April 12, 2006, 7:18 a.m.
2,000 Flavors and Counting
The more race categories you add, the more pointless they are.

By Peter Kirsanow

Discussions concerning racial categorization almost inevitably devolve into the surreal. And so it was at last Friday's U.S. Commission on Civil Rights briefing on the 2010 census. Although the next enumeration is four years away, the U.S. Census Bureau is conducting a series of test censuses to "improve" the census questions pertaining to race and ethnicity. "Improvement" in this regard means more precise and accurate racial/ethnic data that may better inform the governmental programs and policies that rely on such data.

The problem with any such data collection is that classifications based on race and/or ethnicity are inherently arbitrary, unscientific, and inaccurate. Consequently, their utility in informing government programs and policies is, if not marginal, then at least suspect.

The problem isn't the Census Bureau. The agency does a remarkable job collecting and disaggregating mountains of information (at a cost of over $10 billion) pursuant to directives received from Congress and the Office of Management and Budget (OMB). The problem is that race and ethnicity are extraordinarily malleable concepts that evade precise biological, anthropological, or sociological definitions. The law, on the other hand, has been remarkably resourceful in defining race — sometimes hilariously, frequently illogically, but too often tragically so. (See my NRO article "Black Like Me.")

The original census in 1790 had but three racial categories: White, Black and Indian. These categories satisfied the Article I, Section 2 requirements pertaining to the apportionment of representatives and direct taxes among the states. The number of classifications expanded in the late 1800s to account for the increased Chinese and Japanese populations. The number of categories remained relatively stable throughout most of the twentieth century. By 1978, there were still only four "race" categories (White, Black Indian or Alaskan Native, Asian or Pacific Islander), as well as one "ethnicity" category (Hispanic). Since then the categories have multiplied rapidly. By the 2000 Census respondents could choose from 126 categories.

The rapid proliferation of racial and ethnic classifications does little to dampen suspicions that the categories are, at a minimum, arbitrary — and probably specious: Someone may have been Black of Hispanic origin in 1990, but today that person might be Cuban of "some other race." The accuracy of longitudinal comparisons becomes, therefore, more questionable.

Moreover, why limit the number of categories to 126? Anthropologists and sociologists may not be able to agree on the exact number of ethnicities, but the lowest figure seems to be around 2,000. Any number below that has less to do with demographic accuracy than with politics. For example, at one point OMB considered whether "Cape Verdeans" should be included as a separate category on the census. A quick tabulation of the number of Cape Verdeans in the country shows that approximately two thirds of the states have fewer than 100 and several have none. Further, it turns out that Cape Verdeans are themselves an amalgam of races and ethnicities: Approximately 78 percent are Creole ; 21 percent are African; and 1 percent are from various European subgroups. The agency decided against inclusion.

The utility of census racial/ethnic classifications in providing data for the government programs and policies that the classifications purport to inform is selective. Outside of antidiscrimination statutes, few governmental programs make specific reference to race. Rather, the types of programs that employ racial and ethnic data are generally those designed to address disparities in, e.g., education, income, housing, health care, etc. — disparities that are a function of factors for which race/ethnicity are imperfect proxies. Even the collection and application of racial and ethnic data for civil rights and antidiscrimination enforcement are inconsistent. Consider: Title VII of the 1964 Civil Rights Act prohibits employment discrimination based on race, sex, color, religion, and national origin. Yet the census seeks no data on color or religion. Nor does the census collect precise protected-class information directly applicable to other civil-rights laws. (It was suggested at the hearing that the census doesn't inquire into religious affiliation because the topic is too controversial. Perhaps. But it's unlikely Tocqueville or Myrdal would concur that religion is prohibitively more controversial than America's dilemma.)

There was no evidence adduced at the hearing that any specific governmental program or policy has been improved demonstrably by the atomization of racial and ethnic classifications in the census. Ward Connerly, one of the witnesses, mischievously advocated that a multiracial category be added, for the express purpose of causing the whole regime to collapse upon itself. But even Connerly acknowledged that in today's hyper-racialized society, that's unlikely to happen any time soon.

— Peter Kirsanow is a member of the U.S. Commission on Civil Rights. He is also a member of the National Labor Relations Board. These comments do not necessarily reflect the positions of either organization.