(FORTUNE Magazine) – Among the more dispiriting events of recent weeks was the publication of Fairness in Employment Testing, a highly depraved document produced by the highly prestigious National Research Council. (The NRC is an arm of the National Academy of Sciences.) Especially gloom-engendering was the response to Fairness of the umpteen thousand business folks who speechify about productivity and competitiveness. Their landslide response was to not notice the document. To be sure, it did not get a lot of ink from the media. Still, Fairness is horrifying news for productivity, so you might have expected a scream or two from the speechmongers. To truly savor the depravity of the NRC's report, you must first focus on certain practices of the U.S. Employment Service. The USES, run by the Department of Labor, provides job tests to state employment agencies -- tests that these agencies can administer to job applicants whenever a company calls up and asks for, say, six machinists and five secretaries. By far the most important of the tests is the so-called General Aptitude Test Battery (GATB). The GATB is somewhat like an IQ test: It measures a number of particular mental abilities and also provides a gauge of ''general cognitive ability'' -- what the IQ testers call g. Fortified by an avalanche of studies, the Labor Department has come to believe in recent years that g is the best available predictor of job performance. The department's data suggest that g scores predict most accurately for performance in high-level jobs but that they are unsurpassed as a predictor even for most unskilled jobs (the exceptions being those in which you do better with tests of motor ability). Taking the labor force as a whole, the correlation between GATB scores and actual job performance was stated to be 0.5. (A perfect positive correlation would be 1.0.) This figure was substantially higher than that for any tests of specific job-related knowledge or for hiring based on the credentials, experience, and education of the applicants. So, obviously, widespread use of the tests would do a lot for productivity. But there was this little problem about the GATB. Racial minorities, especially blacks and Hispanics, score lower on average. This means that a state agency sending over the six highest-scoring machinists would be repeatedly sending over an all-white cast of characters and maybe thereby violating the Civil Rights Act (although not everyone agrees this would in fact be a violation). To avoid such outcomes, Ronald Reagan's Labor Department began recommending in 1981 that the states opt for ''within-group norming.'' In other words, black (or Hispanic) applicants would get a percentile score reflecting only their competition with other blacks (or Hispanics). Not surprisingly, this solution created other problems. It meant that minority applicants would often be sent over for the job even if their true scores were lower than those of rejected white applicants. Furthermore, Ronald Reagan's Justice Department thought that within-group norming violated the Civil Rights Act. When assistant attorney general Brad Reynolds began muttering about a suit against the scoring system, the National Research Council was brought onstage to evaluate the system. Bringing in the NRC always looked like a bit of a non sequitur, as the question on the table -- do we want to give job preference to minorities even ; if that means less efficiency? -- was ultimately not scientific; the question was obviously political. The NRC's logical irrelevance led to suspicions that its real role was to lend the shiny-bright prestige of science to the dirty business of institutionalizing racial preference. James C. Sharf of the U.S. Office of Personnel Management has quoted one leader of the NRC group as denying that its mission was ''to put a scientific fig leaf on a naked political argument.'' Sharf himself has been quoted as disgustedly concluding that this is precisely what the group has now done. Among its basic findings: -- The GATB is an extremely useful test and is a good predictor of job performance. The committee quarreled with the correlation coefficient of 0.5, however, and said it would not itself spring for more than 0.3. -- The test is not biased against minorities: A given score, i.e., before norming adjustments, predicts the performance of blacks and Hispanics as well as that of whites. -- Nevertheless, the test is not perfect, and since minority-group members would be disproportionately affected by top-down score reporting, it would be unfair to leave their scores unadjusted. But wait. How can they say unadjusted scores would be unfair when they just got through acknowledging that the GATB is not biased against minorities? Friends, we have looked carefully through the report's 354 pages and cannot find a crisp answer to that question. The report dances all around the question. It alludes affirmatively to government policies calling for ''inclusive discrimination.'' It reminds you, in case you forgot, that efficiency isn't everything. It worries that USES staffers might draw the wrong inferences about the potential social contributions of minority-group members. It never does explain how racial norming of test scores will lead them or anybody else to the right inferences.