For example, banks might choose to allocate indirect costs uniformly across accounts. This has the benefit of simplicity, but it establishes a minimum threshold for how much revenue is required for a customer to be profitable. Is this why the industry insists their lower-balance accounts aren’t profitable absent the imposition of significant fees? That’s a claim bankers will sometimes make explicitly, but more often is implicit in the assertion that if they are forced to reduce overdraft fees, they will begin imposing higher monthly service charges on any customers who can’t maintain a sufficient balance. The Open Banker piece lays out a stylized version of these dynamics, showing the burden from a uniform allocation of indirect expenses makes lower-balance accounts money-losing absent higher fees (Table 1). In contrast, allocating indirect expenses pro-rata with net-interest revenues (and therefore with deposit balances) would allow lower-balance accounts to break even without these higher fees. The implication is that the tradeoff between overdraft and other fees may be a false choice driven by how banks allocate expenses.

Table 1: Assumed Checking Account P&L by Balance ($ per annum)

Source: FDIC Call Reports (Schedule RI), Federal Reserve Survey of Consumer Finances (2022), CFPB Data Point; Checking Account Overdraft (2014), Pulse 2020 Debit Issuer Study, StrategyCorps: The Profitability of the Average Checking Account and author’s calculations

Stepping back, the piece also challenges the premise that profitability targets set for the corporation as a whole should be applied at the level of individual customers. Given the large taxpayer subsidy banks receive in connection with their deposit business, they shouldn’t ration access to the banking system by pointing to arbitrary profitability thresholds and subjective methods for allocating expenses. They could instead acknowledge that their charters are a privilege and come with responsibilities, among them that they should meet the needs of their communities, even if doing so isn’t always profitable (i.e., they have a duty-to-serve). They could substantially reduce overdraft fees without reducing access to overdraft credit or raising minimum balance fees, providing their youngest and least affluent customers with a flexible and affordable bank account. If this results in a loss under their preferred method for allocating expenses, they can cross-subsidize those losses with revenues from elsewhere in their business.

As noted by the team at Better Markets, the report reads as a case of motivated reasoning. The gist of CEA’s argument is that what happens to the stablecoin market will be of little consequence for banks. The not-so-subtle implication is that policymakers should therefore feel free to give the crypto industry whatever it wants. The report offers what appears to be a free lunch, claiming stablecoins will help the financial system absorb a growing supply of U.S. government debt while having limited impact on household or business access to credit. The White House was happy to echo this framing, with Mr. Witt emphasizing how “sharp-minded” PhD economists at CEA had studied the matter and concluded any funds diverted to stablecoins would nevertheless find their way back to the banking system.

The report presents a good deal of math, but CEA’s analysis is hobbled by flawed premises. They lay out a theoretical model for how prohibiting yield on stablecoins will affect bank lending and therefore overall societal welfare. It assumes a tight quantitative relationship between deposits and lending. This lets them claim, when they show that stablecoin flows are returned to banks as “deposits”, that the impact on lending will be negligible. However, no such tight relationship can be found in the data. At the same time, the report underplays the qualitative value of transaction deposits in providing banks with a base of stable funding for long-term, fixed rate and difficult-to-sell loans. Deposits aren’t homogenous, and the report ignores a spectrum that runs from “sticky” transaction accounts (that depositors treat as “money”) to institutional liquidity pools that are little different from short-term wholesale borrowings.

Ignoring this disparity is what lets CEA argue stablecoins are a free lunch: if all bank funding is the same, and the flow of funds is conserved within the financial system, then allowing transaction deposits to shift into stablecoins isn’t a big deal. In his interview, Mr. Witt portrays the report’s depiction of a recirculating system as a profound insight analogous to discovering the conservation laws in physics. However, it’s not really all that insightful to show that debits and credits balance: they must, that’s the whole point of the double-entry accounting when banks create deposits by extending credit. What matters is how changes in the funding mix will impact bankers’ confidence when making loans. The report is an exercise in sophistry that relies on complex math, academic jargon and exaggerated precision to distract from the truth: there is no free lunch.

Bottom line, acceding to the crypto industry’s wishes will plant the seed for a long-term threat to a hybrid banking model that ensures financial stability. Promoting stablecoin adoption will eventually drain stable transaction deposits from the banking system and imperil the maturity transformation that underpins lending to households and small/medium-sized businesses, thereby supporting demand and employment. Stablecoins conflate a settlement technology (distributed ledgers) with a structured finance vehicle (the coin itself) whose only real purpose is to circumvent the banking system. Any efficiency benefits from the technology can be adapted to traditional deposit accounts without disrupting maturity transformation. Congress shouldn’t capitulate to lobbying pressure and allow banks to be displaced by an industry that lacks the same regulatory protections and does nothing to channel capital to the real economy.

The Link Between Deposits and Lending is More Tenuous than CEA Suggests

The key premise of the report’s economic model is that the supply of bank lending is constrained by available deposit funding after accounting for banks’ effective reserve requirements. However, modern scholars and policymakers doubt this kind of “fractional reserve” or “money multiplier” framework is a realistic description of how banks manage their balance sheets or the way credit flows through the economy.1 In practice, banks satisfy 100% of the loan demand they deem profitable. If core deposit funding isn’t sufficient to the task, they will borrow the remainder in the wholesale markets (debt will then be repaid if core deposit growth subsequently outstrips loan demand). The historical data highlight this dynamic. Over the last fifteen years, bank lending has been fairly stable relative to the economy despite a rising supply of deposits (Chart 1). This is true even after adjusting for inflows associated with the post-crisis and COVID-era accumulation of safe assets (dotted line). As a result, banks have reduced their reliance on large time deposits and wholesale borrowings. The model presented in the CEA report portrays an ironclad relationship that doesn’t comport with the data.

Chart 1: Bank Credit, Core Deposits and Wholesale Funding (Quarterly % of GDP)

Source: Federal Reserve H.8 via FRED and author’s calculations. Data is for “All Commercial Banks. Core Deposits is Total Deposits less Large Time Deposits. “Adjusted Core Deposits” deducts “Cash Assets” (mostly reserves held at the Fed) and Treasury and Agency Securities to reflect the portion of deposits available to support lending.

This is not to say deposits aren’t critically important for a bank’s lending business. In addition to being an independent source of profitability (earnings that should accrue to taxpayers and therefore represent a subsidy for banks), government-backed deposits provide useful ballast, ensuring the bank has a reliable source of long-duration funding for the majority of assets held on balance sheet. They effectively provide the bank with insurance, reducing the liquidity and interest rate risk that would otherwise be connected with lending.

