Is the Subprime Segment of the Credit Card Market Concentrated?

On Feb. 19, 2024, Capital One announced its plan to acquire Discover Financial Services, a combination that would create the largest credit card issuer in terms of card balances and third largest in terms of purchase volume.[1] Because of this size factor, the merger proposal has generated some reflexive criticism, based on the oversimplified notion that it would create a dominant firm capable of raising prices.[2]

Some critics have gone further, claiming that subprime consumers constitute a distinct submarket for considering competitive effects, and that they would be more vulnerable to price increases because they would have fewer options than prime borrowers. However, no evidence has been presented to support the view that the subprime segment comprises a relevant submarket, which seems inconsistent with how the market functions in practice.

One calculation recently receiving attention assessed the effect of the proposed merger on concentration in the subprime segment based on consideration of 12 major credit card issuers and a definition of subprime based on a specific credit score threshold. This analysis concluded that the proposed merger could increase the Herfindahl-Hirschman index of concentration in the subprime segment to a level that exceeds the generally applied threshold for indicating potential competitive concerns in the banking merger context (the 1800 HHI threshold).[3]

In this blog post, we examine the impact of the Capital One-Discover merger on concentration levels in the subprime and prime consumer segments of the credit card market, as well as on the overall market. We show that after expanding the analysis to include just a few additional card issuers with significant subprime presence, beyond the 12 included in the aforementioned analysis, the subprime segment is far less concentrated than critics suggest. And even our expanded analysis understates the degree of competition, as it omits from the concentration measure numerous other card issuers and does not consider competition from providers of substitute forms of credit such as “buy now pay later” (BNPL) products.  

Our analysis demonstrates that even the subprime consumer segment of the credit card market is quite competitive and that the proposed merger would not come close to creating a level of concentration meriting concern. The finding that post-merger concentration in the subprime segment would exceed 1800 is based on restricting the market concentration calculation to 12 banks and does not hold up to a more accurate assessment. Thus, the analysis demonstrates that even if subprime is to be considered a separate market for evaluating the competitive effects of the proposed merger (and there is little reason to believe it should be), there is no indication that the merger raises competitive concerns.

Data Sources

To calculate the level of concentration in the subprime segment of the consumer credit card market, we use various sources. First, we obtain the amount of consumer credit card loans held by each institution held as of the fourth quarter of 2023 from bank or savings-and-loan holding company FR Y-9C reports (or, in the case of independent banks and credit unions, their Call Reports). Second, we gather information about each company’s share of prime versus subprime credit card loans from annual corporate SEC 10-K filings and annual reports to shareholders.

Although annual filings also report total credit card loans, we use credit card balance outstandings in FR Y-9C and Call Reports because corporate 10-K filings or annual reports to shareholders are not available for many credit card issuers. To best assess the degree of concentration in the subprime consumer segment of the credit market, we include all depository institutions with at least $1 billion in credit card outstandings, using FR Y-9C and Call Report data on card loans outstanding as these are the most consistently available sources. 

For companies that reported a breakdown of credit quality in their 10-K filings or annual report, we use that information directly to determine the institution’s subprime share. If this information is not available, but the lender is known to serve only subprime borrowers (for instance, an institution that issues only “credit builder” or secured cards), then the institution’s share is calibrated at 100 percent. Otherwise, the institution’s subprime share is estimated based on the credit card delinquency rates from Y-9C or Call Report data.[4]  

To alleviate the concern that the use of the information on total balances from the Y-9C and Call Report while relying on company 10-K and annual reports for information on subprime borrower share may produce inconsistencies, we conduct a validation exercise described in Appendix 1. There, for the banks with information available from 10-K or filings or annual reports, we compare the credit card balances from those sources to the balances from the Y-9C and Call Reports and find that they align closely.

