David E. Bernstein is on an intellectual crusade which, like many crusades, contains points of both insight and blindness. With commendable brevity he introduces the thesis of his new book, Classified: The Untold Story of Racial Classification in America:

Official American racial and ethnic classifications are arbitrary and inconsistent, both in how they are defined and how they are enforced. The categories are socially constructed and historically contingent. They evolved from older racist categories and have barely been updated since the 1970s…. Modern American racial and ethnic classifications do not reflect biology, genetics, or any other objective source.

His basic and incontrovertible point is that the standard five-fold classification of white, black, Hispanic, Asian, or American Indians (Native Americans) is utterly useless for deciding who should get preferences in education, business, or anywhere else given a large, heterogeneous population that must be sorted into distinct, but internally disjointed, groups.

Bernstein, a distinguished law professor at the George Mason University School of Law, opens with an anecdote that neatly illustrates the challenge: Kao Lee Yang, a Hmong-American neuroscience Ph.D. student, was denied a prestigious fellowship for members of “groups historically excluded from and underrepresented in science” because she was classified not as Hmong but Asian, and thus as a small member of a broad category full of gifted students from both East Asia and South Asia.

It is easy to conclude that this administrative ruling, even if lawful, is arbitrary and capricious. But it is far harder to figure out how to fix the problem. The government might break out the Hmong into a separate category—but which other subdivisions will be required from the diverse Asian population? And if some educational institutions entertain a Hmong subclassification, should they make special allowances for underprivileged individuals from better-represented Asian groups like Chinese and Indian Americans?

Similar adjustments will have to be made for all other groups. For example, the Hispanic or Latino label covers both Hungarian Jews who migrated to Argentina before coming to the U.S. and impoverished Mestizos descended from Spanish conquistadors. No consensus exists either within or across government agencies about what mix of ancestry, name, place of birth, or use of the Spanish or Portuguese language defines one as Hispanic or Latino. The ironies increase given that most Spanish-speaking people consider themselves white—at least before checking the right box that will provide some tangible benefit. And though the African American category may seem to contain less internal diversity, some members—like Michelle Obama—descend from slaves, and others—such as her husband, Barack—descend from more recent immigrants, either from different countries in Africa or the Caribbean. Should the law recognize a difference between them?

***

Bernstein notes with evident unhappiness that the National Minority Supplier Development Council—a non-profit organization that assists minority-owned businesses—uses “a combination of screenings, interviews and site visits” to determine who qualifies as a member of a minority group. In its view, members of such groups have at least one quarter Asian, Black, Hispanic, or Native American ancestry. These large, diffuse categories are blessed by most race-based interest groups because it allows at least four of the five racial classes to amass political power and ultimately wealth, opportunities, and systemic preferences, even if dividing up the spoils within any particular group becomes ever more perilous.

The increasingly common practice of intermarriage between individuals of different groups makes racial classification no easier. Bernstein writes: “As of 2017, 46 percent of Asian and 39 percent of Hispanic American newlyweds born in the United States married a spouse from a different category.” That fluidity and the absence of any obvious classification principle makes self-designation the first step in the selection process. But it cannot be the last step, given the obvious risk that some Robert Leo will become Roberto Leon, solely to gain a prized position. Who will police these exceptions? Firms, agencies, or courts? None of these alternatives is particularly attractive, least of all courts. Constitutional oversight is at best a faint possibility, as courts will hesitate to wade into this thicket. A durable consensus exists that some past discrimination justifies some racial preferences, but Americans have reached no clear agreement on how those preferences should be structured or why. Thus, the entire matter smolders under an all-forgiving rational basis test—the judicial equivalent of a rubber stamp. When courts pass, firms and agencies must pick up the slack by default.

***

Bernstein tracks these twists and turns with admirable perseverance. His search for viable reforms, however, leaves many stones unturned. At one point he argues for a better mousetrap: “I propose shifting from vague, ill-defined, and often overly broad and non-specific racial and ethnic categories to more precise, objective categories defined by history, politics, sociology, and, in the scientific context, genetics.” Unfortunately, this amounts to jumping from the frying pan into the fire, because Bernstein doesn’t offer any way to operate his new categories that would satisfy his craving for objectivity. Given the incredible lengths to which experts and administrators have gone attempting to unscramble this omelet, it seems unlikely that anyone could do the job. To be sure, genetic patterns among groups can matter. In medicine, for instance, certain groups are susceptible to certain diseases, such as sickle-cell anemia or Tay-Sachs disease. But tests for these conditions make up just a tiny portion of the practice of medicine, and the difficulty of how to divide populations—by sex, by pregnancy, by age, by race, by genetic marker—clouds clinical trials. As with every other regime, overly precise classification leaves the numbers in each category too small for statistical significance; overly broad classification causes the latent variations within each group to matter.

