(FORTUNE Magazine) – We begin here by raising a question that every presidential candidate should be ordered, preferably at gunpoint, to answer squarely. The question is at the center of an emerging huge row in social policy and furthermore happens to be tres interessant. Question: Where do you stand, Mr. Candidate Man, on the issue of within-group percentile conversions? ''Huh?'' would possibly be the chap's opening rejoinder, so leave us back off and creep up to the question more deliberately. For many years, the U.S. Employment Service (USES) has offered the states a cluster of tests collectively known as the General Aptitude Test Battery (GATB). The tests measure verbal ability, quantitative skills, ability to visualize spatial relationships, and reasoning ability. Basic purpose of the GATB: to help the state employment agency predict how well the testee would be expected to perform in various jobs, a task the agency typically undertakes on behalf of employers trying to fill particular slots. GATB results are now being used to predict performance on every one of the 12,860 jobs in the Labor Department's Dictionary of Occupational Titles. The validity of GATB in predicting performance is now pretty well accepted by industrial psychologists. Over the years, an avalanche of ''validity generalization'' studies have powerfully made a case for them, and both the USES and the U.S. Office of Personnel Management now support the tests. They believe, furthermore, that there is an essentially linear relationship between test scores and job performance (i.e., the higher the score, the better the performance). This is now the position not only of the Reagan Administration's Labor Department but also of 42 state employment agencies that use GATB. But the view of GATB as a powerful predictor of job performance has an implication that a lot of people don't like. On average, minorities do not score as well on the tests as white folks. This is not in fact surprising, since GATB is quite similar to an IQ test, and minority IQs -- at least those of blacks, Hispanics, and Native Americans -- run lower on average than those of whites. (Asian-American IQs run higher.) Does this mean that the U.S. Employment Service is now offering tests that leave employers predisposed to hire nonminority workers? Friends, USES was not born yesterday. Somebody high up there obviously figured out that the bottom line had better show good scores for minorities. The USES solution, not mentioned in this item since the first paragraph: within-group percentile conversions. In other words, a white, black, Hispanic, or Native American job applicant gets a percentile standing that ranks him only against other members of his own group, with the scores of all the groups ''normed'' to show overall equality. When the state employment agency sends workers over to the employer, it doesn't simply send those with the best raw scores; instead it sends those having the highest within-group percentiles (and tries, as usual, to ensure that all groups are proportionally represented). This deal was not universally popular in the Reagan Administration. Back last February, Associate Attorney General Brad Reynolds was briefly threatening a Justice Department reverse-discrimination suit against within- group percentiles. Confrontation between Justice and Labor was avoided, however, with a cop-out compromise leaving the percentiles intact while they are massively studied by a ''blue-ribbon panel'' from the National Academy of Sciences. As you might imagine, within-group percentiles considerably reduce the utility of GATB tests. Professor John Hunter of Michigan State concluded after a lengthy study that the separate percentile arrangement has reduced the test's value to employers by 15%. So we think it would be just great to get a little feedback on the arrangement from all the presidential candidates. Especially when they get around to their set speeches on making the U.S. more competitive. You may need the gun, at that.