The testing trap.

American companies find themselves in a dilemma: In an increasingly competitive environment, with people as their most significant business asset, they can afford no less than to hire the best. Rigorous, scientific studies demonstrate that testing enables them to do that. But testing generally produces results disproportionate to the broader racial makeup of a community or marketplace. This puts the merit-based approach on a collision course with both Constitutional law and legislative and regulatory pronouncements. It also puts companies at risk of costly litigation.

Are tests unfair to minorities? Should companies continue to rely on them? The answers to both of these questions - and the decisions based on them - are complicated, subjective, and carry significant, potential assets and liabilities. The immediate stakes are no less than the ability of U.S. companies, schools, and other institutions to judge people according to their merits - rather than the color of their skin. In the longer term, of course, the stakes are the competitiveness of American business.

IBM, Xerox, and American Express are among the major companies that set up aggressive hiring programs to increase minority representation. The list seems likely to grow longer as brow-beaten companies fall into a politically correct lockstep.

A BALANCED WORK FORCE?

Some examples: Liberty National Bank & Trust faced a discrimination suit in the early 1990s when federal regulators found that, although 16 percent of the Louisville, KY-based bank's employees were black, 32 percent of those applying for jobs were black. Ultimately, Liberty agreed under duress to offer jobs to 18 African-Americans it previously had rejected, with back pay. The case cost Liberty millions of dollars in legal fees and compensation. Similarly, statistical imbalances in its work force impelled Eagan, MN-based Northwest Airlines in 1991 to spend $40 million to accelerate the hiring and promotion of minorities, establish minority scholarships, and award back pay to minority employees.

As these examples illustrate, all hiring standards that show a "disparate impact" on racial groups are suspect. Since 1971, the Supreme Court has insisted that tests on which minority groups do not perform as well as whites generate a presumption of discrimination. Companies have the burden of producing studies to convince the government that their merit standards are "demonstrably related" to job performance and required by "business necessity."

In practice, many companies avoided government intrusion and legal liability and bad publicity - by abridging their hiring standards to ensure adequate minority representation. Companies such as IBM, Xerox, American Express, Walt Disney, KFC, and Price Waterhouse have established aggressive preferential hiring systems for minorities. Some leading CEOs are finn proponents of racial preferences, either to right past wrongs or avoid litigation. "We've got to get right with the future," asserts David Lawrence, publisher of The Miami Herald. Warns James Preston, CEO of Avon, "Companies that do not adjust to diversity are asking for trouble."

The main reason racial preferences are promoted by the government, enforced by the courts, and tolerated by private and public institutions is that hiring standards - in particular standardized tests are widely considered to be racially (and sometimes sexually) biased. One example from the public sector demonstrates that sometimes even the best intentions don't satisfy critics: In 1994, the Chicago police department was faulted by minority leaders after it spent $5 million to produce a bias-free test that failed to even scores among applicants of all races (out of 500 candidates who scored highest on the test, only 40 were black, and 20 were Hispanic). Critics assumed that in a city that is more than 50 percent minority, test results that did not mirror the population must be racially biased.

ASSESSING THE TESTS

The two most widely used forms of standardized testing are the General Aptitude Test Battery, administered by the U.S. Employment Service and used for millions of job referrals to state and private employment agencies, and the Scholastic Assessment Test, conducted by the Educational Testing Service and considered by most American colleges for undergraduate admission. In addition, standardized tests are used for admission to graduate school, as well as for numerous private- and public-sector jobs.

In the past two decades, thousands of studies have sought to assess how well standardized tests such as the SAT and the GATB predict performance, and whether such tests are biased against minorities. These studies have produced results so unanimous they are considered definitive by scholars, although they are not well known to the public at large.

The SAT, it turns out, is an excellent predictor of college performance, much better than high school grades, which depend on where the student went to school, and interviews and recommendations, which are subjective. Study after study has shown that the SAT is not biased against blacks - indeed it slightly overpredicts the performance of African-Americans in college.

Similarly, a review by industrial psychologists John Hunter and Frank Schmidt of many studies of the GATB showed that the test effectively predicts job performance, as measured by independent criteria such as supervisor ratings or on-the-job evaluations. A review of almost 100 studies by the National Research Council found that the GATB is equally predictive for whites and minorities.

As psychologist Lyle Jones argues, tests do not create discrepancies of ability and performance; they merely measure them. All racial groups in our society do not perform equally on those measures of merit that selective colleges and companies use to choose the most qualified applicants. Indeed, virtually every test shows large average differences of performance among groups, with whites and Asians scoring at the top, Hispanics in the middle, and African-Americans at the bottom.

This is the heart of our social problem. Although we want to live in a pluralistic society, we are moving toward a racial hierarchy: Merit produces inequality, not just among individuals but among groups. Our civil rights laws don't help, as they ferret out discrimination by looking at group outcomes. If all groups do not hit the finishing tape at the same time, the Equal Employment Opportunity Commission and other civil-rights agencies insist the race has been rigged.

This reasoning is both a non-sequitur and a chimera, yet for almost a generation, U.S. companies have been living under laws based on it. Tama Starr, president of a small New York service company, wrote in a trade magazine that government regulators descended upon him for firing an employee who had been "late or absent 70 out of 249 consecutive workdays." Starr reports that to avoid a civil-rights lawsuit, his company "had to provide reams of data comparing lateness and absenteeism by observed skin color for all apprentices within his particular trade over a 52-week period."

What can companies do? One option is to forgo testing altogether. But that can lead to productivity losses as companies substitute less-effective ways to measure performance. Those measures can be challenged if they do not produce racially proportional results.

A second option is for companies to use standardized tests for whites, while hiring minorities by other criteria calculated to assure sufficient representation. This strategy can help minimize the risk of government harassment or a lawsuit filed by a minority plaintiff, yet if racial preferences are egregious, it can invite suits from whites. Companies must weigh the risk of being sued by minorities on the one hand, and whites on the other. Under current law, minority claims are taken more seriously, but as white grievance escalates, the pool of potential white plaintiffs can become formidable.

A third choice - offered for the bold - is for companies to be strictly meritocratic, understanding that virtually all standardized tests are likely to produce racially disparate outcomes. If the company is willing to have a work force that does not "look like America," and is willing to defend its merit standards before the courts, it is likely to reap the highest productivity gains.

It is time to face up to the national challenge, which is twofold. First, we must work for an improvement of skills and standards among all groups, but especially blacks who fall so far behind others. Second, we need to return to Martin Luther King's vision of an American standard that is truly meritocratic and color-blind. Thus, we can simultaneously serve the goals of economic productivity and social justice.

Dinesh D'Souza is the John Olin Scholar at the American Enterprise Institute. His new book, "The End of Racism," has just been published by the Free Press.