HHI Calculations

Before conducting our own calculations, we attempted to replicate the analysis cited previously that purported to demonstrate lack of competition within the subprime card market based on a calculated concentration measure. That analysis measured concentration based on the lending activity of 12 U.S. banks, using the annual 10-K filings or investor reports of these banks (Appendix Table 2). It found that the proposed merger would increase HHI for subprime credit card loans from 1420 to 1922. Upon revisiting this analysis, we found a minor error and calculated the corrected HHIs to be 1404 pre-merger and 1884 post-merger HHI, indicating that the merger would still exceed the 1800 threshold.

However, measuring concentration in the subprime segment of the consumer credit card market analysis using just these 12 banks is insufficient because it fails to account for numerous other significant credit card issuers serving subprime borrowers. While data on the full set of issuers serving this segment of the mark admittedly are difficult to come by, with reasonable effort the analysis can be expanded to include a larger number of participating firms. We have taken two additional steps toward a more comprehensive assessment.

First, we obtained information on subprime share of total card balances for three banks  with total card balances exceeding $1 billion, in addition to the 12 already noted. We then incorporated these three, plus three others known to be subprime specialists (for which we assume a portfolio comprised entirely of subprime.) and recalculated the HHI for the total market and the subprime and prime credit segments of the market pre- and post-merger.

Second, we extended the analysis yet further by incorporating lenders that exceed the $1 billion size threshold but for which information on share of subprime balances is not directly available. More specifically, we use regression analysis to estimate the prime/subprime breakdown for banks that do not explicitly report this information. This estimation is based on banks reported delinquency rates on credit card loans, which serve as a proxy for the proportion of subprime borrowers in their credit card portfolios.

It is important to emphasize at the outset that even this expanded analysis falls short of including the full range of competitors in the market, as it omits numerous smaller banks (below the $1 billion aggregate loan balance threshold) that issue credit cards and various nonbanks, including fintechs.[5]  Information is scarce about the subprime segment shares of nonbanks. For these reasons, even these expanded calculations will still overstate concentration in the subprime segment.

Table 1 presents the HHI calculations based on the 18 credit card issuers with balances exceeding $1 billion for which we have direct information on subprime share of card balances.[6] The results are as follows:

  • Panel A shows that the proposed merger increases the HHI for the total consumer credit card market from 1064 to 1301.
  • Panel B indicates that the proposed merger increases the HHI for the prime credit card segment from 1117 to 1318; and
  • Panel C shows that the proposed merger increases the HHI for the subprime credit card segment from 1107 to 1488.

These findings demonstrate that when considering all major credit card issuers in the market, the post-merger HHIs for both the prime and subprime segments remain well-below the 1800 threshold indicating potential competitive concerns.

Table 1. Credit Card Market Concentration for Issuers with Breakdowns, by Consumer Profiles.

Panel A: Total Credit Card Balances

table 1 panel a-total credit balances

Panel B: Prime Credit Card Balances

table1-panel B-prime credit card balances

Panel C: Subprime Credit Card Balances

table 1-panel c-subprime credit card balances

As noted earlier, not all banks and credit unions report a breakdown of their credit card lending by range of borrower credit score. To incorporate other financial institutions with significant roles in the credit card lending space, we used regression analysis based on the reported credit card delinquency rates from the call reports and Y9-C fillings.

We used a sample comprised of 20 banks from Table 1 for which we had both subprime market shares and delinquency rates. Navy Federal Credit Union and Bread Financial are excluded from the regression sample because one is a credit union and the other is a nonbank, and Stride Bank was excluded due to not reporting a delinquency rate. The regression analysis, which used the subprime market share as the dependent variable and credit card delinquency rates as the independent variable, yielded an Adjusted R-Squared of 80 percent. The regression results suggest that a 1-percentage-point increase in delinquency rates is associated with an approximately 7 percent increase in the predicted subprime market share, with a constant value around -4 percent.

We then applied the regression equation to predict the subprime versus prime consumer shares for all other financial institutions with at least $1 billion in credit card balances that submitted a delinquency rate in their financial reports. Table 2 presents the HHI calculations for this expanded sample, where the estimated shares are highlighted in blue font.