While this support is valuable for banks and their borrowers, it’s an exaggeration to claim deposits mechanically drive the supply of credit. What matters is that the available supply of stable deposits remains sufficient to reassure banks and their other creditors. For the relatively modest (i.e., <10% of total deposits) stablecoin volumes contemplated in the report, it isn’t clear whether there would be any visible impact on lending activity.2 However, the chances of a material impact rise sharply if stablecoins enjoy “catastrophic success” (e.g., rising to >25% of total deposits) and begin supplanting bank money as a medium of exchange. Once the loss of core deposit funding passes a critical threshold, banks will move to boost loan spreads, reduce maturities and curtail their less liquid exposures. It won’t matter if investors recycle the same dollars back into banks. That funding simply isn’t reliable enough to support long-term lending.

Mapping Deposit Flows: Quantity vs. Quality

The CEA seeks to finesse the narrower question of how stablecoins will impact banks’ access to deposit funding. Their core insight is that the money market is a closed system: every asset is owned by someone, and when it is purchased, the seller is left with proceeds they must deploy. So, while allowing stablecoins to pay a yield may trigger additional withdrawals from deposit accounts, those funds will find their way back to banks. For example, if depositors withdraw funds from a bank to purchase stablecoins, the issuer will invest those funds in qualifying reserve assets, and the previous owners of those assets will need to reinvest the proceeds from the sale. In the absence of any unowned assets magically lying around, those funds must ultimately reenter the banking system.3 This reasoning is entirely correct as far as it goes (and getting at least one thing right clears the very low bar set by their farcical report on the CFPB).4 However, it’s a gross oversimplification. Not all forms of bank funding are perfect substitutes and it’s misleading to treat them as such.

The report considers three scenarios for how stablecoin flows might play out. In the first, issuers receive $1 withdrawn from bank deposits and use those funds to buy reserve assets (e.g., USTs) from an investor. They assume the proceeds from this purchase will then be redeposited at another bank, but the result would be the same even if the funds circulated through a series of trades among investors. Again, the key point is that all assets are already owned. When the stablecoin issuer steps in as an incremental buyer, it leaves the investment community with excess proceeds they need to leave on deposit with the banking system. This closes the loop created by the original deposit withdrawal. Hence the purported free lunch: stablecoin issuance leaves bank “deposits” unchanged, but total payments instruments (including stablecoins) and total securities owned by financial intermediaries increase (Table 1).

Table 1: CEA “Scenario 1”: Stablecoin Inflows Used to Buy Securities

Source: Author’s calculations.

The CEA’s second scenario depicts a case where the stablecoin issuer doesn’t purchase securities but instead holds bank deposits as reserves. The report notes that aggregate deposits would again be unchanged, but the 100% assumed LCR outflow factor on short-term liabilities to financial institution would increase effective reserve requirements. While the report wrongly claims this “restricts [the bank’s] ability to lend against this deposit” (as noted above, this isn’t how banks manage their lending businesses), it would nevertheless impact the bank’s regulatory liquidity position. If the Liquidity Coverage Ratio (“LCR”) or other measures fell below targets, the bank would need to buy additional high-quality liquid assets (“HQLA”). All else equal, these purchases engender additional deposit inflows and are therefore self-financing (again, it’s a closed system, and sellers will ultimately direct their proceeds back to banks). However, those incremental balances would have their own assumed outflows, necessitating additional HQLA purchases. Unless the bank short-circuited the process by issuing term debt, the process would iterate until the bank reached its LCR target (Table 2).

Table 2: CEA “Scenario 2”: Stablecoin Inflows Deposited at Banks

Source: Author’s calculations. Assumes an LCR outflow factor of 100% for deposits from stablecoin issuers. This necessitates the purchase of additional HQLA which engenders additional deposit inflows. Assuming an average outflow factor of 50% on those deposits, incremental HQLA purchases are 2X the deposit inflow from stablecoin issuers.

While regulatory liquidity requirements are an important consideration, banks will also make their own risk assessment. If holders use stablecoins the same way they previously used bank accounts for making payments, there’s no reason to believe potential outflows will be any higher under normal conditions. However, the lack of formal deposit insurance or an explicit lender of last resort might lead stablecoin holders to run more quickly than they would from traditional deposits at times of stress. In addition, stablecoin issuers may be more prone than retail depositors to preemptively withdraw deposits if they have concerns about their bank partner. Banks might therefore assume stablecoin issuers would accept lagged changes in deposit rates but hold additional liquidity reserves (directionally consistent with regulatory requirements) against the elevated risk of stress case outflows.

With respect to pricing, it’s not clear why banks should pay more for these wholesale deposits than they do for core deposits from the underlying household and business customers. Indeed, unlike institutional clients who can offer the promise of future business in exchange for a higher yield on their deposits, stablecoin issuers will become competitors as they grow. For issuers, holding deposits at banks is therefore likely to be a money-losing proposition, more so if banks act strategically to throttle a nascent competitor. Similar dynamics may play out in the repo market. On the one hand, repo should offer stablecoin issuers better rates than deposits. On the other, issuers may find themselves at the back of the line for collateral and pricing when dealing with bank counterparties. Given these potential downsides, issuers will likely want to minimize their reliance on the banking industry and the leverage it provides incumbents. Doing so may require help from policymakers to ensure reserve assets unconnected to banks are available at the necessary scale.

How Will Policymakers Manage Stablecoin Demand?

In the CEA’s third scenario, they assume issuers hold money market fund shares as reserves. If money funds use these inflows to purchase Treasury Bills in the open market, the flow of funds will resemble the first scenario, with excess investor liquidity winding up back in the banking system. However, relative scarcity in the Bill market may lead money funds to instead lend those funds against longer-dated USTs through the Fed’s Reverse Repurchase (“RRP”) facility. What happens next depends on the Fed’s balance sheet strategy. If it’s unwilling to expand its portfolio, it could sell other securities and extinguish a portion of the reserves held by banks (Table 3). The report somewhat wryly notes “this leakage is a feature of the broader non-bank financial system and is not unique to stablecoins.”5 That elides the extent to which a growing stablecoin market would put strain on the Fed’s desire to deliver “ample” reserves (as part of the floor system for setting policy rates) while also maintaining as small a balance sheet as possible.

Table 3: CEA “Scenario 3”: Stablecoin Inflows Placed at the Fed as RRPs

Source: Author’s calculations.

