Site icon Truth on the Market

The CFPB’s Misleading Slant on Competition in Credit-Card Markets

In yet another example of interagency cheerleading from the Federal Trade Commission (FTC), Chair Lina Khan recently touted the work of the Consumer Financial Protection Bureau (CFPB) on payments networks:

 Hmmm, does it? How so? And what ought one to do with that information?

A caveat: I’ve spent many years on competition issues, but I haven’t done much work on credit- card competition. I’ll focus on some rather straightforward points, but for deeper dives on specific issues to do with credit cards and competition (and regulation), see my International Center for Law & Economics (ICLE) colleague Julian Morris on the Credit Card Competition Act here; and Julian, Todd Zywicki, and Geoff Manne on payment-card interchange fees here.

Of course, this issue is not exactly in the FTC’s wheelhouse, either. As they say on their credit-card web page, “most credit cards are issued by banks, which are outside FTC’s jurisdiction.” And a tweet is just a tweet. Still, there are staff at the FTC with considerable experience in economic research and the FTC does have jurisdiction over nonbanks that deceptively market credit cards. 

There are other connections between the FTC and the CFPB. For example, the agencies share enforcement responsibility for the Fair Credit Reporting Act (FCRA), which sets out requirements for companies that use data to determine creditworthiness, insurance eligibility, suitability for employment, and to screen tenants. And, as it happens, Rohit Chopra, the current CFPB director, served as an FTC commissioner from 2018 right up until he assumed leadership of the CFPB (and perhaps even a bit after that, via “zombie votes”). 

Turning back to Khan’s tweet (yes, it’s just a tweet, not an article, or congressional testimony, or a lawsuit) touting the CFPB finding, it links not to CFPB research, but to a press release, which itself links to a “data spotlight”  based on, among other things, an October 2023 CFPB report on the consumer credit-card market

That’s fine, in and of itself. The underlying 175-page report is required by statute and issued biannually. Both the CFPB and the FTC have reporting obligations. Moreover, consumer education can be a useful and low-cost intervention that better enables consumers to participate in competitive markets. So, an agency—independent or otherwise—might do actual research, and it might report on that research in various ways for various readers. Primary research can be translated into a substantial (if, perhaps, less technical) report for lawmakers and others. A long report might get an executive summary. And a report can inform more accessible publications aimed at consumers, businesses, or others. A gloss here and a graphic there can provide easily digested, material information.

In this case, the report itself responsibly notes some of its limitations:

The limitations inherent to the CFPB’s methodology in this report are substantially similar to those inherent in the CFPB’s previous reports on the credit card market. All results reported from data throughout this report aggregate results from multiple industry participants. Each source has particular limitations, as not all data rely upon consistent definitions or cover the same periods, products, or phenomena. Additionally, the available data generally do not allow for definitive identification of causal relationships. Accordingly, correlations presented throughout this report do not necessarily indicate causation.

So, among other things, we see here lots of talk about averages, disparate—perhaps not easily integrated—data, and correlations, but not very much about causation.

There’s a fair bit going on in the report, if not the data spotlight. For today’s purposes, I want to focus on just a couple of things.

Concentration and Competition

Much of the report’s discussion regards competition among cards and issuers, describing various dimensions of competition and innovation over a highly differentiated product space. Consumers likely know that card offers may vary along multiple dimensions—including, among others, signing bonuses; rewards programs; interest rates (reported as maximum annual percentage rates); over-limit charge policies; fees (late fees, cash-advance fees, etc.); and–perhaps of specific relevance to offerings from larger banks–international fees and purchase protection.

As the report notes, over-limit transaction fees are highly regulated, and have largely been eliminated. And, of course, a given consumer’s credit limit may vary (and may be adjusted) even for a given card, obtained at a time certain. 

