As noted by Michael Fox in a recent post on his Employer’s Lawyer blog,, an OFCCP Administrative Law Judge (ALJ) just released a 66-page decision in a case that began with an audit notice in 1993. The case was bogged down in large part due to the bank’s contention that it was not selected for audit in accordance with its constitutional right under the Fourth Amendment to be free from unreasonable searches and seizures. That claim was ultimately unsuccessful. As a result of the delay, though, the bank found itself litigating claims about hiring practices dating back to 1993. Not surprisingly, the recollections of key witnesses such as the recruiters were foggy on some points.
But, in essence, the trial boiled down to a battle of the experts, who each advocated his or her own method of statistically analyzing the hiring data. The analysis of the OFCCP’s labor economist/statistician disregarded several of the bank’s legitimate business reasons for rejecting applicants because of evidence provided by the recruiters regarding how they coded applicants.
For each applicant, the recruiters were to use a code to indicate the outcome of the application. For example, they used a certain code to indicate that the applicant was not interested in working the hours that were available, and another code to indicate that the applicant had failed the credit check. Unfortunately, the recruiters testified that they did not use the code consistently.
If someone told the recruiter that her or she was not interested in the hours and/or the wages being offered, the recruiters sometimes used the code for “no position available” rather than the code used to indicate that the hours or wages were not acceptable to the applicant. To the OFCCP’s expert, this justified treating the hours code as entirely unreliable.
He also disregarded the code the recruiters used to indicate that the applicant was rejected based on his or her credit report for several reasons: (1) the recruiters did not have a consistent system for screening based on a credit report, (2) there was no evidence validating the use of credit reports as a test for success in the job, (3) the bank stopped using credit reports in 1994, and (4) the use of credit reports as a screening device adversely impacted African-Americans. The bank had not retained copies of the credit reports, so it was not possible to determine whether the recruiters used the credit reports in a consistent way as between white and African-American applicants.
When the employer’s expert analyzed the hiring decisions and excluded the people who had been rejected based on hours preferences or the credit check, the outcome was that there was no statistically significant evidence of discrimination. When the OFCCP’s expert analyzed the same hiring decisions but included the applicants who had been rejected based on the hours and credit check results, there was strong statistical evidence of discrimination.
The ALJ also rejected the bank’s expert’s opinion that the bank had hired more African-Americans for the jobs in question than would be predicted if the analysis had been based on the overall availability statistics for the Charlotte metropolitan statistical area for 1993. The ALJ wrote that “it is well established that the applicant flow data, which documents the actual labor pool relevant to the hiring decisions at issue, is ‘highly relevant evidence of an employer’s labor market.’”
This proposition is one that, in my experience, is theoretically appealing but completely out of sync with reality. The reality is that applicants’ self-identification of race and gender by applicants is voluntary, and a large number of them do not self-identify. Consequently, the employer, the courts and labor economists running statistical analyses will never have an accurate picture of the racial characteristics of the “applicant pool” from which the hires were made. Given that the information about the race and gender of the “applicant pool” is always incomplete and inaccurate, it is difficult to understand how applicant flow data can be more relevant and reliable than census data.
Anyway, this case still is not over. The ALJ has to decide what the damages number will be, and after that, if the case does not settle, appeals seem likely.