Facts Matter: The Case for Rate Caps Doesn’t Hold Up — Part II

While the first part of this blog series provided a quick overview of how a rate cap would restrict access to credit using Shearer's own analysis, our deeper dive in part 2 provides a more technical analysis of the paper’s flaws

Recent debates over a proposed 10% interest rate cap on credit cards continue to cite a Vanderbilt Policy Accelerator paper arguing that such caps would not materially restrict credit access. In part one of this blog series released last month, the CBA Data Desk team showed how the Vanderbilt paper's analysis reveals what CBA has long maintained: a 10% cap would dramatically reduce credit card access for most Americans, undermining the consumers policymakers aim to help. We also laid out gaps in the paper's analysis—specifically the treatment of opportunity cost and returns in credit markets.

In part 2, we dive deeper, identifying additional technical flaws in the Vanderbilt paper.

Key Findings

1. Shearer’s own analysis shows how a rate cap would reduce access to credit for current and future cardholders. 
2. The Vanderbilt paper fails to consider or understand core risk, capital, and liquidity constraints in bank lending. 
3. The paper suggests that interchange fees could offset losses by supporting unprofitable credit score tiers, ignoring how interchange income would be impacted during a recession.
4. The paper relies on older data on the share of balances held by transactors and borrowers. Using more recent figures we show how Shearer’s profitability figures are inflated.   
5. The paper also relies on data from a period of strong credit performance and fails to account for default risk premiums, inflating its profitability figures. 
6. The paper contends that when losses increase during a recession and risk bank stability, the interest rate on cards could be increased, ignoring well-established economic and monetary policy about reducing rates during a recession to spur the economy.

The author’s proposal would reduce access to credit for current and future cardholders. 

Impact on current cardholders.

As we covered in our previous blog, the paper’s own calculations show that credit card lending under a 10 percent cap would generate losses for cardholders with credit scores below 800. With 75 percent (over 150 million) of current cardholders falling below this threshold, a 10 percent rate cap would make credit cards economically unviable for the vast majority of Americans — eliminating credit cards as a source of short-term liquidity.

The impact from credit line management.

The Vanderbilt paper presents the decision to lend and manage credit risk as binary: either the consumer gets a credit card or they don’t. In reality, banks also reduce credit lines to manage their exposure to risk.^[1]

For example, in 2020, as the economy entered a brief recession at the start of the pandemic, banks began decreasing credit lines and significantly limiting line increases. Once conditions improved, the number of line increases rose, and the number of line decreases fell (see Chart below).^[2]

Significant reductions in credit lines would drastically reduce consumer spending as consumers would not have as much capacity to spend on their credit cards. With nearly 20% of GDP tied to spending on credit cards, this would have a negative impact on America’s sustained economic growth.

Additionally, material reductions in credit lines would immediately increase utilization rates (the ratio of credit owed to credit available) for nearly every cardholder—especially for those carrying a balance. Utilization is a key factor in determining credit scores with higher utilization generally equating to lower scores. Even if line decreases are not uniform and affect some cardholders more than others, millions of borrowers would still see their scores drop quickly, affecting not only their future access to credit but also the terms under which they could obtain it including higher rates on auto loans and mortgages.

Impact on future card availability. 

The impact would not only be felt by current cardholders. Future consumers will be impacted too. Based on the analysis above, consumers with a FICO score of less than 800 looking to obtain credit through a credit card would be impacted as well. Without the ability to price the risk appropriately and avoid consistently negative returns, people looking to obtain credit who do not have 800+ FICO scores would now be effectively excluded. This would fundamentally disrupt the ability of millions of consumers to access the credit they need to weather financial shocks, spread out payments on large purchases, and fill gaps in their income to pay for essentials. 

The importance of opportunity cost, capital, and liquidity in credit card lending.

A core problem with the Vanderbilt paper is that it uses the wrong profitability measure. It focuses on return on assets (ROA) rather than return on equity (ROE). ROA measures how much profit a bank generates relative to its total assets, while ROE measures profit relative to the equity, or capital, shareholders have invested and banks have to hold. Unlike ROA, ROE accounts for the capital banks must hold to ensure they remain solvent in the event of a downturn and the relative opportunity cost of investing capital in other assets.^[3] This matters because banks are required to hold more capital against credit card loans than against less risky assets and loan products.