Two credit cards, both from the same issuer (say, for example, Citibank) using the same network (say, for example, Visa) obtained in the same week under the same credit rating (e.g., a certain FICO score) do not necessarily—or even likely—offer the same bundle of terms. That makes for some complexity. Then again, it’s a complex space (consumers, cards, issuers, networks, retailers, etc.), inputs vary, and different consumers value different terms differently.

So far, so good. But then there’s this:

About 4,000 financial institutions offer credit cards, yet a handful of issuers represent an overwhelming majority of credit card debt. The top 10 issuers by average credit card outstandings represented 83 percent of credit card loans in 2022, continuing a decline from 87 percent in 2016. The next 20 issuers by reported credit card debt accounted for 12 percent, an increase of four percentage points over the past six years. 3,800 smaller banks and credit unions account for the remaining five to six percent of the market. No single issuer outside the top 15 represented more than one percent of total credit card loans in regulatory filings.

What’s the point? Well, basic structural features of the market may be of interest, and may signal something, even if—as many have noted in discussing the new FTC/U.S. Justice Department (DOJ) merger guidelines (and as I did here):

economic learning and agency experience have tended to diminish the role of structural presumptions over the course of several decades (at least). My ICLE colleagues and I spent a good many pages (and citations) on this in our response to the draft merger guidelines. The structure/conduct/performance paradigm has been largely abandoned, because it’s widely recognized that market structure is not outcome–determinative. The view is shared, as we note, by scholars across the political spectrum.

We link to this from Fiona Scott-Morton, Martin Gaynor, and Steven Berry, and this from the Global Antitrust Institute, but there are scores of relevant comments based on a well-developed body of literature. 

Still, it seems a bit odd, and not just because 10 firms is not “a handful.” It seems odder still if we look at the CFPB data spotlight, which tells us that: “Lack of competition likely contributes to higher rates at the largest credit card companies.” 

Does it? 

Specifically, we are told: “CFPB research has found high levels of concentration and evidence of anticompetitive behavior in the consumer credit card market.”

Well, that sounds bad, even if it’s not so easily found in the underlying study. 

The data spotlight further explains: “the top 30 credit card companies represent about 95 percent of credit card debt, and the top 10 dominate the marketplace.” More specifically, the report tells us that “[a]bout 4,000 financial institutions issue credit cards,” with the top 10 issuers (by average outstanding debt) accounting for 83% of credit-card loans in 2022 and the next 20 issuers accounting for another 12%. That tracks the numbers in the report, which, not incidentally, indicates declining market share by the top 10, from 87% in 2016.  

But, 30 companies? That’s the high level of concentration?

How Low Can You Go?

According to the 2023 merger guidelines, markets with a Herfindahl–Hirschman index (HHI) score of greater than 1,800 are “highly concentrated.” Under the 2010 Horizontal Merger Guidelines, “highly concentrated” markets were those with an HHI greater than 2,500, and “moderately concentrated markets” were those with an HHI between 1,500 and 2,500. 

Many readers know that HHI is a concentration measure that sums the squares of each individual market participant’s market share. For sake of simplicity, a market with five firms, where each possesses a 20% market share, would have an HHI of 2,000; that is, “highly concentrated” under the new guidelines, but only “moderately concentrated” under the 2010 guidelines. 

If there are 30 firms, each with what I’ll round to a 3.3% market share, the HHI is 30(10.89) = 326.7. The new merger guidelines do not say what constitutes an “unconcentrated market,” but the 2010 Horizontal Merger Guidelines did: any market with an HHI below 1,500. Which is greater than 326.7, yes? 

Of course, the issuers don’t all have equal market shares. According to at least one source, the top 10 have the following shares: 17.9%; 13.3%; 12.4%; 11.4%; 10.8%; 4.5%; 3.8%; 2.6%; 2.4%; 2.3%: voila, that’s 82.4% (just did it in my head and hoping I haven’t botched the sum). 

What happens when we sum the squares of those numbers? 

320.41 + 176.89 + 153.76 + 129.96 + 116.64 + 20.25 + 14.44 + 6.76 + 5.76 + 5.29 = 950.16.