Could rapid growth in the stablecoin market and related demand for RRP capacity persuade the Fed to abandon its commitment to ample reserves and return to a corridor system based on reserve scarcity? The report treats the floor system as a base case, but now that Miran is at the Fed, he has mused about the possibility of reverting to the old system. Under a new Chair, the Fed might decide the monetary policy benefits of the floor system can be sacrificed to meet the needs of a politically influential new industry. If the Fed abandoned the floor system, it could accommodate as much as $3Tn of stablecoin balances invested in RRPs without expanding its balance sheet above current levels. Of course, banks would lose the contingent liquidity these reserves ostensibly supply, so the Fed would need to be find a way to lend to banks on an emergency basis without the stigma associated with use of the Discount Window. Liquidity regulations would also need to be amended to explicitly recognize that borrowing capacity.

What else could policymakers do to accommodate growing stablecoin issuance? One step would be to better align the supply of short-dated reserve assets with issuer demand. When the Fed borrows against longer-dated USTs through the RRP facility, it reduces the government’s effective debt maturity profile. Treasury could acknowledge this reality and eliminate the Fed’s intermediary role by shifting its issuance strategy to increase the availability of Bills. Secretary Bessent and Fed Governor (and former CEA Chair) Stephen Miran have both emphasized the potential for stablecoin demand to directly absorb a rising supply of Treasury issuance. While the proceeds from Bill purchases would still find their way to bank balance sheets (again, it’s a closed system), those funds would likely be indistinguishable from unsecured wholesale borrowings. For example, while “jumbo” CDs are classified as “deposits”, they aren’t used a medium of exchange and therefore price at levels comparable to unsecured bank debt.

Conclusion

For the moment, the threat to banks remains mostly hypothetical. Stablecoins will not reach a scale where they pose a meaningful threat unless a broad cross-section of businesses and the public see them as a practical alternative for settling purchases of goods and services. This seems highly unlikely given banks’ incumbency advantages (low cost, universal vendor acceptance and guaranteed access to funds) and the risks associated with using stablecoins. Of course, the crypto industry could persuade policymakers to put a thumb on the scale, “priming the pump” by signaling to stablecoin holders that their funds will remain fully accessible in a crisis, asking large payees to accept stablecoins as payment and pushing large payors (including government agencies) to offer payment in stablecoins as an alternative to bank transfers. Absent these interventions, stablecoins are likely to remain a specialized form of payment in the crypto trading market, with some potential additional use for cross-border consumer payments and other transactions where banks choose not to compete.

In the scenario where policymakers succeed in pushing stablecoin adoption, the loss of core deposit funding would eventually reach a tipping point where it threatens the supply of credit. Taken to its logical extreme, the result would be a shift to a narrow banking model. Stablecoin issuers (i.e., the narrow banks) would intermediate payments flows and hold safe assets while legacy lenders would become entirely wholesale funded. If policymakers wanted to limit the impact on credit supply (including the possibility of an outright funding squeeze and related credit crunch), they would need to offer lenders guaranteed term funding. This would eliminate much of the reputed benefit from shifting to a narrow bank model and fully absorb the additional Treasury issuance capacity Bessent and Miran have been hyping.

Some readers might wonder what’s wrong with a little competition and react negatively to the idea that banks might attempt to suppress new entrants. While this is understandable, it misinterprets banks’ role in our hybrid banking system. Banks may be privately owned corporations seeking to maximize profits, but they are also executing core monetary functions as franchisees on behalf of the U.S. Government. It would be an act of self-sabotage on the government’s part to promote “competition” (free riding?) from less regulated entities that don’t perform the same role in supporting financial stability and credit formation. If policymakers aren’t satisfied with how banks are performing, the first response should be to more vigorously enforce the requirements of their charters. Competition shouldn’t come at the expense of endangering the public goods a bank charter creates.

This “quasi-sovereign” role is most clear with respect to the money supply. Very little consists of physical currency directly issued by the Federal Reserve. Most is held as bank deposits, with payments then reflected as transfers among depositors. The guarantees extended as part of the hybrid system were intended to eliminate frictions in the payments system by convincing the public to treat their deposits as “money.”3 However, an unintended consequence was that it drove a spread between the resulting convenience yield deposits enjoy and banks’ alternative cost of debt. My previous article (Quantifying the Taxpayer Subsidy on Deposits) estimated that banks earn as much as $150Bn per annum (i.e., roughly 50% of industry earnings) from this spread, even after accounting for the cost of running the business. It’s one thing to know banks have access to a government backstop if there is a crisis, but quite another to realize it drives so much of their earnings. A modern fiat currency system is ultimately a covenant between the people and their government. The public shouldn’t be expected to pay for a service they themselves are providing.

Proposed Alternatives

This exploitation of a public resource by private businesses has prompted calls for root and branch reform to eliminate free riding. For example, some argue for shifting to a “narrow banking” model where deposits can only be used to fund safe assets (e.g., Treasury securities or central bank reserves).4 For these scholars, the goal is to eliminate the run risk inherent in using transaction deposits to fund long-term loan portfolios and limit banks’ ability to stretch their deposit guarantee to cover high-risk lending or capital markets activities.5 In contrast, Saule Omarova has argued for abandoning the hybrid model altogether, shifting deposit-taking to a “public bank” model. Her focus is less on limiting systemic risk and more on “democratizing” access to the banking system by lowering costs and improving access. In one specific variant, she suggested that the Federal Reserve offer deposit accounts directly to the public.

What narrow bank advocates neglect is the critical role government-backed deposits play in maintaining a reliable supply of credit. Because they are perceived as money, they can offer depositors the assurance of immediate access to cash while simultaneously providing banks with a stable and low-cost source of funding. This “maturity transformation” means banks can offer borrowers long-term fixed-rate loans with little concern as to interest rate and liquidity risk. It also lets them hold small-balance and other illiquid loans that couldn’t be readily financed in the capital markets. The benefit of deposit funding is especially important at times of stress when lenders’ access to wholesale funding may be in doubt. If banks didn’t have access to stable deposits, they would require some other form of guaranteed funding. Otherwise, the resulting credit crunch would throttle consumer spending, investment and employment. In other words, a narrow banking model doesn’t reduce taxpayer exposure; it just rearranges the ledger items.