If credit card lending becomes riskier while generating less return, as it would under any APR cap, relative to other loans or investments, banks will shift away from credit card lending and/or limit credit access.

Return on equity and opportunity cost.

The Vanderbilt paper assumes that credit access would remain unaffected if a bank’s overall returns exceed some minimum hurdle rate – seemingly anything marginally greater than zero. The paper contends that banks have excess deposits that would be better served lending to credit cards that earn even the most marginal returns rather than sitting on a bank’s balance sheet or earning the same as the Fed Funds rate. This view is misguided and fails to consider real opportunity costs. 

The “hurdle rate” to justify investment in any lending product is not zero and is linked to investor opportunity costs. Banks need capital to support their lending. To attract that capital (shareholder equity) they must demonstrate that investors can achieve competitive returns. This opportunity cost helps to set the true hurdle rate. To justify investment in credit cards, banks must show investors can earn returns comparable to alternatives. That is the standard against which profitability must be measured.

For example, if a bank makes 0.1% return on credit cards but can make 2% returns on installment lending or investing in other markets, they will shift to the safer, more profitable alternative. 

They may not shutter their entire credit card program altogether, but they will significantly reduce the investment in credit cards and likely shift it towards the least risky consumers with the highest credit scores. 

In banking, returns need to be sufficient not just to be profitable, but to justify the allocation of equity to one activity rather than another and attract capital.

The link between credit cards, risk, and capital.

Higher risk means higher capital requirements as banks must hold more equity in reserve, reducing what they can deploy elsewhere. As a result, they must charge higher prices on riskier assets to achieve the same return, making high-yield but risky loans more expensive relative to alternatives.To illustrate how much riskier credit cards are compared to other loan types, consider the most recent Federal Reserve Stress Test results.^[4] These tests project losses under adverse economic scenarios to determine required capital levels. Higher projected losses generally indicate higher capital requirements. The chart below shows that projected loss rates on credit cards are three times higher than consumers loans and eight times higher than mortgages.^[5]

This helps illustrate why credit card portfolios require banks to hold more capital, why those capital requirements have been rising, and why card rates are higher to justify the investment. A rate cap may produce a positive return from an ROA perspective, but one that is unprofitable once equity considerations are taken into account.  

With increased risk and equity held against them, credit card products must generate a higher return on assets to justify that allocation of equity, especially compared to other, less risky lines of business.

Two loan portfolios with identical returns on assets (ROA) can have very different returns on equity (ROE) depending on their capital needs. This relationship is what makes ROE the binding constraint. Credit card businesses are evaluated on whether they generate acceptable returns after accounting for equity allocation, not before. 

Deposits and liquidity

Separate from ignoring capital constraints, Shearer fails to understand the liquidity constraints banks have to consider. In a follow-up blog, Shearer suggests banks cannot deploy deposits into anything other than loans, or that banks would prefer to lend into credit card portfolios earning marginally more than the Fed Funds rate rather than hold deposits idle. These assertions are used to reinforce the idea that banks have plenty of funds to lend to credit cards under a rate cap, even if only marginally profitable. Neither assertion reflects how banks actually operate.

Deposits are liabilities that must be repaid to customers on demand. Banks are therefore required to manage their liquidity—the ability to convert investments and capital into cash-- to meet these obligations to their depositors. This can make it more prudent to place deposits in safer, more liquid options like an account at the Federal Reserve, short-term Treasuries, or other high-quality liquid assets. These options carry significantly less risk and preserve the ability to return funds to depositors when needed, even without offering higher returns. 

By taking any excess deposits and putting them into only marginally profitable credit cards, banks will have given up this liquidity. This puts deposits at risk in a product whose returns no longer cover the risk they present and are hard to convert back into cash when depositors ask for it.

How additional data and context undermine the Vanderbilt paper’s findings.

Interchange Revenue is an Insufficient Backstop for Credit Losses.

The Vanderbilt paper seems to suggest that interchange income (fees generated when consumers use their cards) could bolster banks’ returns for cardholders whose losses exceed the revenue they generate under a 10% interest rate cap. The argument is that losses could be offset by interchange revenue and reduced rewards expenses. There are two significant problems with this framework.

^[6]Spending falls during a recession while losses increase, making interchange an unsustainable offset to overall profitability.

