Stephan Thernstrom is Winthrop Professor of History at Harvard and a Senior Fellow at the Manhattan Institute. He and his wife Abigail are editors of Beyond the Color Line: New Perspectives on Race and Ethnicity, to be published by the Hoover Institution Press in June 2000.

"The United States is the only country in the world in which a white mother can have a black child but a black mother cannot have a white child."

Not long ago, when the civil rights movement was unquestionably on the side of the angels, liberals were fond of pointing that out with a mixture of dismay and smugness. How stupid! they said. You were deemed black if either of your parents were. Indeed, you were deemed black even if your only black ancestor was one great-grandparent — the infamous "one drop" rule made it so.

That rule was essential to the either-or logic of the Jim Crow system, in which being part-black was as inconceivable as being part-pregnant. Nor was there room in this scheme for people with other racial identities; those who were not white were thrown into the category of — as the Census Bureau traditionally termed it — "Negroes and other races." Under this system, if you recognized the remarkably mixed ancestry of a Tiger Woods, what public school would you assign him to? Where would you have him sit in the bus?

Jim Crow is now dead, but its legacy lives on in current racial-classification practices. Ironically, though, it is now those on the left, who are pleased to call themselves liberals, who insist that we all belong in rigid, mutually exclusive, color-coded compartments. Their views, alas, are reflected in the decisions recently made about the Census of 2000, now in mailboxes throughout the country.

In the past, the Census Bureau was not been totally oblivious to the reality of the American marital melting pot. To the question about "ancestry or ethnic origin" that appeared in both the 1980 and 1990 censuses, it was allowable to give as many as three answers. You could report your ancestry as German, Italian, and Irish, or as Polish, Greek, and Swedish. When it came to identifying yourself racially, though, no such latitude was given. The "races" of the United States were assumed to be mutually exclusive categories, and you could only name one.

This rigid rule has come under growing criticism in the past several years, because of the spectacular surge in intermarriage between Americans of different "races." In some Asian groups — not so long ago categorized as "Orientals" and forbidden to marry non-Asians in some states — a majority of current marriages are to non-Asians, mostly whites. Fully half of all American Indians today are married to non-Indians.

Even black/white intermarriages, once illegal in all southern states and taboo in all but the most cosmopolitan communities, are by now far from uncommon. More than one in eight African Americans who married in 1994 had white spouses, and figure is doubtless higher by now. The rapid growth in the number of children born of racially mixed unions, and the resentment of their parents at being forced to put the child in just one racial box, forced the planners of the 2000 Census to recognize the new reality. It would now be possible, they announced, for racially-mixed Americans to identify with more than one racial group when they filled out their forms.

In the in-fighting that led up to the decision, the battle to preserve the old, mutually exclusive conception of race was led, not by white dinosaurs, but by the major civil rights organizations. If a significant proportion of African Americans reported themselves as partly white, they worried, it could mean fewer affirmative action slots at colleges and universities, fewer jobs for black applicants, a smaller share of public contracts set aside for black-owned businesses, fewer safe black seats on city councils, in state legislatures, and in Congress. All of these programs take proportional representation as the goal, so a decline in the total numbers of a minority group means a reduced entitlement.

In a 1995 Vanderbilt Law Review article, an African American attorney pursued the implications of the new "civil rights" position on this issue to its appalling but logical conclusion. Since race was now being used to award societal benefits, and properly so, the author argued, it was necessary to devise new legal tools to prevent "racial fraud." Whites would have a much better chance of getting into the school of their choice or finding a desirable job if they claimed they were black.

To prevent that, the author called for a stricter racial-classification system for all Americans, with "fines and immediate job or benefit termination" for those who "falsified their racial identity." What was needed was a new body of law that recognize d the "permanent importance" of racial divisions.

Of course, many of our states had such laws before; these laws were essential to the Jim Crow system. For example, a 1924 Virginia statute required statewide racial registration and provided a year in jail as the penalty for making a "false" report of one's racial identity. Obviously, this was designed to discourage blacks from attempting to "pass" as white. One of the arguments advanced by the plaintiffs in Plessy v. Ferguson was that Homer Plessy was entitled to sit in the "white" railroad car because he was white, being only one-eighth black. But the Supreme Court was not impressed, declaring that it was up to the state of Louisiana to decide "the proportion of colored blood necessary to constitute a colored person." To be one-eighth black in Louisiana was to be black period, and that was that.

It is breath-taking to find old ideas about "blood" now resuscitated in a respectable law review, and by an African-American scholar. Now the fear, apparently, is that some whites will try to "pass" as black. We have come a long way from the days of the Brown litigation, when Thurgood Marshall insisted that "classifications and distinctions based upon race or color have no moral or legal validity in our society."

Fortunately, the attempts of the civil rights organizations to preserve racial purity were unsuccessful. The Census Bureau agreed to modify the instructions that accompanied the race question on the 2000 Census, and to indicate that it was possible to choose two or even more racial identities. Homer Plessy could say that was both white and black, if he chose. Racially mixed Americans would be counted as they wished to be counted.

Counted in the simple sense that they may record their perceptions of their own racial identity. But what they report themselves to be, it has now become clear, will be of no practical import. The Office of Management and Budget has just issued guidelines governing how all federal agencies are to employ the racial and ethnic data gathered on the 2000 Census. Those guidelines reveal that the civil rights lobby lost the battle but won the war. When it comes to figuring out the racial composition of the population for purposes of "civil rights monitoring and enforcement" (i.e., to set quotas, goals, or targets in education, employment, and public contracting and to determine whether racial minorities can claim that their right to political representation has been "diluted), the complex identifications of racially-mixed Americans will be ignored.

