If tests were perfect, and if test scores were always fair and accurate measures of a person's ability to perform, it would be reasonable to ban adjustments to test scores.
Unfortunately, tests are often culturally biased and the scores people obtain on them are often of limited use in predicting their ability to perform. Where the test scores themselves are unfair, banning adjustments to test scores will interfere with merit selection, not protect it.Suppose that you are an employer hiring several inexperienced typists. You use a typing test, and find that every increase in test score corresponds to greater ability to type. If these are all the facts, it would clearly be unfair for you to add points to the test scores of blacks just because they obtained lower scores on the typing test.
Suppose, however, that there were other facts clouding this simple picture: some of the lower-scoring black applicants impressed you in their interviews. You found that their inner-city schools still used manual typewriters in typing classes, and they were not used to the electric typewriters used for the tests. The white applicants had all used electric typewriters in their typing classes.
Now what do you do? You know that the scores obtained by black applicants cannot directly be compared with the scores obtained by white applicants. You check with a friend at another firm, and she confirms that black hires routinely outperform white hires who had scored up to 10 points higher on the same typing test.
If you are still interested in hiring based on merit, you have to regard a typing score of 75 for a white applicant as being the equivalent of a typing score of 65 for a black applicant in order to make hiring decisions truly based on ability to perform the job.
This kind of situation occurs frequently. In 1989, a federal court barred New York State from awarding Regents' scholarships based solely on the scores obtained on the Scholastic Aptitude Test, a test developed to predict first-year college grades.
Women receive lower scores than men on the SAT, so men received most of the scholarships. However, women's first-year grades are higher then men's even when controlling for course of study. The SAT substantially under-predicted women's ability to perform well. Without a score adjusted, its use would be unfair.
The Labor Department's Specific Aptitude Test Batteries are used by State Employment Service in deciding which applicants to refer to employers with job vacancies.
Whites do a lot better on the tests than they do on the job, and blacks do a lot better on the job than they do on the test.
The "intelligence" tests administered by the Army during World War I are a chilling example of what can occur when differences in test scores are treated as if they have intrinsic meaning - 38 percent of Irish draftees and 68 percent of black draftees were regarded on the basis of these tests as having a mental age of 8 or younger.
The lower scores received by blacks and by recent immigrants from Ireland and southern and eastern Europe were used to support theories of racial superiority and were relied upon in efforts to limit the immigration of southern and eastern Europeans.
Clearly, there was something seriously wrong with the tests, and not necessarily anything wrong with the test-takers. With this example in mind, we should not again make the mistake of treating test scores as automatically meaningful in themselves.
We know that minorities score lower on most tests than whites, that women score lower than men, and that southerners of any race score lower than northerners of the same race.
Before we engrave test scores in stone, we must first ensure that the test scores themselves are fair measure: that they really are related to ability to perform, that the differences in performance they predict are meaningful, and that their predictions are equally accurate for all race, gender, and regional groups affected by the tests.
(Richard Seymour is a civil rights lawyer in Washington, D.C.)