by Jeff Sovern
The article, by Kate Berry and Ian McKendry, is headlined Fight to kill CFPB arbitration rule could rest on whose data is right. Here's some of what the article says about the OCC claims, though the article has more than I can insert here.
"The uncertainty of the OCC's estimate is very large, it's a very noisy estimate, and there is quite a high chance that
the true effect is zero," said John Campbell, the Morton L. and Carole S. Olshan Professor of Economics at Harvard
University, who is a member of the CFPB's Consumer Advisory board. "The CFPB argues in its defense that the
effect can't be a large number like 3% because it's ridiculously large, relative to any reasonable estimates or costs
based on past class-action suits."
* * *
Other economists challenged the OCC's data, casting doubt on anyone's ability to predict consumers costs
resulting from the rule.
"Any statistically significant results would be highly questionable given the level of noise in the data," Alexei
Alexandrov, a senior economist at Amazon and a former senior economist at the CFPB, who had written a paper
that was the basis for the CFPB's own arbitration study.
* * *
However, Ted Frank, a senior attorney and director at the Competitive Enterprise Institute's Center for Class Action
Fairness, said that even if the CFPB's study found that an increase in credit card rates was not statistically
significant, that "does not mean it's not significant."
"They often cherry-pick their data," said Frank. "To say it's not statistically significant, therefore there's no effect,
that's wrong. What you can say is, we can't be confident there is an effect but there probably is an effect, and more
study is needed."
I'm not sure what the basis is for Frank's claim that there probably is an effect, but I'm not a statistician by any means. If not using arbitration clauses increased lending costs, I remain confused why more credit card issuers don't use arbitration clauses out of self-interest (roughly half don't use arbitration clauses), or alternatively, where are the credit card issuers that don't use arbitration clauses and are charging more than those who don't? I keep waiting for someone who supports the OCC position on this to address that point, but I'm not aware of anyone who has. Perhaps that's because they can't.
UPDATE: Please see the comment below for the basis of Frank's claim and an argument for why some credit card issuers don't use arbitration clauses.
Ted,
Per the OCC paper, the sample size in both studies was 308,737 observations for which the total cost of credit was calculated. The number of variables included in the regression was not egregious; given the structure of the analysis there is no credible economist out there that would call such a sample small.
Personally, I am not persuaded by the argument that credit card issuers haven’t gone through the hassle of adding arbitration clauses because (1) courts formerly didn’t enforce arbitration clauses; and (2) the Bureau will invalidate them anyway. I don’t see why it would be a hassle to amend the agreements to be used for new customers because the issuer could simply add a new clause to the contract, which consumers don’t read anyway. As for existing customers, many issuers retain the right to add new terms simply by sending them to cardholders (so called “bill-stuffers”); if the customer continues to use the card after receipt of the new terms, the customer is deemed to have accepted the new terms. The issuer can even include the new terms with the bill to avoid the expense of an additional mailing. If the reduced prices are as dramatic as the OCC claims, that hassle would be well worth incurring. While some courts have refused to enforce arbitration clauses, plenty of others have, including the United States Supreme Court. That is probably one reason why roughly half of all credit card issuers use arbitration clauses. As for the Bureau, it was not obvious to me that the Bureau would invalidate them, though I hoped they would; indeed, I heard a leading industry lawyer as recently as last spring predict that the Bureau would not issue its rule. Given the challenges that rule faces, including a CRA challenge and litigation, it remains unclear to me that the rule will go into effect, though again I hope it will.
Benefits of arbitration are reduced because of litigation and regulatory uncertainty. It’s only very recently that courts have started following the law and enforcing the clauses, and they still do so inconsistently. It’s rational for a bank to decide that an arbitration clause isn’t worth the hassle if it’s just going to add litigation expense to put the entity right back where it would be in the first place without a clause or if it will be retroactively invalidated by a regulator who made its biases clear.
The basis for Frank’s claim that “there probably is an effect” is the CFPB’s own data, which showed a p-value of 0.88. The CFPB then dishonestly characterized this as “There is no effect” and “There is no evidence of an effect,” rather than “We have a small sample size that shows an effect, but can’t be 100% certain that this isn’t noise in the data.” Given the large magnitude of the effect, and that there was a greater likelihood that the effect was larger than 6% than that the effect was 0 (and a better than even chance from the data that the effect was greater than 3%), the CFPB’s approach to its own statistics has been politicized and irresponsible.