AN EMPLOYER buys into the American Dream and starts a new business. He hires ten workers, but only one of them is black. Should these statistics alone constitute proof that this employer is a racist and in violation of employment-discrimination law? If the Kennedy-Hawkins Civil Rights Act of 1990 had become law, the answer would be yes.
The bill was intended to overturn Supreme Court decisions that rejected the claim that underrepresentation of minorities in a work force by itself creates a presumption that the employer has discriminated against them in his hiring practices, even when no discriminatory practices are found.
Quotas would become irresistible and inevitable under such a bill, even though it expressly said they would not be required. Employers exercising basic common sense would hire along statistical lines rather than risk incurring the significant time and expense necessary to litigate under rules that all but guarantee losing at trial and thereafter being branded in the community as discriminators.
But can mere statistics, by themselves, testify to discrimination? Consider a group of 1,000 employers who each employ 10 workers. Assume that the local populations from which they draw their work force are 30 percent black and 70 percent white. Assume further that these employers are fine, outstanding citizens who are all absolutely immune from the taint of racism and employment discrimination, and who all feel a moral, not merely legal, obligation not to make race-conscious decisions. Let us even hypothesize that they interview and hire potential employees behind a screen, sight unseen, or that they retain personnel by hiring, at random, every tenth person who walks into the room seeking a job.
It might seem that each of these 1,000 employers, subject to the above constraints, would end up with three black and seven white workers. In fact, only 267 of the 1,000 employers would show this ''expected,'' ''optimal'' result. Most interestingly, 121 of the 1,000 employers would hire but one black employee, and 28 of these non-discriminating employers would not hire a single black worker! Others would have majority-black work forces.
What this statistical representation clearly shows is that many employees will be able to make a prima facia showing of ''disparate impact'' against an employer because the minority proportion of the employee population fails to mirror that of the general population. But that result obtains merely because of the random statistical behavior of populations.
In other words, statistical principles dictate that deviation from the mean is not only usual, but actually expected. Even in a discrimination-free world, it would be a statistical oddity of the highest order, bordering on the miraculous, for each employer to match the racial and gender distribution reflective of the qualified work force. Human beings simply do not produce convenient, statistically balanced results.
Employers actually attaining the optimal racial mix of the local population in their work forces, then, are the ones who ought to be carefully scrutinized, because it is most likely that these employers are the ones who are making race-based decisions.
An old statistical fable tells of a mathematician who, over a period of weeks, weighed the daily loaf of bread sold to him by the local proprietor and discovered that each loaf weighed less than the advertised average weight.
''I accept the fact that, given the expected distribution of weights, some loaves will be underweight,'' said the statistician to the merchant, ''but the uniformity of the underweight result cannot be attributable to mere chance and, unless you undertake an honest effort to provide the proper weight, I shall register a complaint with the authorities.''
In response, the sly merchant ordered his employees to specially prepare a very slightly overweight loaf to be given the mathematician each day, and otherwise continued his short-changing criminal practice.
The wily old statistician nevertheless quickly deduced the scheme because he knew, as the unsophisticated baker did not, that if the loaves were truly made and randomly selected, it was equally implausible to consistently obtain loaves at the proper weight.
By analogy, any employer consistently reflecting the proper race distribution in his work force must be making race-conscious decisions because the odds are against attaining this result by mere chance.
An employer should be hauled into court if an employee can allege a specific practice or practices that lead to a ''disparate-impact'' result and that a severe burden should be imposed on such an employer to defend that allegedly discriminatory practice. But the mere recitation of statistics is simply not enough to establish causation or even, as we have shown, to establish a prima facia claim that the employer can reasonably defend against. Any law that must, by its very nature, sweep away the innocent with the guilty is simply a bad law. It is good that Kennedy-Hawkins was vetoed.
The writer, formerly an actuary, is a Baltimore attorney.