The paper does not account for a fundamental dynamic in interchange revenue: in a recession, consumer spending falls while credit losses increase. As spending declines, interchange revenue declines with it. At the same time, losses from lower FICO tiers rise. With transactors generating less interchange revenue to maintain profitability, continued credit card access becomes economically untenable. This is a primary reason banks do not rely on interchange alone to manage profitability and losses — they must plan for the full credit cycle, including downturns, not just the current environment. 

The paper's combined ROA calculations are flawed. 

The data Shearer uses to calculate “combined ROA” overestimates the returns one would expect in the current market.

Borrower and transactor balance shares have shifted since 2015.

The Federal Reserve Staff Report (“Staff Report”) Shearer uses to create his table provides the percentage of average daily balances (ADB) that a FICO tier has, given it is a transactor or borrower. It does not provide the specific shares of ADB held by borrowers and transactors within in a FICO tier.^[7] Shearer estimates this by using information from an earlier table in the Staff Report finding that 86.48% of overall balances are held by borrowers and 13.52% of overall balances are held by transactors.^[8] He then uses these estimates to help calculate the “Combined ROA” figures in his table. Critically, the Staff Report data Shearer uses to estimate his figures is derived from accounts originated in January 2015, skewing his ROA calculations.^[9]

More recent data helps to reveal the miscalculation. A recent Federal Reserve Board analysis using Y-14M data found that as of May 2020, heavy and light revolvers held 78% and 15% of balances respectively, with transactors accounting for the remaining 7%.^[10] A separate Federal Reserve Report using 2014–2019 data found transactors held 9.65% of balances.^[11] Both figures diverge from the 13.52% transactor share assumed by Shearer. 

Publicly available Y-14M data and CFPB Credit Card Market Report data further show that the share of accounts making full balance payments increased significantly after Q2 2020 and has remained elevated, indicating more transactors relative to revolvers than the 2015 data suggests. 2024 Y-14 data from the same CFPB report shows that the share of revolving accounts in 2024 was 49%, compared to 58.95% in the 2015 data Shearer relies on.^[12]

Using these more recent figures and the paper’s same methodology, we show that the paper overestimates the combined ROA for all credit score tiers below 840 and underestimates the combined ROA for all credit score tiers 840 and above (see Table below). For example, rather than a positive combined ROA for 800 FICO cardholders, the adjusted figure is negative. Further, it suggests that the Vanderbilt paper overestimates the return for almost all cardholders by 20 to 40 basis points, representing a change in net income of tens to hundreds of millions of dollars depending on the size of a bank’s credit card portfolio. 

10% APR Cap (or 5% + FFR)
FICO Tier	Industry’s Return on Assets (ROA) after rate cap (%)
	ROA from Borrowers	ROA from Transactors	Combined ROA before cost-cutting	Adjusted Combined ROA (using May 2020 data)	Difference (positive values indicate overestimate of ROA)
600	-5.98	16.43	-5.20	-5.60	0.40
620	-5.37	9.22	-4.80	-5.09	0.29
640	-4.49	6.68	-4.03	-4.26	0.23
660	-4.36	4.39	-3.98	-4.17	0.19
680	-4.51	2.89	-4.17	-4.35	0.17
700	-3.90	2.70	-3.52	-3.71	0.19
720	-3.12	2.89	-2.65	-2.88	0.23
740	-2.37	2.77	-1.82	-2.09	0.27
760	-1.64	2.87	-0.98	-1.29	0.32
780	-0.83	3.01	0.00	-0.38	0.38
800	0.09	3.21	1.10	0.68	0.43
820	1.20	2.01	1.54	1.41	0.13
840	2.17	0.16	1.22	1.56	-0.34
850	2.79	0.29	1.46	1.91	-0.45

The Data Used Excludes a Period of Deteriorating Credit Performance. 

The Staff Report uses data that examines credit card accounts opened from 2015 to 2017 and tracks them for six years (December 2023 being the latest). Shearer mentions this as a limiting factor in terms of interest rate spreads, but not in terms of risk and credit losses. This period was characterized by historically low charge-offs and good consumer credit card performance. In contrast, the period directly after the one covered by this data saw performance on credit cards worsen. This period, where defaults increased and consumer performance worsened, materially impacts the figures in the paper’s analysis by inflating what credit card profitability would look like through a credit cycle downturn. 