Narrow banking proposals are often unclear about how they will address the subsidy issue. Some just assume competition will lead banks to pass through the after-expense earnings from their reserve assets to depositors. However, the convenience yield of deposits generates a pricing wedge relative to securities, one that is potentially larger than the cost of operating the bank. Paying depositors more than the market demands will drive substitution into deposits from other assets, inflating narrow bank balance sheets more than is necessary in a competitive market. Potential distortions can be avoided if banks only pay the convenience yield demanded by depositors. In a public bank model, this won’t result in a subsidy as any profits are returned to taxpayers. However, in a privately-owned model, eliminating subsidies would require some mechanism to capture those profits.6

An alternative approach from Lev Menand and Morgan Ricks addresses both these concerns. Their proposed “New National Banking System” leaves the deposit-taking function with banks but requires them to be rechartered as public utilities. It echoes narrow bank proposals in seeking to limit the use of deposits to support risky assets. An explicit goal of their proposal is to undo the post-1980 shift toward deregulation and restore the clarity of the New Deal regulatory framework. Utilities wouldn’t be allowed to own stocks or commodities, and the proposal anticipates that securities trading would be conducted by nonbank financial firms. These nonbank firms would be prohibited from issuing short-term, runnable debt and would instead rely entirely on long-term debt and equity raised from investors. Utilities could engage in permissible forms of lending (i.e., those promoting “productive ends”, including meeting the needs of low- and moderate-income borrowers in their communities) that would continue to benefit from a stable source of long-term funding. They address the subsidy issue directly by requiring any spread between the all-in cost of deposits and the alternative cost of debt to be conveyed to Treasury as a “franchise royalty.”

Menand and Ricks’ proposal is an important addition to the public debate that offers an elegant solution to the subsidy problem. Unfortunately, their desire to force a clear separation between the utilities and riskier activities is unworkable given current leverage levels in the economy. Form follows function, and the legal framework needs to reflect economic realities. In the immediate postwar period, the stock of household and noncorporate business debt was roughly at parity with bank deposits (Chart 1). Given the balance between less liquid loans and stable funding, it was possible to erect a high fence around the banking system while letting corporations and investors fend for themselves in the SEC-regulated capital markets. However, over the last fifty years, the supply of these illiquid loans has grown well beyond the capacity of the banking system and is now funded in the capital markets by nonbank intermediaries.

Chart 1: Bank Deposits vs. Household and Noncorporate Business Debt (% of GDP)

Source: Federal Reserve Z.1 and author’s calculations. Balances includes data for commercial banks, savings institutions and credit unions. Deposits are shown net of reserves as deposit inflows that result from the Fed’s creation of reserves are not available to fund lending.

While large portions of the long-term funding for this risk (e.g., GSE MBS and student loans held by the Department of Education) are protected by separate government guarantees, origination, aggregation and distribution activities are not. The legal and regulatory changes Menand and Ricks criticize did not cause the increase in financial instability we’ve seen since the ‘80s. Higher debt levels were driven mainly by changes in tax policy and the related rise in asset values. Policymakers weakened New Deal rules in a haphazard effort to accommodate a new environment where too much illiquid debt was chasing too few sources of stable funding. It is this imbalance that is the underlying cause of the liquidity shortfalls that drive systemic risk. If policymakers had sought to maintain the New Deal framework in the face of rapidly rising debt, the result would have been even greater instability (i.e., the cause of increased nonbank intermediation was “supply push”, not “demand pull”).

We are where we are, and it would be dangerous to narrow the regulatory perimeter as Menand and Ricks suggest when debt levels are so high. Restoring the clarity and simplicity of the New Deal structure is impossible unless we also reverse the post-1980 shift of resources from labor to capital that has inflated equity values and increased leverage (there is a link between “financialization” and inequality, but the causation flows in the opposite direction from how it’s usually depicted). This is highly unlikely absent a major rupture in the political system, one larger than just the normal rotation in the partisan balance. Even if such an effort were to succeed, it would only halt the accumulation of new debt. Unwinding existing balances would take decades given the long-term nature of most loans.

In the interim, nonbank intermediaries (e.g., the GSEs, nonbank mortgage lenders, automakers, independent broker dealers, etc.) depend on a flexible supply of short-term liquidity from insured banks to manage volatile origination volumes, hedge interest rate exposures and meet investor demands for secondary market liquidity. Cutting them off from bank liquidity would make their business models unworkable, disrupting the flow of credit that supports aggregate demand. As such, it’s a nonstarter politically: the revealed preference through multiple crises is that the safety net extends beyond the payments system to include supporting lending relied on by a critical mass of voters. Policymakers should publicly acknowledge this reality and design around it.

In theory, the perimeter could be left untouched, and policy could instead focus on eliminating deposit subsidies by imposing a franchise royalty. However, even this aspect of Menand and Ricks’ proposal would be controversial. The industry will decry what they would label as a devastating “tax”, dispute the details of whatever methodology is used to calculate assessments, and deploy armies of lobbyists and lawyers to block the move in Congress and/or tie it up in the courts.7 Banks might also argue that if they “paid in full” for the public goods they employ, they can abjure any requirement to serve the public interest. In short, waging a protracted political battle over recapturing subsidies puts the principle of a public service obligation at risk for an amount of money that, while certainly meaningful, is not decisive for the country’s fiscal trajectory. The status quo is unfair, and eliminating subsidies by charging banks outright may eventually be necessary, but the complications argue for pursuing an alternative approach.

A Duty-to-Serve Mandate is a More Realistic Alternative

Given the challenges connected with trying to eliminate deposit subsidies altogether, policymakers should focus instead on ensuring they are used to improve outcomes for the public. This would start with insisting on candor: given how much of banks’ earnings derive from taxpayer support, complaints about the costs of necessary regulation should be mocked loudly and often. Banks’ exclusive access to guaranteed liquidity from taxpayers is comparable to the way railroads, telephone companies and other “common carriers” benefit from access to various scarce public goods (e.g., radio spectrum).8 If banks want to profit from this access, their products should be available to the public at the “just, reasonable, and affordable rates” required of other common carriers. This can be best assured by subjecting banks to an explicit “duty-to-serve” mandate that ensures affordable access to bank accounts and other services. It would be akin to the one already applicable to the GSEs that ensures mortgage credit is available to support the purchase of manufactured housing, preserve affordable housing, and bolster lending in rural markets.

