The third in a series on the CFPB's use of disparate impact. Part one is here, and part two is here.

WASHINGTON — Dr. Bernard Siskin may know more than any other outsider about how the Consumer Financial Protection Bureau cites big banks and other lenders for unintentional discrimination -- and he has turned that unique knowledge to his advantage.

Public records show the agency has paid Siskin's statistical analysis firm, Philadelphia-based BLDS LLC, more than $1 million since 2012 to consult on the CFPB's disparate impact methodology and appear as an outside expert witness on enforcement matters dealing with fair-lending cases. Yet at the same time, Siskin has also been hired by the biggest banks in the country to defend them against unintentional discrimination charges, including those brought by the CFPB in citing auto lenders for bias.

"I'm a consultant to the CFPB. I do a lot of their consulting work. I have a blanket contract for the enforcement group, as well as I represent, on the other side, most of the major banks in America. I represent Chase, Citi, Bank of America and others," Siskin testified on a separate case related to voter IDs in Pennsylvania in 2013.

Serving both sides at the same time is allowed, according to sources familiar with government contracting. But it has created problems for the agency.

In at least one major case against a large lender, CFPB officials delayed other investigations after finding out that Siskin was arguing against the methodology that his firm helped the agency use to cite lenders, according to internal documents reviewed by American Banker. The lender, meanwhile, made sure to point out that its expert witness, Siskin, was "well known to the agency," according to internal documents.

Siskin declined repeated requests for comment for this story. Sam Gilford, a CFPB spokesman, said Siskin's work in helping the CFPB and big banks is not unusual and does not indicate a potential conflict of interest.

"As a general matter: Working for a particular company within an industry does not disqualify an expert from providing his or her services to a federal agency for purposes of an action involving a different company," Gilford said. "Experts typically do not recommend a single approach, but instead offer advice on a variety of options that they view as reasonable."

But former regulators said that when agencies do hire an outside expert witness, they usually require the witness to refrain from working with banks while under contract.

"It's not unusual for the Justice Department to hire outside experts on cases, but it is unusual for that expert to play both sides of the fence," said Michael J. Bresnick, a former top DOJ official who now chairs the financial services investigations and enforcement practice at the law firm Venable. "And working for a government agency would not preclude that expert from representing the company in an unrelated matter, but it would get troublesome when that expert starts punching holes in the agency's method."

Some former CFPB officials wondered why the agency was hiring an outside expert well before any case was being brought to trial.

"While it's not unheard of to have a government agency pay an expert witness, many regulators have internal economics or research departments that provide the analysis at no additional cost to the government. Considering that the CFPB will always be supervising and enforcing fair lending laws and has a research office of economists—it is surprising that the CFPB would not develop this expertise 'in-house,' " said Gerald S. Sachs, former senior counsel at the CFPB's enforcement office who is now with Paul Hastings. "It brings into question whether a second look from an independent inspector general or a commission would have made this process more efficient."

American Banker reached out to all the prudential financial regulators to ask about their practices in hiring external expert witnesses for enforcement actions. Regulators replied that they usually use internal economists or their own research departments to conduct analysis in fair-lending cases and then refer any actionable cases to the Justice Department. As a result, it's rare that the agencies would need to hire an outside contractor. The CFPB also has its own internal office of research that helps conduct the analysis in fair-lending cases. The Justice Department declined to comment.

Gilford, the CFPB spokesman, said the agency's hiring of outside consultants was appropriate.

"The bureau retains the services of experts to use in litigation or disputes, including any reasonably foreseeable litigation or dispute; expert services include but are not limited to assisting the government in the analysis, presentation, and defense of its claims," he said. "In procuring the services of an expert, the CFPB complies with all relevant laws and regulations."

It is not atypical that regulators and banks would seek advice from the same statisticians, given that there is a limited marketplace for those experts.

"I always tell clients that they are responsible to do their own analysis -- and prior to the regulator coming in," said Ed Kramer, executive vice president of regulatory affairs at Wolters Kluwer, and New York's former deputy superintendent of banks. "My advice to every auto lender is to hire the good statistician before the regulators do."

Yet Siskin has a "blanket" contract with the CFPB on fair-lending matters, making him a particularly valuable commodity to banks facing potential actions by the CFPB.

An Inside View

The agency has paid a significant amount of money to secure Siskin's services. Exactly how much is unclear and depends on which website is used to sift through publicly available government data.

The USAspending.gov website shows 65 transactions between the CFPB and Siskin's firm, BLDS, totaling more than $940,000 for work since 2012. But insidegov.com, an independent website that collects government data, shows a $1.6 million contract the CFPB awarded to BLDS for work from June 2012 that ended Sept. 25 of this year. The CFPB would not disclose the exact amount it paid Siskin.

"The CFPB follows all relevant laws and regulations in procuring services, and information about contracts is reported on USASpending.gov as appropriate," the CFPB's Gilford said.