Charge-off rate differences since 2015

The average quarterly net charge-off rate for Y-14M banks during the covered period (2015Q1 to 2023Q4) was 3.14%. From 2024 to the second quarter of 2025 it was 5.74%, an increase of over 260 basis points.^[13] This would materially decrease the ROA figures in the paper’s “ROA from Borrowers” column of their 10% rate cap table as the ROA figure is pulled from adding and subtracting subsets of costs and revenues (of which, charge-offs are a large part).^[14]

Without access to the underlying granular data used by the original Staff Report, how much these ROA figures would change and how it would spread across the different FICO bins is unclear as charge-offs are not evenly distributed across FICO scores. 

We agree (as Shearer mentions in his paper) that additional analysis could be helpful if performed on cardholders during this portion of the credit cycle. However, by only considering a period in which charge-offs were comparatively low and credit performance good, conclusions drawn from this analysis exaggerate the profitability figures and downplay the negative impacts of the proposed rate cap, especially during a credit cycle downturn.

Risk premium and pricing the entire credit cycle.

The paper analyzes a static environment with historically low charge-offs. But banks must price credit across the entire credit cycle, not just favorable periods. They are also constrained from increasing the price of credit (the APR) on existing customers’ credit cards as those customers become riskier during downturns.^[15] To account for this, banks must "bake in" the future default risk of loans originated today.^[16]

The Staff Report itself acknowledges this default risk premium, though it is not included in the ROA figures the Vanderbilt paper presents.^[17] The Staff Report authors adjust their ROA figures to account for this premium.^[18]This added premium is an essential layer of credit card pricing that CBA has discussed before. Like everyone, issuers can’t predict the future, and they have to price for that risk. Incorporating this factor would push the profitability of every FICO tier <820 negative and double or triple the losses, even after accounting for the paper's proposed interchange revenue changes and marketing spend reallocation.  For example, Shearer’s analysis shows that a cardholder with a 720 FICO has a combined ROA of -2.65 before cost cutting. Accounting for the risk premium, this ROA would be -7.56.^[19]

10% APR Cap (or 5% + FFR)
FICO Tier	Industry’s Return on Assets (ROA) after rate cap (%)
	ROA from Borrowers	ROA from Transactors	Combined ROA before cost-cutting	New Combined ROA(Using Table 4 Risk Adjustment)	Difference (positive values indicate overestimate of ROA)
600	-5.98	16.43	-5.20	-13.46	8.26
620	-5.37	9.22	-4.80	-12.92	8.12
640	-4.49	6.68	-4.03	-10.73	6.70
660	-4.36	4.39	-3.98	-10.64	6.66
680	-4.51	2.89	-4.17	-10.32	6.15
700	-3.9	2.7	-3.52	-8.40	4.88
720	-3.12	2.89	-2.65	-7.56	4.91
740	-2.37	2.77	-1.82	-5.74	3.92
760	-1.64	2.87	-0.98	-4.35	3.37
780	-0.83	3.01	0.00	-2.51	2.51
800	0.09	3.21	1.10	-0.55	1.65
820	1.2	2.01	1.54	0.53	1.01
840	2.17	0.16	1.22	0.30	0.92
850	2.79	0.29	1.46	0.63	0.83

The paper ignores well-established economic and monetary policy tools and political realities. 

The paper argues that in a recession, banks can simply draw on required capital reserves for the brief period before regulators or policymakers step in to raise the cap — pointing to the fact that ROA only dipped negative for a couple of quarters during the Great Recession. But this framing understates the stakes considerably. Those reserves exist to guard against insolvency during genuine financial crises, not to absorb the consequences of government-mandated price controls. Requiring banks to simultaneously absorb multi-quarter losses, drain their capital buffers, and then — only once solvency is threatened — seek a political remedy, layers serious systemic risk onto an already stressed economy. Critically, a recession is precisely the worst moment to raise interest rates on consumers. Doing so at the very point when the Federal Reserve is trying to stimulate spending and economic activity would work directly against sound monetary policy.