At the moment, access to the banking system suffers from significant gaps. Branch closures in lower-income neighborhoods have fed concerns about “banking deserts.” While the vast majority of Americans have a bank account, a significant percentage of lower-income households do not. According to the FDIC, in 2023 approximately 14% of households with annual incomes below $30,000 did not have a checking or savings account. These 3.3MM households represented the majority of the 5.6MM total that were “unbanked”. The reasons offered for not having an account often revolve around costs, with a third claiming the level or unpredictability of fees as the main cause. Overdraft/NSF fees have fallen in response to negative publicity and regulatory pressure, but in 2025 consumers still paid over $12Bn in aggregate deposit fees.9 Given how unpopular fees are, why are they such a mainstay of banks’ pricing structures? When challenged, defenders of the status quo claim fees are necessary to maintain a sufficient level of profitability, particularly for lower-balance customers who don’t produce meaningful net-interest revenue. This is a flimsy excuse, one that policymakers should reject.10

Instead, banks should be required to maintain an acceptable level of branch density and offer low-cost transaction accounts to all applicants residing within their footprints as a condition of their charter (this is an element of the Menand and Ricks proposal as well). The structure of these accounts should account for the struggles faced by households living paycheck to paycheck. Families with modest savings and inconsistent expenses will occasionally run short of cash, so even the most basic accounts should offer overdraft credit at reasonable interest rates. The industry has publicized its willingness to offer low-cost checking accounts (e.g., the “Bank On” program). However, despite offering relatively modest fees (e.g., a maximum $5 per month with no minimum balance), consumer uptake in this product has been limited. Only 23.2mm accounts have ever been opened, and the 14.3MM accounts open at the end of 2024 represented only $18.2Bn in deposits. Even if low-balance accounts represent only 5% of transaction balances, that would still indicate an addressable market of $300Bn that this specialized offering has failed to penetrate. Given the adverse consequences of late payment on rent or utility bills, depositors understandably place a premium on access to cash liquidity. Banks shouldn’t take advantage of that preference to extract excessive fees.

The implementation of a duty-to-serve mandate for deposit accounts should be accompanied by more vigorous enforcement of the Community Reinvestment Act (“CRA”) and other rules with respect to lending. Banks shouldn’t neglect credit-worthy borrowers who are seeking small-dollar unsecured credit, mortgages and other loans. Nor should they cross-subsidize their wealthy and corporate customers from excessive interest rates and fees imposed on vulnerable borrowers who lack viable alternatives. Trump’s proposal for a temporary 10% cap on credit card drew renewed attention to the excessive spreads card lenders earn from their subprime customers. If, for example, rates were capped at 18% in today’s rate environment, a reasonable estimate is that it would cost the industry no more than $14Bn per annum of net-interest revenue.11 Separately, when the CFPB presented its final rule on credit card late fees, it noted that those fees were approximately $14Bn in 2022. The implication is that the industry could cap credit card interest rates, waive deposit service charges and eliminate credit card late fees, all while suffering a combined revenue loss of less than $50Bn per annum. This is well below the estimated $200Bn pretax subsidy they receive on deposits. In practice, many fees apply universally, so it would probably cost banks far less than $50Bn to make their products affordable for their most vulnerable customers.

How Should a Duty-to-Serve Mandate be Implemented?

Ideally, a robust duty-to-serve mandate would be implemented through explicit new legislation. However, given the lack of bipartisan consensus when it comes to helping the disadvantaged, this would likely require an end to the filibuster or a 60-seat Senate Democratic majority. Of course, policy often evolves more elliptically, with major legislation codifying already existing regulatory standards. For example, when the FDIC Improvement Act of 1991 codified standards for “safety and soundness”, it built on decades of regulatory work to define practices that would violate an earlier statute's prohibition on “unsafe and unsound” practices. A similar dynamic could play out with respect to a duty-to-serve. The National Bank Act’s requirement that the OCC assure “fair access” to financial services, coupled with the Community Reinvestment Act’s finding that banks must demonstrate their “deposit facilities serve the convenience and needs of the communities”, signal a clear intent that chartered banks have a duty to meet the needs of the general public. Regulations that further elaborated that standard could serve as the basis for future confirmatory legislation.

Unfortunately, any effort to impose these regulations without industry buy-in will face immediate backlash and legal challenges that are likely to succeed given the current makeup of the federal courts. However, this need not be the end of the story. The post-Trump political environment will be volatile, and a more populist future Congress might consider more sweeping policies, including imposing a $100Bn+ deposit franchise royalty of the sort recommended by Menand and Ricks. Faced with that prospect, industry leaders may see the wisdom of engaging constructively, embracing a duty-to-serve mandate to stave off potentially more costly requirements. Both sides would avoid a protracted legal battle, industry would protect most of their legacy privileges, and policymakers will have established the principle that those privileges are earned through a more tangible commitment to serving the public.

Conclusion

We shouldn’t throw away the elegance of a hybrid banking model that underpins household and small/middle-market business lending, thereby sustaining the demand that supports employment. Maturity transformation conjures a public good out of thin air - this is a good thing! However, it gives rise to a subsidy, one that is difficult to quantify with precision but is large and benefits bank shareholders at the expense of taxpayers. A pragmatic strategy would avoid protracted fights about specific dollar amounts and instead prioritize the principle that universal service is a condition of a bank charter. Banks don’t have a viable business model absent taxpayer support, and this reality should be acknowledged through explicit duty-to-serve requirements. Indeed, simply having access to stable funding already gives banks a competitive advantage against nonbank lenders. Shareholders’ property rights shouldn’t be assumed to extend to that stable funding being available at a subsidized cost. A duty-to-serve mandate would force them to use a portion of the subsidy to provide universal, affordable access to bank accounts and lend money to creditworthy borrowers at a reasonable cost. It wouldn’t require banks to make any changes to their business models, other than to acknowledge the benefits and obligations that accrue from their charters. If they balk and use a sympathetic court system to resist change, Congress could always choose to impose a fairer system through legislation.

Background

Banks offer a range of deposit products with diverse features and target audiences. Certificates of deposit (“CDs”) provide a guaranteed interest rate for a fixed term and appeal to customers with discretionary savings seeking a return. In contrast, checking and (misleadingly named in most cases) “savings” accounts are fully available on demand and used to make payments. Balances in these “transaction” accounts fluctuate as they represent a residual left over from a much larger volume of outlays and receipts the depositor can’t fully predict nor control. Interest rates paid on transaction balances are typically near zero for small-balance accounts held at traditional “brick and mortar” institutions. Rates can be higher for customers who have just opened their accounts or for households and businesses that hold larger balances. However, the average rates paid across all depositors are well below those available in the fixed income market.

Savings accounts offered by online banks offer a very different value proposition, one more intuitively consistent with the product’s name. The rates they offer for even small-balance accounts rival those available from money market mutual funds. However, they typically lack a linked checking account or ATM access, meaning withdrawals only become available to spend once funds have been transferred to another bank. This may impose a delay of up to three days for the transfer to clear.

