The Profitability Puzzle: Why Rising Card Losses Do Not Automatically Erase Returns
Credit card banks—not the Visa and Mastercard networks that carry transactions, and not an individual cardholder’s monthly bill—are the subject of this series. The question is how these banks can remain highly profitable when delinquencies and charge-offs are elevated. The useful answer is not a slogan about high interest rates: it is a way to read a portfolio’s income statement, separating the money earned on balances from the cost of funding and operating the accounts and from the provisions made for credit losses.
That distinction matters beyond bank accounting. Credit cards are both payment tools and revolving loans. The same institution may serve a customer who pays a statement in full and never incurs purchase interest, and another who carries a balance at a much higher rate. A bank’s reported return combines those different relationships with millions of others. A consumer’s APR, meanwhile, describes a price on one product under particular terms. Confusing the two makes the market look either more mysterious or more straightforward than it is.
The return belongs to a defined kind of bank
The Federal Reserve’s 2025 report measures the profitability of specialized “credit card banks.” For this purpose, a bank must have more than half of its assets in consumer lending and at least 90% of that consumer lending in credit cards or related plans. Eight banks met that definition at the end of 2024, holding 29% of the outstanding credit-card balances reported on all banks’ balance sheets. This is a useful lens on card operations, but it is not a claim that every issuer has the same business model or result. Diversified banks can have different costs, pricing, customer profiles, and returns.
Within that defined sample, pretax income was 3.87% of average quarterly assets in 2024, up from a revised 3.31% in 2023. For comparison, the return for all banks was 1.38%. The word “return” is doing important work here. The denominator is the bank’s average quarterly assets, not the balance on one card. The numerator is pretax earnings before extraordinary items, not the APR charged to a borrower. A 3.87% return therefore cannot be read as “the bank earns 3.87% on every account,” nor can it be compared directly with a card’s posted annual percentage rate.
The comparison also has a time-and-sample boundary. The specialized group has become smaller: the report counted 42 credit card banks holding 77% of outstanding balances in 1996, but ten banks held less than 40% in 2019. Reorganizations can move a card operation into a broader banking subsidiary and out of the specialized sample. There were no changes in the included sample from 2023 to 2024, which makes that one-year comparison more straightforward, but it does not turn the series into a perfectly comparable history of every issuer.
The headline is still striking because the business is exposed to consumer credit risk. The Federal Reserve says credit-card delinquencies rose in the first half of 2024 before ticking down in the second half, while charge-offs remained elevated. Yet the annual return increased. That is the tension this series examines: a portfolio can absorb substantial losses when the revenue and cost streams around those losses move favorably, but absorption is not immunity.
A portfolio is a set of offsetting streams
The basic model is simple enough to fit on one page and rich enough to prevent several common mistakes. A card bank begins with assets—principally balances and other account exposures—that generate interest income. It pays interest to fund its balance sheet. It incurs noninterest expenses to acquire, service, manage, and support accounts. It receives some noninterest income and fees. It also records provisions for expected loan losses. What remains, after these offsets, is the reported return on assets.
The Federal Reserve’s 2024 table makes the relationship concrete. Total interest income equaled 13.94% of average quarterly assets, while total interest expense was 3.78%, leaving net interest income of 10.16%. Net noninterest income was negative 2.69%, because noninterest expenses of 9.03% exceeded noninterest income of 6.34%. Provisions for loan losses were another 3.62%. The rounded components come close to the reported 3.87% return; they do not sum perfectly because the published figures are rounded and the reported measure includes smaller accounting details.
The year-over-year movement is more revealing than any one number. Interest income rose from 13.30% of average assets in 2023 to 13.94% in 2024, an increase of 0.64 percentage points. Interest expense also rose, from 3.38% to 3.78%, but by 0.40 percentage points. Net interest income therefore improved by 0.24 points. At the same time, provisions declined from 3.92% to 3.62%, a 0.30-point decrease. Noninterest income and noninterest expenses both edged up, leaving their net effect approximately unchanged. The Federal Reserve identifies higher interest income and lower provisioning as the main reasons profitability improved, with higher interest expense partly offsetting them.