But Siskin's work on both sides of the fence was clearly problematic for the CFPB at times.

Under the legal theory of disparate impact, a company can be cited for unintentional discrimination if data shows disparities between what minorities were charged and what white borrowers were charged. But using different models can lead to different conclusions about how many borrowers were actually affected and about the harm done to them.

Additionally, while disparate impact is contentious in most areas where it is used, it is particularly so in auto lending because lenders are forbidden from asking a borrower for information about their ethnicity or gender, information that mortgage lenders are required to collect. The CFPB relies on proxies that examine a borrower's surname and neighborhood, among other factors, to identify minorities.

Which proxies the CFPB uses — and what other factors the agency puts in place — can make a difference in the disparities it finds. For the most part, the industry claims it doesn't know how to apply the CFPB's methodology to its compliance systems. Except for a white paper and software code that the CFPB released last year, which banks argue are not detailed enough, the agency has not publicized further specifics about its methods.

As a result, Siskin is particularly valuable to banks because he has inside knowledge of the agency's methodology. In public speeches, Siskin has talked about different ways the agency's proxy method can be applied to identify potential discrimination.

When hired by lenders, he has gone further, helping the banks to suggest how to use the CFPB's methodology while adding other factors that would show less discrimination.

In April 2013, for example, a large lender under investigation by the CFPB in an indirect auto lending case challenged the CFPB's methodology by hiring Siskin. As a result, the CFPB delayed investigations against other lenders until it could assess Siskin's and the bank's argument.

"Pending evaluation of this methodological contention, we have delayed sending PARR letters [Proposed Action, Response Request] in other indirect auto lending exams," wrote Patrice Ficklin, assistant director of the CFPB's Office of Fair Lending, in a drafted memo to agency officials at that time.

The delay did not last long, but the CFPB ultimately concluded Siskin had a compelling case — and that his disparate impact methodology, which showed less potential discrimination, was valid. Officials suggested they may make changes to the agency's methods at a later stage, but continued to use the existing calculations for now.

"First, the alternative method proposed by [the bank] is not invalid or unreasonable, and thus could potentially suggest a lower bound on disparities that we should bear in mind as we make decisions on how to proceed in the current auto lending matters," Ficklin wrote. "Second, [the Office of Research] will continue to evaluate ways of enhancing its method, and additional PARR responses and discussions with other regulators and academics may help identify adjustments or alternative methods for consideration. For now, though, we would like to proceed in reliance on the existing [Office of Research] method."

It's unclear whether or not the CFPB has since made adjustments to its method. The CFPB's Gilford would not discuss any ongoing investigation but said the agency could adjust the method at any time.

"We are committed to continuing our dialogue with other federal agencies, lenders, advocates, and researchers regarding the methodologies we use in our work addressing discrimination. Regarding the BISG methodology, we wrote in the September 2014 white paper that the 'methodology has evolved over time and will continue to evolve as enhancements are identified that improve accuracy and performance,'" he said. The "CFPB's statistical analyses of disparities are tailored to individual lenders' policies, and we evaluate potential findings from the perspective of alternative methodologies known to staff at the Bureau or presented to us by institutions under review."

CFPB's Uncomfortable Position

The documents make it appear that officials were discomfited by Siskin's role at the lender. In internal discussions, CFPB staff sought to remove references to the bank's "expert witness" (Siskin) and instead just refer to the bank.

Officials "would rather attribute the argument to [the bank] and not Siskin, given we have engaged him for other institutions," CFPB officials wrote in one drafted internal response.

They also acknowledged Siskin was raising legitimate issues. Several documents show that CFPB officials considered Siskin's alternative methodology, which included some controls, "reasonable" and similar to the agency's own method — which CFPB officials were not expecting.

"Regarding the estimation issue raised by [the bank], our initial expectation was that OR's [Office of Research] analysis of the two estimation methods would reveal that one or the other was plainly superior," Ficklin wrote in April 2013. "However, OR has concluded that the two methods are both reasonable, but under different assumptions about the underlying cause of the disparities."

Ficklin also acknowledged in the memo that the CFPB's method can overestimate the disparities but it was not enough to change its mind about citing the lender.

The CFPB's Office of Research proxy "method is valid and reasonable under the circumstances here, and although there may be some risk of overestimating disparities, the alternative presents an equal (and perhaps greater) risk of underestimating disparities and thus consumer harm," she said.

Still, some observers said it's perplexing that the CFPB continues to aggressively pursue indirect auto lenders for disparate impact when it is internally acknowledging concerns raised by its own expert. That could pose problems if a case were to reach trial, something that has yet to happen.

"If the CFPB's own expert witness says it is undertaking its fair-lending analysis wrong, that could pose real issues for the CFPB should someone decide to challenge the bureau in court," Sachs said.

Subscribe Now

Access to authoritative analysis and perspective and our data-driven report series.

14-Day Free Trial

No credit card required. Complete access to articles, breaking news and industry data.