The Bureau will not tabulate data for every possible racial mix reported by respondents. But they are keeping track of a mind-boggling 63, so that we will have figures, for example, on the number of people who consider themselves part-white, part-black, and part-American Indian. But when the figures are to be used to determine whether civil rights laws have been violated, these distinctions will disappear. Citizens who say they are not only white but also a member of a "minority race" will be counted as belonging to the "minority race" — period.

In other words, for these purposes, it is still the case that "a white mother can have a black child but a black mother cannot have a white child." If you have enough black or Asian "blood" to mention it on the census form at all, you will be put in the black or Asian column that the Equal Employment Opportunity Commission and similar agencies will be using to decide whether some company has "enough" employees of the right kind.

This decision means that civil rights groups need not fear that the number of their constituents will be reduced. Indeed, they will grow, if Americans of mixed race who reported themselves simply as white on past censuses now choose to mention that they have black ancestry as well. That mention will put them in the black column for purposes of civil rights enforcement.

The most dramatic, indeed bizarre, effect will be on the enumeration of the American Indian population. A mere 2.1 million people checked the "American Indian, Eskimo, Aleut" box on the 1990 Census race question. That is the figure everyone uses for the Indian population in that year. But the separate question about "ancestry or ethnic origin" on the same census, on which multiple choices were possible, yielded a staggering 8.7 million people who claimed to have some Indian ancestors.

If all of the additional 6.6 million identify themselves as both Indian and something else on the 2000 Census race question, now that multiple race choices are possible, the Indian population figures used for civil rights enforcement purposes will suddenly quadruple. And most of these born-again Indians will be thoroughly assimilated Americans who have no contact or cultural connection to any Indian tribe today, but they will now count as if they had never left the reservation.

Some of us hoped that making it possible to give multiple answers to the census race question would reveal a picture of such intricate complexity that it would call into question the very effort to divide us all into racial boxes. Sadly, our federal government today is attempting to prevent this fate, by retabulating the raw information in a manner that seems designed to insure the "permanent importance" — as that Vanderbilt Law Review contributor put it — of race.

Deplorable though this is, it is too simple to say flatly that the remedy is simply to stop gathering any information at all about the racial and ethnic composition of the population. I have been strongly tempted to take the abolitionist position on this issue, and respect Ward Connerly and others who do. I am certainly disgusted to see that the five out of six Americans who have been sent the census short form are asked to provide virtually no information except what racial box they belong in. Nevertheless, I have regretfully concluded that it is impractical and unwise to go the extreme that many of my comrades advocate.

Why? As early as the middle of the nineteenth century, the United States made its decennial census into something more than a simple head count of the population. The 1850 Census was the first to inquire into such matters as who had been born abroad and in what country, the occupations Americans were engaged in, and the school attendance of their children. Subsequent enumerations added questions about income and unemployment, housing, modes of travel to work, and dozens of other topics. All modern societies have felt compelled to compile such information, not least because businesses find such information extremely useful. The libertarian view that none of these matters are a proper concern of government is hopelessly utopian.

Even if all racial and ethnic questions were removed from the census, surely we would continue to hear complaints that blacks are far more likely than other Americans to live in poverty. Indeed, in the absence of authoritative data, the magnitude of such social problems would likely be even more exaggerated.

Furthermore, it is only by gathering the facts that we are able to discern the close connection between black poverty, the remarkably high proportion of female-headed black families, and the huge racial gap in levels of literacy and numeracy. Surely we want to know if black third-graders are far behind whites in reading and math competence, and the correlation between that difference and family structure or economic circumstances. It is at least possible that social science knowledge will guide us towards solutions.

The problem is not with the gathering of social and economic facts. The problem is what might be termed, "the privileging of race." The idea that mankind is divided into a very few, physically and culturally distinctive, racial groups is a thoroughly outmoded and discredited nineteenth century notion.

In setting out the officially-recognized racial categories, the Office of Management and Budget cautions that "these classifications should not be interpreted as being scientific or anthropological in nature." That's for sure — but the admission only makes OMB's decision to utilize them more dubious. If these categories are not "scientific" or "anthropological," what are they? Why should the U.S. government distinguish some citizens from others on a basis that is not "scientific" or even "anthropological" (whatever that means) and use those distinctions in allocating public resources? "Race" is a bad idea whose time has come and gone. It does not belong in statistics that appear with the stamp of approval of the government of the United States.

Ethnicity (a concept that includes all so-called racial groups), however, is another matter. We cannot pretend that African Americans are not a distinctive group with an uniquely oppressive history they are still struggling to overcome, nor that Chinese Americans or Ukrainian Americans are like everyone else in every way. But we need not assume that Americans of Chinese origins are a separate "race", while classifying Ukrainians as "white." The census question on "ancestry or ethnic origin," though, is quite sufficient to identify subgroups of the population, and it will do the job nicely — with one important modification.

In the past, the Census Bureau has framed the instructions on the ancestry question in a manner that discourages the response "American." Despite that discouragement, 13 million people — 5 percent of the population — said that their ancestry was "American" or "United States" on the 1990 Census, and another 27 million — 11 percent of the population — said that were unable or unwilling to specify any ancestry. If it were made clear that "American" is a perfectly acceptable response, those numbers might soar. That would not be a bad thing at all.