That bridge is informative but deliberately incomplete. The table tells us how the aggregate lines moved; it does not identify how much of the increase in interest income came from larger balances, higher rates, or a change in the mix of accounts. Nor does it match a particular consumer’s APR to a particular bank’s funding or servicing cost. Those unanswered questions are not reasons to discard the result. They are reasons to keep the result at the level it actually measures: the combined performance of a defined portfolio.
This is why “losses are rising” cannot be the entire explanation. First, different loss measures describe different moments. A delinquency indicates payment trouble; a charge-off records a loss after a later process; a provision is an accounting expense for expected credit losses. Second, even a higher loss expense is only one subtraction from a larger set of revenue and cost streams. If interest income rises more quickly, or if expected-loss provisioning falls in the relevant accounting period, the final return can increase. The converse is equally important: a positive return in one year does not promise that a future deterioration will be absorbed.
The two relationships that make the puzzle consequential
The first relationship is between borrower behavior and the revenue base. Credit cards offer an interest-free grace period to customers who repay by the statement due date. The Federal Reserve reports that more than 60% of cardholders in a cited consumer survey nearly always repaid in full before incurring interest. It also estimates that about one-fourth of outstanding balances primarily reflected convenience use. Those customers may be valuable to the payment business without being major sources of purchase-interest revenue.
A much smaller group carries the revolving exposure. The report says that the 20% of accounts that revolved a balance in each of the previous 12 months accounted for about two-thirds of all revolving balances. Revolvers therefore occupy a disproportionate place in both sides of the portfolio equation: their balances are central to interest income, and their repayment capacity is central to credit risk. That does not mean every revolver is costly or every full payer is profitable. It means that the bank-wide average hides distinct economic roles.
The second relationship is between issuer structure and the prices consumers encounter. In a CFPB survey covering January through June 2023, the median reported purchase APR at large issuers—the top 25 issuers—was higher than at small issuers in every credit tier. The gap was roughly eight to ten percentage points. For an average balance of $5,000, the CFPB estimated that choosing a small bank or credit union card instead of one from the largest issuers could mean about $400 to $500 in annual savings. Large issuers’ products were also more likely to charge an annual fee, and their average annual fee was higher.
That dollar estimate is an illustration of the price difference, not a promise attached to every card or borrower. It uses an average balance and reported purchase APRs; it does not fold in rewards, grace periods, balance-transfer terms, eligibility, or every other fee. The CFPB’s sample included national, regional, and single-state products, so the existence of a lower-priced alternative does not mean that every consumer can obtain it. The comparison is valuable precisely because it makes the consumer consequence visible while leaving the underlying cause open.
These observations do not establish that the entire difference is profit, or that concentration alone determines every card price. A large issuer may have a different borrower mix, rewards package, funding structure, operating cost, or acquisition channel. But the comparison changes the practical stakes of the profitability question. A bank can report a strong portfolio return while consumers who carry balances face materially different prices depending on which issuers they can find and qualify for.
The market is concentrated enough that this is not a minor detail. The Federal Reserve reports that the ten largest issuers held 83% of outstanding credit-card balances in 2024, while several thousand other institutions offered cards but held small shares in aggregate. The CFPB likewise describes high concentration and raises concerns about practices that may limit price competition, including the visibility of alternatives. Those are evidence-based concerns, not a settled finding that every high rate is unlawful or that every smaller issuer is a better choice.
The central idea, then, is conditional. Credit card banks can stay profitable amid elevated credit stress when interest income and other revenue, relative to funding and operating costs, are sufficient to absorb the period’s loss provisions. Borrower mix determines where income and risk are concentrated. Pricing and market access influence how much revenue can be collected and which alternatives consumers see. The reported return records the outcome of all those forces together; it does not identify one force as the whole cause.
The next useful step is to put those forces on the same balance-sheet scale and follow exactly how the income and expense lines changed. That is the difference between asking whether losses “matter” and understanding how much room a portfolio has to absorb them.