Nonetheless, Bernstein never gives up on his mission. He shifts to limiting the scope of the various government special programs to get, at least in part, out of the classification game forever. Yet completely cutting these programs troubles him. He thus suggests, “As in the MBE [minority business enterprises] context, affirmative action preferences, if pursued, should be limited to African American descendants of slaves and members of American Indian tribes who live on reservations.” But high intermarriage rates make eligibility tests necessary, and many would object if prosperous African Americans who previously benefited from affirmative action programs get a second bite at the apple while many poor whites or poor blacks not descended from slaves are forced to subsidize programs from which they receive no direct benefits—including reparations, a topic that Bernstein does not discuss.

These multiple and inescapable difficulties give ample reason for a fresh start, one that reframes the question to avoid using “the official”—that is, government—definition of racial and ethnic classes. Oddly, at no point does Bernstein address how private institutions might help tackle the challenges of racial and ethnic classifications. Their ability to do so is constrained by the presence of a strong set of antidiscrimination laws that undermine the public-private distinction so crucial to the principle that the 14th Amendment’s Equal Protection Clause only applies to state action.

***

To rethink the issue, start with the proposal that an antidiscrimination norm should govern private parties only when they are analogous to common carriers and public utilities, whose monopoly position puts them under the nondiscrimination rules. Ordinary educational and business organizations do not fall into that category. Let those organizations develop, each for itself and in its own way, the distinctions it finds relevant without the threat of government oversight under theories of either disparate treatment or impact.

This position rests on Friedrich Hayek’s claim that decentralized market decisions work better than centralized government ones. Decentralized decisions allow competition to help determine the applicable norms. This requires admitting (or insisting) that not only are there meaningful differences between various racial and ethnic groups, but there are also differences in the various institutions that teach, treat, or hire individuals.

Each institution will have to make, at the retail level, the same difficult trade-off that governments now make (badly) at the wholesale level. Two advantages favor decentralized dispute resolution. First, each institution will have more downstream information, allowing them to make the trade-offs they think appropriate without worrying about whether others agree. Nothing suggests these inconsistent outcomes will reflect some form of bad faith or incurable ignorance. Rather, local information should allow institutions to better make the relevant trade-offs when they no longer must work in lockstep with each other. Their classifications may be arbitrary, but if so, in different ways, and their variations will be legitimated by consent.

Second, once this strategy is adopted, those unstandardized paths increase the diversity in attacking these problems. Some groups might refuse to concede that there are any benefits to ethnic or racial diversity, but that does not mean the emergence of color- or race-blind standards should be dismissed out of hand. Indeed, some institutions may decide to revert to a whites- (or blacks-, etc.) only policy. But if it is possible for us all to allocate our dollars and allegiances elsewhere, why block any of these alternatives? Even if no government sanctions disfavor these choices, various consumers, employees, and students will make subtle and not-so-subtle social choices about where they want to go. Institutions will have the incentive to cater to the desires and attitudes of their targeted customers.

***

I have no doubt that, as an empirical matter, most of the choices will reflect the same set of racial and ethnic preferences that now drive government programs. It is highly unlikely that there will be a large-scale reversion to some Jim Crow system. Jim Crow, after all, was no market outcome. Its implementation required massive amounts of government coercion and private violence. Whether public institutions should operate on a color-blind or race-sensitive basis remains a problem, but learning from the successes and failures of private institutions might help ease the problem. In other cases, the distribution of various benefits in the public sphere (but never in the enforcement of criminal law) could track the patterns adopted on a widespread basis in the private sector.

I hope these mixed and somewhat untidy results will commend themselves to David Bernstein. He is a respected academic, courageous writer, and seasoned litigator. He is also the author of Rehabilitating Lochner: Defending Individual Rights Against Progressive Reform (2011), which defends the principle of freedom of contract on moral, economic, and constitutional grounds. Classified would have benefited from a larger dose of those insights.