Civil Rights: Testscam


When the Waterloo, Iowa, factory for which he had worked 16 years closed, shipping and receiving supervisor Larry P. Holman gladly accepted his sister's recommendation that he move to Richmond, Virginia, where she figured there were plenty of industrial jobs. Holman applied for a job with the James River Corp., which referred him to the Virginia Employment Commission for a job aptitude test in December 1989. But unbeknownst to Holman, the test was rigged against him.

Holman took the General Aptitude Test Battery. A handout accompanying the test explained that he would receive percentile scores of 1 to 99 in each of five job categories, and that a score of 50 would mean that "you scored higher than 49 percent of the people nationwide who have taken the test."

That was a lie.

In 1981, the U.S. Employment Service, a branch of the Department of Labor, had given state-run employment services a new system for converting actual scores on GATB to percentile rankings. Bureaucrats and approving social scientists called it "within-group scoring." In truth, it was a system of scoring based upon race. No longer would job seekers be graded purely on their individual merit against all other candidates. Instead they would be measured strictly against members of their own race.

Here's how it worked. Suppose that a company had an opening for a bookkeeper and asked the state employment service to refer the top-scoring GATB tested applicants. Suppose further that four candidates—a black, a Latino, an Asian, and a white—each had a composite raw score of 300. The black applicant would receive a converted score of 83, the Latino candidate would get a 67, and the white and Asian applicants, lumped together in an "all others" category, would receive a dismal score of 45.

Neither job seekers nor job providers would ever see the real scores from which these percentiles were concocted. The purpose of secretly manipulating test results, say USES officials, is to eliminate the "adverse impact" on certain minorities resulting from the fact that they tend to score lower on standardized tests than members of other racial groups.

The race-norming supposedly results in equal referrals by compensating for disparities in group averages. But a 1989 National Academy of Sciences report acknowledged that the formula for converting scores was based on data that were as much as 17 years old and "by no means nationally representative samples." So the test results may have been slanted toward minorities even more than the USES intended, since minority averages on tests such as the SAT have risen during the past two decades.

The Supreme Court's landmark Griggs decision of 1971 requires employers to show that intelligence tests and other requirements that disproportionately screen out minority job candidates are related to job performance. In response to Griggs, USES set out in the late 1970s to help businesses that found it hugely expensive to devise and validate hundreds of tests for different jobs.

Psychologists John Hunter of Michigan State University and Frank Schmidt of the University of Iowa took the GATB, whose job-relatedness was well established, and found a way to generalize its validity for 500 positions to all 12,000 occupations in the U.S. economy. They called the new system Validity Generalization.

But USES took the court dictum a giant step further and determined that even tests that accurately predict job performance must be finagled toward equal outcomes to avoid adverse impact on minorities. USES piggybacked within group scoring onto V.G. in an attempt to eliminate adverse impact, even though the Supreme Court had said adverse impact was illegal only if the test was not job-related. Thus USES intentionally subverted a test designed to eliminate discrimination to defeat that very objective.

Had USES at least leveled with individuals about the nature of this social experiment, it might have been entitled to credit for good intentions. But when Larry Holman took the test in 1989, he knew nothing about the race-based scoring. Indeed, USES told none of the millions of people who took the test between 1981 and 1990 that the scores were racially biased in favor of blacks and Latinos. The truth did not come out until the summer of 1990, when a government whistle-blower gave the Richmond, Virginia, Times-Dispatch a copy of the percentile conversion tables.

The effect of this social experimentation on individuals like 47-year-old Larry Holman was sometimes devastating. Unconverted, Holman's scores might have been good enough to win him a job interview. But converted, the scores that arrived in the mail from the state job service were a disaster. Soon Holman received a letter from the James River Corp. stating that because of his low GATB scores, it was placing Holman's file on "inactive" status.

"It was one of the hardest things I can remember in my life," he says. Holman was unemployed, and now the government was telling him he was stupid; and a major employer was telling him he wasn't wanted. He says he doesn't think he could have pulled through that bleak period had it not been for the support of his sister and brother-in-law. Fortunately, Holman's story had a happy ending when the Waterloo plant reopened and his old boss called him back to work—with a raise in pay. And no one even asked for his GATB score.

Ironically, since Holman says he is "one part Indian—one very small part," he could have registered to take the GATB as a Native American and received hefty bonus points, since USES has a separate table for Indians.

On July 10, the public outcry that followed disclosure of the conversion tables forced then Secretary of Labor Elizabeth Dole to suspend GATB use for a two-year period of study, after which she said she hopes to "reestablish GATB as a state-of-the-art measurement tool by making it more comprehensive and job-related and by minimizing disparities in scores among various groups." Dole seemed to anticipate the invention of some new subterfuge to equalize test results.

Dole's decision to suspend use of the test altogether reeks of political expedience. During its 42 years of use, testing experts have repeatedly found the GATB a valid gauge of job aptitude. The problem, Dole acknowledged, was in the "experimental feature" of within-group scoring. So out went the baby with the bath water.

The adverse impact of race-norming may not stop at blatant reverse discrimination. The ostensibly favored minorities may also be victims, because the system assumes that the causes of group disparities in scores are not poor education and limited opportunities, but race alone.

In the words of Clint Bolick, director of the Landmark Legal Foundation's Center for Civil Rights, race-norming ensures that "we will never address the reasons blacks score at a low level. It sweeps many problems under the rug. It is really a big fraud."

Unfortunately, political forces may allow that fraud to continue. Certainly Dole's remarks make it clear that USES wants to maintain race-based scoring. The two-year moratorium allows the public furor to die down and USES to find a way to make the GATB's racial bias less obvious. After that happens, race-based scoring will probably be reintroduced. And personnel offices may even welcome a race-normed GATB as protection against civil rights suits. When race-normed testing is reinstituted, bureaucrats will sit back and proclaim victory against racial discrimination.

Robert G. Holland is associate editor of the Richmond (Va.) Times-Dispatch.