That difference in timing is critical to the question of required interest rates. If a deposit is available to make immediate payment (i.e., with zero days delay or “T+0”), depositors consider them to be “money”, no different from the dollar bills in their wallets.1 The value of money as a medium of exchange means it doesn’t need to offer the same return as securities that might be considered equally safe but can’t be used to settle transactions (i.e., as noted above, it has a convenience yield). Physical cash obviously earns nothing, whereas deposits typically offer a non-zero yield. What this yield will be is a function of supply and demand. Traditional brick-and-mortar banks still possess an effective monopoly on the ability to offer T+0 accounts to households and small businesses. This owes in part to their extensive physical branch and ATM networks, but also to merchants’ willingness to accept bank liabilities (e.g., checks) for immediate payment. The privileged role regulated banks play in the system extends to the deposits large corporations and investors use to settle transactions. The rate paid on deposits is therefore a function of competition among traditional banks and is relatively insensitive to changes in the rates available on products that lack immediate cash liquidity.

Where should this competition settle out? It is costly to operate the branch network and other infrastructure necessary to offer deposit accounts, and these expenses (net of any fees collected) are part of a bank’s total cost of deposit funding. Banks have no incentive to pay more for deposits inclusive of these costs than the rate they would pay on debt raised in the capital markets. The question is whether competition will eliminate any excess profit banks can earn if deposits’ convenience yield leaves their all-in cost below the cost of debt.2 Theoretically, there is no need for deposits to make any profit at all. Not only do they not pose risks that might attract economic or regulatory capital, but they actually reduce the liquidity and interest rate risk connected with banks’ lending businesses. Any consistent profit would suggest banks are extracting a subsidy from their deposit franchise, one derived from the perception of deposits as money, a status that in turn depends on banks’ relationship with the government. Centuries ago, the sovereign’s ability to mint coins worth more than their value as metal was referred to as “seigniorage.” When modern scholars refer to seigniorage, they mean the spread earned from issuing money claims to buy other assets. In the modern U.S. financial system, banks capture much of this seigniorage benefit for themselves.

Simulation Analysis: JPMorgan Chase

Estimating the implicit subsidy on bank deposits is a complex exercise. For example, what is the right benchmark alternative cost of borrowing to which deposit rates should be compared? Transaction deposits have no defined maturity and are available “on demand.” However, in practice they tend to be “sticky” and large-scale withdrawals are rare absent a crisis. While the rates banks pay move through time, they can typically retain deposits without fully matching increases in market interest rates.3 The appropriate pricing benchmark is therefore neither an overnight floating-rate note nor a perpetual fixed rate bond.4 The executives managing the deposits business will work with the bank’s Asset-Liability Committee (“ALCO”) to set rates in a way that balances profitability, market share and risk. While we don’t have access to the full range of inputs they are using, we can infer the effective maturity they are managing against from how they respond to changes in market rates.

What is the bank trying to accomplish? A reasonable objective is to maintain a stable margin against its own asset portfolio. However, management understands competitors are doing the same thing and deposit pricing must remain competitive. So, a reasonable first approximation is that deposit rates will move in tandem with the yield of a “composite” portfolio that reflects the larger industry’s asset mix. Banks might also be collectively willing to take some degree of interest rate risk, for example seeking to profit from an upward sloping yield curve by targeting a shorter maturity on deposits relative to assets. The composite portfolio represents a competitive equilibrium where banks believe they can hold market share while managing their asset-liability profiles consistent with an acceptable risk tolerance. Every bank makes both fixed- and floating-rate loans while holding a portfolio of generally fixed-rate securities. It is therefore reasonable to expect the composite portfolio to include both fixed- and floating-rate components.

To get a sense of how these numbers lay out in practice, we can look at how Chase has priced its deposits. From an analytical perspective, Chase offers the advantage of having a large enough market share that it should be representative of the industry as a whole. It has also held that leading market share for over two decades, so its deposit rates can be evaluated through multiple interest rate cycles. Through trial and error, those rates can be compared to different configurations for the composite portfolio. The chart below depicts Chase’s average quarterly average rate paid on interest-bearing deposits over the last three decades and compares that rate to a benchmark rate that is 60% the 3-Month Treasury rate and 40% a trailing average of 5-Year Treasury rates (Chart 1).5 It also shows a proxy formula rate equal to the benchmark rate less a spread of 1.40% but subject to a floor rate of 0.05%.6 The change in this formula relative to a change in 3-Month rates (i.e., “beta” as conventionally quoted by market analysts) will be roughly the 60% 3-Month coefficient from the formula.7

Chart 1: Chase Cost of Interest-Bearing Deposits vs. Proxy Formula (Q1’95 – Q4’25)

Source: Company financials, Federal Reserve H.15, and author’s calculations. Trailing average calculation incorporates the most recent quarter spot rate based on the assumed maturity (e.g., 50% per quarter for a 6-Month, 25% for a 1-Year and 5% for a 5-Year)

The rationale for using a trailing average formula is straightforward. Banks price their loans relative to market rates, so maintaining a spread against the loan portfolio will lead deposit rates to track with market rates as well. While loans are originated in discreet “vintages”, with pricing reflecting the rate environment when each vintage was originated, changes in deposit rates for the most part apply uniformly to all depositors irrespective of when their accounts were opened.8 Deposit yields will therefore reflect the average yield of the asset portfolio, much of it bearing fixed rates set in earlier periods.9

Lest there be any confusion, it is not at all surprising that the proxy formula doesn’t provide a perfect fit vs. Chase’s reported deposit rates. After all, it is based on only two variables and relies entirely on aggregate public data. Indeed, it is rather remarkable that a two-factor model with a static fixed/floating mix and constant spread can reproduce Chase’s deposit rates as well as it does over a period of thirty years, a period during which Chase did several large mergers, grew its deposit base by a factor of 15, and saw the emergence of new competitors like online banks, all while weathering a serious recession, a global financial crisis and a once-in-a-century pandemic. The implication is that the Chase’s target asset mix and return objectives have been relatively stable. Over the twenty years I advised bank clients, I ran this same analysis on dozens of banks as part of our asset-liability and merger advisory efforts. Based on that experience, I am comfortable saying Chase is not an outlier: most banks displayed similar stability with respect to their deposit pricing.