Table 2 shows less pronounced effects of the proposed merger on market concentration when including an even broader set of banks, as expected. The results are as follows:

  • Panel A shows that the proposed merger increases the HHI for the total consumer credit card market from 984 to 1202.
  • Panel B indicates that the proposed merger increases the HHI for the prime credit card segment from 1038 to 1224; and
  • Panel C shows that the proposed merger increases the HHI for the subprime credit card segment from 1004 to 1348.

These findings show that providing more information on the quality of consumers at the credit card issuer level is crucial to assessing the structure of the credit card market and the competitive effects of the proposed merger.  

It is worth noting that credit unions represent a large share of the credit card market, as seen in Table 2 and confirmed in recent issues of the Nilson report. Hence, for the purpose of assessing the market’s concentration we have included credit unions with over $1 billion in outstanding credit card loans in our analysis. However, we also present HHIs without credit unions in Appendix Table 3 to demonstrate that the post-merger HHI remains far below the 1800 threshold, regardless of whether credit unions are included or excluded from the calculation.

Table 2. Credit Card Market Concentration for Issuers with and without Breakdowns, by Consumer Profiles.

Panel A: Total Credit Card Balances

Panel B: Prime Credit Card Balances

Panel C: Subprime Credit Card Balances

Additional Considerations

The preceding calculations clearly demonstrate that were the prime and subprime segment to be considered a relevant submarket for competitive analysis, post-merger concentration in these segments would be well below the level commensurate with competitive concerns. Each segment would remain highly competitive.

Nonetheless, it should not be overlooked that the credit card market and these individual segments are even more competitive than these calculations indicate. In the first place, as previously pointed out, the calculations of necessity omit thousands of smaller banks and credit unions along with many nonbanks that issue credit cards including to subprime borrowers due to lack data on their share of card balances that are subprime.[7] For the nonbanks, delinquency rates are also unavailable, and regarding the numerous small bank card programs we are hesitant to apply our regression model that is based on large bank portfolios. The presence of these numerous smaller banks and of nonbanks, including some that are subprime specialists, implies that the market overall and the subprime and prime segments individually are more competitive than indicated by the calculated HHIs.

Not only that, but credit card issuers face additional competition from very close substitutes to credit cards, including from line-of-credit products and from BNPL products, further ensuring that credit card market pricing will remain competitive. In fact, the Consumer Financial Protection Bureau earlier this month announced that BNPL lenders, which are primarily fintechs, will be classified by the Bureau as credit card providers. According to a recent study published by the Federal Reserve Bank of Boston, BNPL is increasingly being used by “financially vulnerable” consumers as a substitute for credit cards.[8]

Buy now, pay later (BNPL), a short-term, interest-free credit option for retail purchases, is becoming increasingly popular, and evidence indicates that its use is significantly higher among financially vulnerable consumers and disproportionately high among women, Black, and Latino consumers. At roughly 9 percent (as of fall 2023), the share of all consumers using BNPL is still relatively low, but it has increased about 40 percent from two years earlier…  Consumers can use the service to make online or in-store purchases and repay only their purchase amount in installments over a short period, typically four installments over six weeks. … BNPL can thus provide short-term credit to consumers who lack alternative sources of credit, may not have credit cards, or have low credit limits.

It also bears emphasis that while we conducted these calculations separately for the prime and subprime segment, it is by no means evident that these should be considered separate submarkets for antitrust purposes. Any separate treatment of these segments by the regulatory agencies must be grounded in comprehensive evidence that they indeed comprise relevant submarkets for competitive analysis.

Basic facts about how the market functions seem inconsistent with delineating a separate submarket for subprime, especially one that is based on a credit score threshold alone. Issuers compete along various price and nonprice dimensions, including both interest rate and credit limit. For instance, a lender that offers a larger credit limit in exchange for a higher interest rate to borrowers with lower scores may end up with a higher share of that score segment. But that does not mean that they face any less competition—consumers still have the option of the lower interest rate and lower limit.

Moreover, lower score segments are heterogenous. Some consumers have lower scores because of limited credit histories, such as young people or immigrants; others perhaps because of recent credit difficulties that have subsided, and others because of past, serious credit problems, and so on. Various credit card products are tailored to these different situations.

