The Cross-Subsidy: Who Actually Pays for Free Financial Products
The Cross-Subsidy: Who Actually Pays for Free Financial Products
A customer with $40,000 in checking pays nothing for their account. A customer with $340 pays overdraft fees several times a year. Both hold the same product, and the second is funding the first. A shopper paying cash at a supermarket pays the same posted price as one paying with a premium rewards card, and part of that price covers the acceptance cost that funds the rewards — so the cash customer is contributing to someone else's points. Neither transfer was designed. Both follow from the same structure: a bundled product where the person generating revenue and the person receiving the service are different people, and where the price signal linking them is deliberately hidden from both.
In this report
The core thesis
A cross-subsidy exists when a product's price to one group doesn't reflect the cost of serving them, with the gap covered by another group. It's ordinary in many industries and generally unremarkable — restaurants price wine above cost to keep food prices down, and nobody objects.
Consumer finance has three features that make its cross-subsidies worth separate examination.
They're large. Not garnish on a price — in several cases the subsidising revenue is the primary economics of the product.
They're invisible to both sides. The customer receiving free checking doesn't know who's paying. The customer paying overdraft fees doesn't know they're funding someone else's account. Neither party can evaluate a transaction they can't see, which distinguishes this from the restaurant case, where the wine drinker knows the wine is marked up.
They run in a consistent direction. This is the part that matters and the part we'd defend most firmly: the subsidising behaviours are associated with lower balances and thinner buffers, while the subsidised behaviours require the resources to avoid fees. So the transfer flows from less affluent customers to more affluent ones, systematically, across several products at once.
Our position is not that transfers are wrong. It's narrower and, we think, harder to argue against: an invisible transfer running in a regressive direction is one nobody has consented to and nobody can assess. A cross-subsidy defended in the open might well survive scrutiny. These aren't defended in the open because they aren't visible.
The customer receiving the free product doesn't know who pays for it. The customer paying doesn't know they're funding someone else. A transfer neither party can see is one neither party has agreed to.
Free checking
The clearest case. A checking account costs real money to provide — infrastructure, processing, fraud, compliance, branches, support. Most customers pay nothing.
Where the revenue comes from:
- Overdraft and non-sufficient funds fees, historically the largest direct source, and heavily concentrated. A relatively small share of accounts incurs charges repeatedly, and that share generates the substantial majority of the revenue — the pattern our overdraft analysis documents.
- Net interest on balances, where the institution earns more on deployed deposits than it pays out.
- Debit interchange on card transactions.
- Ancillary fees — wires, stop payments, paper statements, out-of-network withdrawals.
The concentration is the finding. The revenue isn't spread thinly across all customers; it's paid heavily by a minority. And the characteristics that produce it — low balances, thin buffers, timing mismatches between income and obligations — are the liquidity conditions our illiquidity analysis describes.
Stated plainly: the customer who never overdraws because they maintain a buffer receives a service funded substantially by the customer who overdraws because they can't. Both are behaving reasonably. Neither chose the arrangement.
The second revenue source deserves its own note. Interest earned on balances means customers with large balances do contribute — through foregone interest rather than fees. That's a genuine qualification and it partly offsets the picture. But it's a contribution the customer also can't see, and in periods when deposit rates lag market rates it's substantial, which means even the apparently free-riding customer is paying something invisible. The whole structure is opaque in every direction.
Rewards cards
The larger transfer, and the one most people would deny they participate in.
The mechanics compressed — the full treatment is in our interchange analysis. A merchant accepting cards pays an acceptance cost that varies by card type, with premium rewards cards costing meaningfully more. Most of that flows to the issuer, which funds rewards from it.
The step that creates the transfer: merchants almost universally set one posted price regardless of payment method. Card acceptance cost is a cost of doing business, embedded in prices charged to everyone.
So:
| Customer | Pays | Receives | Net |
|---|---|---|---|
| Cash | Posted price including acceptance cost | Nothing | Contributor |
| Basic debit | Posted price | Little or nothing | Contributor |
| Standard credit | Posted price | Modest rewards | Roughly neutral |
| Premium rewards | Posted price | Substantial rewards and benefits | Recipient |
And the correlation that makes it regressive: access to premium rewards cards requires the credit standing and income that qualify you for them. A household paying cash because they're unbanked — the population in our unbanked analysis — pays the embedded cost and can receive nothing, by construction.