For those who have trouble with the number line, that’s less than 1,000, which is less than 1,500, which is less than 1,800, which is less than 2,500. 

Oops, I forgot the other 17.6% share of outstanding credit card debt. Well, let’s keep it simple and add a nice big square. Assume that there’s just one more firm (not nearly 4,000). If that were true (it’s not), we’d add 17.6 squared, which is 309.76 (and, of course, the largest sum of squares for any decomposition of that 17.6%). 

Add that to 950 and we’re still well shy of 1,500 (the bottom rung of “moderately concentrated” under the 2010 guidelines). Which is, of course, below both 1,800 and 2,500. 

Well, there may be examples of anticompetitive or otherwise unlawful behavior, but by their own accounting, they have not found high levels of concentration or market failure. 

Economists at the FTC don’t much rely on structural thresholds these days—not for actual competition analysis—even as HHIs might serve as quick and dirty preliminary signals and might be usefully cited before judges or juries in arguing a case. But to the extent that they are useful at all, they don’t support the CFPB’s having “found” a highly concentrated market or any issuer’s ability to exert market power.  Not one iota.

Director Chopra knows that.

The Relationship Between Competition and Rates

The spotlight tells us that “small issuers offer lower rates,” and then that their “median APRs are significantly lower than the largest institutions’ rates.” CFPB says its survey data, gathered on 643 credit cards from 156 issuers, indicates a considerable variation in the reported purchase APR, with a “spread between the largest (top 25) and small issuers across credit tiers (of) between eight to 10 percentage points.” 

I suppose that could be useful consumer information. Certainly, the fact that terms vary considerably should be useful to those consumers who do not know it—perhaps a significant subset of consumers, even if it’s a minority. Maybe that should be the leading message? Although a direct pointer to low rates and favorable terms, or how to find them (and identify them as such) might be much more helpful. 

As one might guess, the underlying details can be complicated. That’s not to say that high-level findings, clearly articulated, are not valuable. They can be. Depending on the audience, they might be far more valuable than the underlying research. But there’s an art to finding the nuggets of information gold, and another to clear communication. 

Consider, for example, that they are identifying issuers, which each issue multiple cards. Pointing to issuers might be useful to consumers, who might note that an issuer is, e.g., Citibank, Capital One, or the Bank of Missouri, and might do well to seek out issuers that offer favorable terms. But if the real issue is the APR (or the APR plus the other terms), the average or median APR of an issuer’s cards might be less important than the APR of a given card available to a given consumer (or given set of consumers). And a range of readily available APRs (however contingent) might be more useful than the large/small issuer divide. 

Also, based on the CFPB’s own numbers, the list of issuers reporting “at least one product” (one card or more) with a maximum APR of more than 30% includes nine issuers from among the top 25 and six smaller issuers. So, while the large/small distinction may be useful, somehow, it’s not exactly a clean split.

Among the reasons that it’s complicated is because that’s the maximum APR for one card and, e.g., Citibank doesn’t just issue one card. Some cards charge different interest rates for, e.g, purchases and cash advances, and not all consumers pay the maximum APR (or, indeed, any). For those who pay their balances in full in a timely fashion each month, the APR may be a curiosity, at best. For those who do not, it’s a key bit of information, although not the only one, and it’s agreement-specific, not issuer-generic. 

Consumers might also want to consider annual fees. The data spotlight tells us that “[i]n general, large institutions were more likely to charge annual fees than small institutions,” and to charge higher fees, on average, at that. But it’s “27% of large issuers’ card products” that are reported to charge an annual fee, which seems to be a way of saying that 73% of their products do not. Perhaps the more important information would be that annual fees, like APRs vary. 

Some of this is obvious to those with a bit of experience with credit, and the CFPB reminds us that most consumers have more than one credit card. Still, not everybody has such experience. And other terms might be relevant, such as the differences between the terms of an “initial offer” and those that follow; various fees, such as late fees; and, of course, rewards, which may come in the form of “miles” or kick-back credits on purchases (or even cash). One of my cards gives me a real-time credit equal to the amount of state sales tax charged to my purchases. 