The factor that must be demonstrated for meaningful delineation of a submarket is limited options for the consumers in that submarket, which is not necessarily reflected in the segment concentration measure. It is well known that there are numerous issuers willing to extend credit in lower credit score segments, although they have small, existing market shares in these segments.[9] At the same time, there may be niches within the subprime segment, such as secured cards, which may be less substitutable but which are at the same time less concentrated in the sense that market shares are more evenly distributed across lenders present in the segment.

Finally, ease of entry has long been viewed as a mitigating factor in bank antitrust analysis. Even if it were the case that the subprime segment of the credit card market was becoming less competitive due to a merger, numerous existing credit card issuers would be waiting in the wings to enter the segment or expand their current activity in it, as there is virtually no barrier to such entry.

Conclusion

Capital One’s plan to acquire Discover has caused considerable debate about the potential implications of the merger for consumers. Regulators might decide to examine competition in the subprime segment of the credit card market if they find reasons to believe that this segment constitutes a distinct and relevant submarket.

If the subprime consumer segment of the credit card market merits separate scrutiny, our analysis indicates that the segment is highly competitive and would remain so even after the proposed merger. Our calculated concentration index for the subprime segment falls well below the threshold for raising competitive concerns. Due to lack of data, our calculation of market concentration excludes numerous additional sources of competition, and therefore, the market is even more competitive than our calculation indicates.

No evidence has been put forth by critics of the proposed merger to define the boundaries of the subprime segment and establish that consumers in this segment are sufficiently isolated for it to be considered a distinct submarket for antitrust purposes. Basic facts about the functioning credit card market suggest that the subprime segment is not a distinct submarket. Any separate treatment of these segments must be grounded in comprehensive evidence that they indeed comprise separate submarkets for antitrust purposes.

An appropriate competitive analysis of Capital One’s acquisition of Discover should provide a complete picture of the concentration and competition in credit card lending. Additional relevant factors to be considered include the role of buy-now, pay-later as a substitute for card lending and the ease of entry or expansion into the subprime segment of issuers who might not currently be active in that segment.

To read the full post with the Appendix, please click here or click on the download button below.


[1] Balance projections are based on FR Y-9C data as described below, and purchase volume projections rely on an article from U.S. News and World Report.

[2] Americans for Financial Reform. (2024). The Anticompetitive Effects of the Proposed Capital One-Discover Merger. Retrieved from https://ourfinancialsecurity.org/wp-content/uploads/2024/04/AFREF-Capital-One-Discover-Merger-Comment-4-2024.pdf

[3] Kress, Jeremy [@Jeremy_Kress]. Twitter, Feb. 27, 2024, https://x.com/Jeremy_Kress/status/1762576947761279089

[4]  We use 10-K filings for JPMorgan Chase, Citi, Capital One, Bank of America, Discover, American Express, Synchrony, Wells Fargo, U.S. Bancorp, Goldman Sachs, Bread Financial, PNC, Citizens, Fifth Third, Regions and KeyCorp. We use annual reports for Barclays and Navy FCU.

[5] The lending activity of the smaller banks and of nonbanks are included only to the degree that they partner with larger banks and some of their lending  is incorporated into the reported card balances of their partner banks.

[6] These 15 institutions are almost the same as Nielson’s list of top 15 credit card issuers (14 of the 15 are on the Nielson list.) The Nielsen Company. (2023). Largest US General Purpose Credit Card Issuers, February 2023. Retrieved from https://nilsonreport.com/newsletters/1236/

[7] The FR Y-9C and bank and credit union Call Report data indicate a total of 3,755 depository institutions having credit card balances.

[8] See Joanna Stavins, “Buy Now, Pay Later: Who Uses it and Why.”  Current Policy Perspectives 2024-3. May 22, 2024. Available here.

[9] The CFPB’s Survey on the Terms of Credit Card Plans for evidence of this fact. The TCCP survey includes information on the range of consumer credit scores covered by issuers’ individual credit card programs.