A second layer compounds it. Rewards cards are most profitable on customers who revolve, and the interest paid by revolvers funds rewards enjoyed by customers who pay in full — the dynamic our revolver analysis describes. So the same product contains two nested transfers running the same direction, and the customer who benefits most is the one who spends heavily and never carries a balance.
The other cross-subsidies
Once you look for the structure, it's widespread:
- No-fee brokerage funded by order flow and margin lending — active traders and margin borrowers subsidise the passive.
- Free credit monitoring funded by lead generation, where users who convert to products fund the service for everyone.
- Deposit rates below market in an environment where they could be higher, which is a transfer from inattentive depositors to the institution and thence to borrowers and to depositors who shop.
- Grace periods on cards, funded by those who don't use them.
- Free tax preparation funded by refund products, per our refund analysis, where users of high-cost refund advances subsidise free filing for others.
- Fixed-rate lending, where borrowers who don't refinance when rates fall subsidise those who do — an inattention transfer of the same shape as our search analysis describes.
The common structure across all of them: a service priced at zero, funded by a behaviour concentrated in a minority, where the behaviour correlates with either constrained circumstances or inattention. Constraint and inattention are the two revenue sources, and both are distributed unevenly.
Why the direction is regressive
This is the claim requiring the most support, since regressivity isn't automatic — a cross-subsidy could just as easily run the other way.
Three mechanisms push it consistently:
Fee-generating behaviours require thin buffers. Overdrafting requires having no cushion. Late fees require a timing mismatch between income and obligations. Revolving requires an expense exceeding available cash. Every one of these is a symptom of the liquidity constraint rather than the wealth constraint — which is why they appear among households that look comfortable on a balance sheet, and why they're concentrated among those who aren't.
Fee-avoiding requires resources. Maintaining a minimum balance requires having it. Paying in full requires cash flow that covers the month. The customer avoiding fees isn't more careful; they're better resourced, and the distinction matters because the fee structure is frequently defended as a charge for carelessness.
Benefit intensity scales with spending. Rewards scale with volume, and volume scales with income. So the largest recipients are the largest spenders, and the largest contributors are those who spend on necessities at posted prices without receiving anything back.
Which is a system that nobody designed and everybody rational participates in. The bank offering free checking is responding to competition; the customer maintaining a balance is optimizing sensibly; the merchant setting one price is avoiding complexity. No participant is doing anything objectionable, and the aggregate is a persistent regressive transfer — the same emergent-outcome-from-reasonable-behaviour structure our poverty premium analysis documents.
What happened when one was unwound
The debit interchange cap for larger institutions is the closest thing to a controlled experiment on this question, and the results are worth taking seriously in both directions.
Research examining affected banks found reductions in free checking availability, increases in monthly maintenance fees, and higher minimum balance requirements relative to unaffected institutions. Removing the revenue that funded a service did not remove the cost of providing it.
Three readings, and honesty requires holding all three:
The critical reading: the reform harmed the customers it aimed to help. Households lost free accounts and faced balance requirements they couldn't meet, while merchant savings were captured or passed through incompletely.
The defending reading: the reform made a hidden cost visible. A monthly fee is a price a customer can see and shop; an invisible cross-subsidy isn't. Making the cost explicit is an improvement in the structure even where it's unwelcome in the amount.
Our reading: both are right, and the episode is evidence for the analysis rather than against the reform. It demonstrates that the cross-subsidy was real and load-bearing — if free checking hadn't been funded by interchange, capping interchange wouldn't have changed checking. The predictive claim was confirmed. Whether the change was net beneficial depends on a distributional judgment the data doesn't settle.
The general lesson for anyone contemplating unwinding one: the subsidised product reprices. Always. A reform removing subsidising revenue without addressing the underlying cost is a reform that moves who pays rather than what it costs — and it should be argued for on those grounds rather than as a straightforward gain.
The honest case for cross-subsidy
A report that only prosecuted would be advocacy. The defence is real and deserves its own section.
Cross-subsidy expands access. This is the strongest argument. If checking accounts were priced at cost — a monthly fee covering the true expense — some households would go without. Free checking funded by fees keeps people banked who would otherwise be unbanked, and being unbanked carries costs our unbanked analysis documents as substantial. A regressive transfer that keeps people inside the banking system may be better than an equitable price that pushes them out.
The fee is avoidable and the service isn't. Overdraft fees are incurred by a choice, in a sense — the customer initiated a transaction without funds. Framing every fee-payer as a victim overstates it, and some overdraft use is deliberate short-term credit that the user prefers to the alternatives.