All-In Cost of Deposit Funding

The simulation analysis was for interest-bearing deposits only and suggests Chase has been able to raise those deposits at a consistent spread of approximately 1.40% against a composite Treasury portfolio, unless rates approach the zero lower bound (“ZLB”).10 The bank also accepts noninterest-bearing deposits, and these have constituted between 20% and 30% of the total deposit base over the last three decades. Given the lack of any interest paid, they essentially act as a zero-coupon perpetual note and can substitute for very long-term borrowings. The analysis therefore assumes they earn a spread equal to a moving average of 10-Year Treasury rates. Raising roughly a quarter of deposit funding at a spread equal to the 10-Year rate historically boosted the effective spread on total deposits by more than a full percentage point (less so recently given the long-term decline in interest rates).11 It implies a prevailing pretax spread for the aggregate deposit portfolio over the last ten years of approximately 2.00%, 0.60% higher than the approximate 1.40% spread on interest-bearing deposits (Chart 2).

Chart 2: Effective Spreads Relative to Benchmark Formula (% of Average Deposits)

Source: Company financials, Federal Reserve H.8 and author’s calculations. Trailing average calculation incorporates the most recent quarter spot rate based on the assumed maturity (e.g., 50% per quarter for a 6-Month, 25% for a 1-Year and 5% for a 5-Year)

The other two drivers of deposit profitability are fees (e.g., related to consumer overdraft, minimum balance requirements and ATM usage, as well as those charged to business customers) and the personnel and other expenses associated with operating the business (e.g., the branch network, IT platforms, etc.). While deposit fees are disclosed publicly each quarter as part of banks’ regulatory disclosure, the cost of servicing deposit accounts isn’t isolated separately. The analysis therefore estimates an allocation based on the Company’s segment disclosure.12 It suggests only a third of Chase’s total Non-Interest Expenses are allocable to deposit gathering (Chart 3). For purposes of estimating the subsidy, this approach is obviously imprecise but is by no means aggressive. Allocating one third of expenses to deposit gathering leaves this function with a historical Efficiency Ratio (Non-Interest Expenses as a share of revenues) similar to the rest of Chase’s business lines. Increasing the expenses allocated to deposits (i.e., raising their all-in cost and lowering the implied subsidy) would imply this activity is less efficient than the rest of the business.

Chart 3: Fee and Expense Spread Contribution (% of Average Deposits)

Source: Company financials and author’s calculations

The Fed’s quantitative easing (“QE”) and ample reserves policies have structurally inflated deposit balances and makes comparisons to pre-crisis levels less relevant.13 At the new plateau the Fed’s balance sheet has reached over the last ten years, deposit fees have ranged between 0.2% and 0.4% of average deposits. Allocated expenses have been more variable, but this is not surprising given the crudeness of the allocation method. In recent years, these expenses have been between 1.0% and 1.5% of average deposits (for a sense of magnitudes, each additional ten percentage points of total non-interest expense allocated to the deposit business raises the allocated expense ratio by approximately 0.4%). Combining all these elements (effective spread on all deposits, plus fees, less allocated expenses) produces an all-in pretax margin from the deposit business (Chart 4). In addition to showing the sum of the input series, the chart also shows what the net-margin would be if deposit spreads were not compressed when market interest rates were at the ZLB.

Chart 4: All-In Pretax Net-Margin from Deposits (% of Average Deposits)

Over the last fifteen tears, all-in pretax net-margins have oscillated around 1.0% (e.g., 2.0% for the interest rate spread on total deposits, plus 0.3% for deposit fees, less 1.3% for expenses). As discussed above, it’s not clear why banks should collect any profit at all. Doing so is akin to charging people to stand under your neighbor’s umbrella!

Implications

If Chase’s estimated 1.0% pretax deposit profit margin is applicable across the entire industry, the subsidy it implies is enormous. While less than $11Tn of the industry’s $20Tn in deposits are fully insured, the analysis above measured the resulting cost advantage for the entire deposit base. A run-rate net profit margin of 1.0% implies the industry receives an approximate $200Bn pretax subsidy ($158Bn after-tax) from its role in administering the money supply (Chase alone would earn $25Bn pretax from its $2.5Tn of deposits). This represents just over 50% of the industry’s 2025 earnings. To put the relative importance of the subsidy versus other revenues in context, the industry’s 2025 overdraft fees were approximately $6Bn.

The subsidy available from deposits is vital for understanding how banks are valued by investors. Monoline finance companies (i.e., those whose only business is lending and who lack access to deposit funding) are typically priced at no better than 1x their Tangible Book Value (“TBV”). This is reasonable, as it is difficult for them to earn more than their cost of capital. In contrast, banks typically trade at a substantial TBV premium given the expectation that they will deliver premium returns. How do they do this when banking is a mature, highly regulated industry? Common explanations include how charter requirements protect incumbents from competition, the contribution from low-risk fee businesses that are ancillary to banking, and the value of the federal safety net in reducing downside risk. Scholars also highlight banks’ ability to pay depositors a below-market rate. However, these discussions of the “deposit franchise” typically focus on deposits’ modest rate sensitivity and the benefit this provides with respect to holding long-term assets in higher rate environments. The analysis laid out above explicitly calculates the value of deposits’ all-in cost advantage versus comparable maturity funding alternatives.

The industry may dispute that they receive any subsidy whatsoever. After all, they pay for their FDIC insurance coverage, coverage which in any case poses no risk to taxpayers given the mutual funding structure of the Deposit Insurance Fund (“DIF”). However, this confuses the issue. FDIC assessments are sized to ensure the DIF is sufficiently large to cover potential losses and avoid draws from Treasury. The credit risk posed by potential bank failures is entirely separate from the liquidity benefit banks receive from issuing money claims. The former is a potential direct cost to taxpayers if the mutual insurance structure fails, the latter is an opportunity the cost the government has failed to capture. Alternatively, banks might argue that any cost of funding benefit they receive from their deposit base is passed through to borrowers. This argument rests on somewhat firmer ground. Access to a deposit base certainly makes banks better able to lend through credit cycles. It is also true that finance companies rarely compete for banks’ prime credit borrowers, suggesting banks can offer a discount versus their wholesale-funded competitors. However, the fact that banks consistently trade at a valuation premium to these competitors implies much of the deposit subsidy is captured by shareholders. Finally, banks might argue some net profit is required to justify the management time and attention necessary to operate a deposit platform. This argument is unpersuasive given the extent to which deposits reduce the liquidity and interest rate risk posed by banks’ other businesses. One typically pays for insurance rather than be paid to enjoy the protection.

1. The first is a “deposit franchise channel” that represents the spread between what banks collect in interest on reserve balances (“IOR”) and what they pay for the marginal deposit funding associated with those balances. The report estimates the net present value of this channel at approximately $150Bn.