The policy's dependence on political actors to adjust the cap in real time compounds the problem. The paper proposes giving Congress or a regulatory body authority to raise the cap when conditions warrant — but offers no account of why they would actually do so. During an inflationary period like 2024, for instance, the Federal Reserve would need higher rates to cool spending, but any governing body controlling a credit card rate cap would face enormous political pressure not to raise it, given that voters would already be carrying high balances due to rising prices. This is precisely why the Federal Reserve was designed to operate outside the political process. Rather than addressing this tension, the paper's own recommendation illustrates the danger: a rate cap as designed in these proposals would undermine monetary policy in both directions, slowing recovery from recessions and prolonging inflationary cycles.

Conclusion

The debate over credit card interest rate caps deserves careful, data-driven analysis, but that analysis must reflect how banks and credit card markets actually operate. Failing to account for what happens in a different part of the credit cycle, misunderstandings about return on assets, return on equity, deposits and liquidity, optimistic assumptions about interchange revenue, and ignoring impacts of reductions in credit lines and access all paint an incomplete understanding of the credit card market and the impacts a rate cap would have on the consumers.

Good policy starts with good analysis. In this case, the analysis doesn’t hold up.

---

^[1] See, for example, Ibid. pp. 132-136

^[2] See also, Ibid Figures 94 & 95.

^[3] When we say “capital” or “equity” what we are generally referring to is the equity that can include owner’s equity (stock), surplus, retained earnings, and other accumulated income and undistributed profits. 

^[4] We can also see the higher risk from discussions of risk-weighting in the Staff Report at p.40-42 and BPI report at https://bpi.com/the-basel-proposal-what-it-means-for-retail-lending/; For brevity we do not discuss risk-weighted assets in this piece, but encourage readers to understand their role in the pricing f loan products.

^[5] DFAST Stress Test Results 2025. Federal Reserve Board of Governors at https://www.federalreserve.gov/supervisionreg/dfa-stress-tests-2025.htm

^[7] For instance, the Staff Report shows that 1.90% of borrower balances are held by those with a 840 FICO, and 10.9% of transactors’ balances are held by those with an 840 FICO, but it does not show us how much of 840 FICO cardholder balances are held by borrowers and transactors respectively. 

^[8] See Shearer footnotes 69 and 72; Staff Report Table 1, Panel B, p.59. ($165.78B/$191.70B = 86.48%)

^[9] Staff Report Table 1, p.59

^[10] See “The Effects of the COVID-19 Shutdown on the Consumer Credit Card Market: Revolvers versus Transactors”. We note that while balances fell during the May 2020 period, they fell consistently for transactors and revolvers. 

^[11] See FEDS Notes “Credit Card Profitability” Table 1

^[12] See CFPB Credit Card Market Report figure 47 on p.74; Staff Report Table 1, Panel B; Shearer paper footnote 69. 

^[13] Source: FR Y-14M Data, Federal Reserve Bank of Philadelphia at https://www.philadelphiafed.org/surveys-and-data/large-bank-credit-card-and-mortgage-data. We note that this is the same dataset used by the Federal Reserve Staff Report cited by Shearer.

^[14] The Federal Reserve Staff Report ROA figure for borrowers involves the following to calculate ROA: Adding together all streams of income and expenses per dollar of ADB, over its lifetime, gives us the banks’ return on assets (ROA) for the account. Specifically, ROA equals interest spread minus net charge-offs (default-adjusted credit spread), plus net interchange income (interchange minus rewards), plus the fee income rate, minus the operating expense rate. The last column of Table 2 reports the ROA across FICO bin and borrower/transactor group

^[15] Regulatory constraints introduced by the Credit Card Accountability Responsibility and Disclosure Act of 2009 (Credit CARD Act) placed restrictions on banks’ ability to adjust the contractual spread (over an index rate) on existing credit card balances in response to evolving risk assessments. See 12 CFR 1026.55(b)(4).  

^[16] A recent paper examining increasing credit card APRs showed how these restrictions on pricing, among other changes in the credit card market, have impacted APR margins over time.

^[17] See Fed Staff Report at 33. This default risk premium is distinct from the charge-off cost and still allows for margin on credit card loans. 

^[18] See Staff Report p. 63; Table 4.

^[19] We apply the risk adjusted ROA figures from Table 4 of the Staff report to borrower ROA only and then recalculate combined ROA using Shearer’s own method. However, this likely underestimates the reduction in ROA as the risk adjustment would likely apply to transactors as well as they would also present an undiversifiable risk.