Interchange funds real services. Merchants receive a payment guarantee, fraud protection, and a settlement infrastructure. Acceptance cost isn't purely a transfer to rewards; a portion buys something merchants genuinely value and would otherwise provide themselves.
Bundling has efficiencies. Unbundling every service into a separately priced item raises transaction costs and produces choice complexity that's its own consumer harm — the comparison burden our search analysis describes.
Weighing these honestly: the access argument is genuinely strong and the others are weaker than usually presented. Which is why our conclusion is about visibility rather than abolition. A cross-subsidy that keeps households banked is defensible. It would be more defensible if anyone could see it and argue about the amount.
The strongest objections
"Customers can avoid the fees, so it's a choice." Partly. Overdraft is avoidable by maintaining a buffer, and maintaining a buffer requires having one. The framing treats the constraint as a preference, and our illiquidity analysis suggests that's frequently wrong even for households with substantial net worth. Where the fee is genuinely a choice — deliberate short-term borrowing — the objection holds and the transfer is better characterized as a price than a subsidy.
"The transfer isn't as large as implied." A legitimate empirical challenge. We've deliberately avoided quantifying the aggregate, because credible estimates vary enormously with assumptions about pass-through rates, cost allocation, and what counts as the counterfactual price. The direction is well supported; the magnitude is genuinely contested, and anyone quoting a precise figure for the total transfer is quoting an assumption.
"This is true of every industry." Substantially yes, and the response is the three features in the opening: scale, invisibility, and consistent direction. The distinguishing feature isn't that cross-subsidy exists but that here it's hidden and it runs one way. A supermarket cross-subsidising loss leaders isn't systematically transferring from poorer customers to richer ones, and shoppers can see the prices.
What follows
What we'd take from the analysis, stated as positions rather than as neutral options:
Transparency before abolition. The highest-return intervention isn't banning cross-subsidies — it's making them visible. A customer who knows their free account is funded by other customers' overdrafts, and a customer who knows their overdraft funds free accounts, are both better placed to act and to complain than either is now.
Expect repricing and say so. Any reform removing subsidising revenue will reprice the subsidised product. Reforms should be argued for including that consequence rather than surprised by it.
Watch what replaces it. Where free checking becomes fee-based, the households priced out don't stop needing payments — they move to the products our selection analysis describes, which are frequently worse. The relevant comparison is against the substitute, not against the ideal.
Note where the structure is absent. A product priced at its cost of service to each user has no cross-subsidy and no hidden transfer, and it's a legitimate design choice with a real cost — some users pay more than they do today. That tradeoff should be made deliberately by anyone building consumer financial products, including us.
The observation to end on is uncomfortable and, we think, correct. The word "free" in consumer finance almost always means "paid for by someone with less money than you." That isn't a conspiracy and nobody arranged it. It's what happens when a bundled product's revenue comes from constraint and inattention, and neither the payers nor the beneficiaries can see the arrangement they're in.
Frequently asked questions
Primarily a concentrated minority incurring overdraft and related fees, plus the spread earned on balances. Those accounts tend to carry lower balances, so customers with buffers receive a service funded by customers without them.
Largely through merchant acceptance costs embedded in posted prices paid by all customers. Cash and basic debit users pay the same price and receive nothing, so the transfer runs toward premium cardholders.
In the main examples yes, though not by design — the fee-generating behaviours require thin buffers while the fee-avoiding ones require resources.
The subsidised product reprices. After debit interchange was capped, research found reduced free checking availability and higher fees and balance requirements at affected banks.
Key takeaways
- Free financial products are funded by concentrated minorities, and neither the payers nor the beneficiaries can see the arrangement.
- Free checking is funded substantially by overdraft fees concentrated in low-balance accounts — the customer with a buffer is subsidised by the one without.
- Rewards are funded through acceptance costs embedded in one posted price, so cash and basic-debit customers contribute and receive nothing.
- The direction is regressive because fee-generating behaviours require thin buffers while fee-avoiding ones require resources.
- The debit interchange cap confirmed the mechanism — removing the subsidising revenue repriced the subsidised product.
- The strongest defence is access: a regressive transfer that keeps households banked may beat an equitable price that doesn't.
This report presents an analytical framework and the authors' interpretation; it is not financial or policy advice. The direction of the transfers described is well supported in the literature while their aggregate magnitude is contested and depends heavily on assumptions about cost allocation and price pass-through; no aggregate figure is asserted here.