And for consumers with deep subprime, subprime, and near-prime scores—that’s the bottom 18.7% of consumers, according to Experian—the ability to secure a credit card at all may be of primary concern. Try renting a car with cash.  

Some More Shuffling of the Numbers 

The CFPB’s biannual report divides consumers by credit scores into six categories, for most purposes (into five, for some, and into seven, for others), using the 300–850-point scale common to the commonly used FICO and Vantage credit scores. 

The data spotlight uses only three categories of consumers: those with “poor,” “good,” and “great” credit scores (619 or less; 620-719; and 720 or greater, respectively).

There are many ways we might divide the 300-850-point scale into categories, and I don’t know how many categories are best (or for whom, or what purpose). I suppose that simplicity is a good thing, all else equal. But one might wonder whether three of them are adequate to capture useful distinctions (for consumers, lenders, lawmakers, or anyone else) among consumers and, if so, where to draw the lines. 

The CFPB report and Experian both mark scores of 579 or less as “poor”; and both place scores of 800-or-higher in their top categories. Most consumers fall in between. What about them? 

Some good news is that average FICO scores in the United States have been climbing steadily for more than a decade: from 689 in 2010 and 2011 to 715 in 2023. I’d rather have them report median scores, but it’s a large n and what’s very likely a normal(ish) distribution, if truncated, so we’ll work with what we’re given. More good news: according to Experian’s numbers, 71.3% of consumers have scores falling into the good or better range, with about half (49.3%) falling into the “very good” or “exceptional” categories (so maybe the mean and median are not too far apart). 

What about the three categories from the CFPB data spotlight? That 715 average falls pretty darn close to the very top of the “good” category in the data spotlight and just five points shy of “great,” even if it’s only a middling “good” score according to Experian, falling below not just “exceptional” credit scores but “very good” ones, as well as the top two (plus) categories in the CFPB’s own report. 

Why tell consumers with scores in the 650-669 range that they have “good” credit, when their credit scores are (a) well-below average; (b) not considered “good” by major credit-reporting firms; and (c) in the fourth tranche from the top (of six overall) in the CFPB’s own report? Should they expect average terms? 

A hallmark of good consumer guidance—of good agency guidance, generally—is consistency between the underlying research and the accessible distillation. Sometimes the question of consistency is plain, and sometimes it’s a judgment call. If this move from five, six, or seven categories to three is a judgment call, it seems a poor one. Telling consumers that large, well-known issuers do not necessarily offer the best terms could be useful consumer information, if best coupled with other information. Telling low-score consumers that they have good credit ratings—not so much.

As presented, it seems overly simplified, obscuring differences across consumers, cards, and issuers, and tilting toward a “big is bad” line central to this administration’s competition policy. A crude implication of the report seems to be that varied terms and profits signal market power and, perhaps worse (as a consequence or as an inference), anticompetitive conduct.

Takeaways

First, the data spotlight on credit-card data touts some misleading numbers.

Second, the data spotlight and the underlying report provide (at best) an odd view of concentration in the industry, and the data spotlight’s conjecture about a “lack of competition” seems just rubbish.

Third, like the FTC’s comments to NIST in support of expanded march-in rights (the subject of my most recent Agency Roundup), it’s too much cant and not enough information. More signal, less noise, please. 

Fourth, they know better.

And fifth (through nth), enough with “big is bad. It might be, sometimes, but it needn’t be, and bigness might also offer countervailing advantages. For example, I don’t want artisanal-complex molecule drugs purchased on Etsy, thanks; and I don’t really want my own personal social network (it’s so hurtful when I don’t get any “likes” from my “friends” and my friends are me). 

Do I want to deal with large banks? Well, large enough. And, for credit purposes, I’ll shop around.