high rates of poverty, crime, and drug abuse.

The Census Bureau has known about the relatively high undercount of blacks since the 1940 census, but only in recent decades, with congressional and state legislative redistricting affected, as well as the allocation of billions of federal dollars to state and local governments, has it become a significant issue. The undercount has prompted controversy and litigation in connection with both the 1980 and 1990 censuses, and promises to do so again with the 2000 one. The Supreme Court laid the 1990 lawsuits to rest last year without deciding the constitutionality of a statistical adjustment, note Margo Anderson, a historian at the University of Wisconsin–Milwaukee, and Stephen E. Fienberg, a professor of statistics and social science at Carnegie Mellon University, Pittsburgh. That, they write in Society (Mar.–Apr. 1997), means that the Court could later construe the Constitution’s reference to an "actual enumeration" as prohibiting a statistical adjustment.

Such a ruling would be fine with Michael Barone, a senior staff editor at Reader’s Digest and co-author of the biannual Almanac of American Politics. He does not question the sincerity of the statisticians who believe an adjustment would improve the accuracy of the census. But he thinks that it would enable politicians to manipulate census figures with relative ease, and that Republicans are right to oppose it. The Framers of the Constitution, he writes in the Weekly Standard (Aug. 11, 1997), "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. The architecture and animating spirit as well as the words of the [Constitution’s] census clause are very much on the Republicans’ side and against census adjustment."