I experienced this firsthand in 1982 when I walked into the offices of a 2-year-old marketing research firm in Chicago that was the first research firm to use data from supermarket scanners. I had just earned an MBA from UCLA with double concentrations in marketing and finance, and had letters of recommendation from the most quant-intensive professors that I was the second best quantitative marketing analyst in that year's graduating class. (The best one, no fool, opened a scuba diving shop in the Florida Keys.) I was looking for a job in marketing model, but they didn't have any openings, so they asked if I was interested in the more general Client Service area.
The Human Resources department gave me their Client Service hiring test, which was the final exam given by a co-founder of the firm, U. of Iowa professor Gerry Eskin, to his class in Marketing Research 402: Senior Seminar on Quantitative Methods (or something like that). I sweated over it for three hours, computing correlation coefficients on my HP hand calculator. The next day, having graded it, they called me back for a day's worth of interviews. I worked there, on and off, through 2000.
The firm continued to use Dr. Eskin's final exams until a number of years later, when it had grown large enough that it was now on the Equal Employment Opportunity Commission's radar. The Griggs decision said that tests that had racially disparate impact, as no doubt this one did, being cognitively demanding, caused the burden of proof in discrimination cases to be placed on the employer, unless the firm demonstrated that it met the stringent level of "business necessity." Some of our clients, such as Procter & Gamble, the paladin of the consumer packaged goods industry, went through the considerable expense of validating their hiring tests to the standards demanded by the federal government, but our HR department decided to give up on using something as idiosyncratic as a Big Ten test.
My personal view is that the quality of our hiring never recovered from this change.
The interesting contrast, however, is that while the firm gave up using this U. of Iowa test, the federal government never had any complaint about the U. of Iowa going right on using Dr. Eskin's test to determine grades, credits, and diplomas, which obviously has an indirect impact in the employment market. Disparate Impact just doesn't much apply to universities.
As Justice O'Connor wrote in her majority decision upholding the U. of Michigan's quota system (just don't call it a quota) in 2003's Grutter decision:
"Our holding today is in keeping with our tradition of giving a degree of deference to a universityâ€™s academic decisions, within constitutionally prescribed limits.... "One might think Supreme Court Justices would logically give less deference to universities — what with all the Justices having at least seven years experience with higher academia — and more deference to, say, marketing research firms, few of the Justices having any experience in that business. Instead, however, the opposite is true.
The long term effect of Griggs's Disparate Impact theory is explained well in The Bell Curve: It pushes the evaluation process required for hiring back from the employer to the university. It exacerbates the tendency toward credentialism in American life.