2. The second is a “maturity transformation channel” that relates to the Fed’s quantitative easing (“QE”) programs. While they don’t offer any specific estimates for this channel, they argue the returns from large-scale asset purchases (“LSAPs”) are typically negative for taxpayers and see most of this lost value accruing to banks.

Given these costs, H&Y believe policymakers should consider eliminating the payment of interest on reserves, contracting the Fed’s balance sheet, and returning to the pre-GFC “corridor system” for setting policy rates. If this is a step too far, they suggest policymakers contemplate alternative strategies that wouldn’t abandon the floor system entirely but would nevertheless reduce the interest paid to banks.

H&Y concede their argument is at odds with conventional wisdom among monetary economists that the level of reserves is “distributionally neutral” (i.e., does not provide a subsidy to banks). Indeed, the report was released in tandem with a pointed response from their Brookings colleagues Don Kohn and Bill English that challenges several of the report’s premises and conclusions. First, it questions whether expanded reserve balances are in fact funded with deposits. Second, it argues competition to fund zero-risk reserve assets should limit profits and notes there has been little change in industry returns following adoption of the floor system. Lastly, it takes issue with judging the costs of a QE program retrospectively and challenges the wisdom of gainsaying a market as sophisticated as the one for Treasuries. Given Kohn and English’s long careers in central banking, including being at the Fed when the floor system was first designed and implemented, it’s intriguing that more of a consensus wasn’t reached prior to the report’s publication and that debates among the fellows are being aired publicly.

Summary Observations

H&Y are correct in claiming banks extract a spread from the deposits backing their reserve assets. However, the spread is likely not as large as they claim. They reasonably assume near zero marginal operating expenses against the incremental deposit inflows funding reserves. Indeed, it is the ability to leverage fixed expenses over a larger deposit base that drives the profitability of those marginal deposits. The report then derives a spread for these deposits using an assumed “beta” (i.e., the change in deposit rates relative to the short-term market rate). Notably, how they use beta in the report differs from how it’s commonly employed by industry.1 Their analysis also excludes data from the “custodial” banks (e.g., BNY Mellon and State Street) that exclusively serve the large investors most likely to have sold securities to the Fed. These banks hold nearly $600Bn of investor deposits, so the rates they pay are highly relevant. Recent pricing from these banks suggests the interest spread between deposits and reserves may be as little as 0.75% per annum. In contrast, H&Y’s valuation of the deposit franchise channel implies an interest spread of approximately 1.15%. If spreads elsewhere in the industry are similar to those paid by the custodial banks, the value of these deposits is well below what is posited in the report.

Even if we stipulate banks are extracting a spread from holding reserves, it’s unclear why this value should be considered a loss to taxpayers. If the Fed’s reserve liabilities are invested entirely in Treasury Bills, there is negligible direct cost to taxpayers.2 The cost of any pricing “wedge” between market yields and the rate paid on deposits is incurred by the depositors. In other words, if there is redistribution related to the Fed’s ample reserves policy, it is within the private sector. However, even this may be overstating things. Investors set the price and yield at which they are willing to sell securities to the Fed. The marginal seller is doing so in the knowledge that the proceeds will be left on deposit with their clearing bank (i.e., they could refuse to sell until yields fell sufficiently to eliminate the spread). Rather than being a loss of value to the investor and windfall for the bank, the spread may simply represent a competitive equilibrium based on the convenience yield deposits enjoy relative to securities. This suggests the real value provided by taxpayers is less the reserve balances themselves, and more the assured safety and liquidity investors associate with deposits given banks’ access to a government backstop (more on this below).

As for the maturity transformation channel, Kohn and English have the better of the argument. As they observe, the report conflates the strategy of delivering ample reserves with a separate decision to pursue LSAPs as a form of stimulus when rates are at the zero lower bound (and the policy rate loses effectiveness in further stimulating the economy). They are also correct when they note the claim that asset purchases will usually result in losses for Treasury overstates the ability to predict the future path of interest rates. Textbook economic theory would argue that in a market as deep and liquid as U.S. Treasuries, whatever the subsequent result, the expected value of raising floating rate debt to purchase fixed rate assets is zero (if it was so obvious rates were going to rise as H&Y surmise, why weren’t they already higher?).

Kohn and English rightly point out that the report’s assertion that the Fed could avoid losses by directing banks to engage in QE themselves flies in the face of supervisors’ interest in having banks carefully manage interest rate risk. Finally, they observe that, even if the Fed was consistently acquiring long-duration securities and then “losing money” because interest rates rose faster than investors expected, this would be evidence that QE had worked in stimulating the economy. The whole point of the exercise is to take risk when the market will not. Any “losses” from the program due to rising rates must be weighed in context of the offsetting benefits, including incremental tax revenue, from a more rapid recovery.

The Real Subsidy is Much Bigger and Mostly Unrelated to the Fed’s Balance Sheet

H&Y focus on the spread banks earn on their reserves, claiming this represents a loss of value for taxpayers. This article argues the spread is less than they estimate and that little of its cost is borne directly by taxpayers. However, this shouldn’t be taken as meaning banks don’t enjoy any subsidy from their deposit franchise. One does exist and it’s much larger than posited in the report. While institutional investors earn higher rates, banks pay far less on the transactional deposits held by households and small/mid-sized businesses. The reason is straightforward; given the same-day liquidity they offer for making payments, as well as the security offered by FDIC insurance, depositors consider these balances to be “money” and don’t expect much in the way of yield. It is costly for banks to maintain branch networks and other deposit infrastructure, but depositors are paid so little there is still a meaningful spread left over as pure profit.

The details of the calculation will be laid out in a future article, but a reasonable estimate is that banks obtain an average pre-tax net profit of approximately 1% per annum on their total deposits. Given system-wide deposits of approximately $18Tn, this suggests banks earn nearly $150Bn per annum after-tax (roughly half of industry earnings) from a business that requires no capital and actually reduces the risk from their lending businesses. They are able to do so because they have charters that give them exclusive access to an FDIC backstop and the Fed as lender of last resort, as well as (mostly) exclusive access to the Fed-managed payments system. Banks are effectively capturing a large portion of the “seigniorage” benefit the government would otherwise earn from issuing money. This can only be described as a subsidy.

If discounted on similar terms to what is presented in the report, the value being sacrificed by taxpayers is well over $1Tn (the real value is likely much higher given the perpetual nature of the flows in question). H&Y are nipping around the edges of a larger problem and doing so by pushing potentially risky changes to the monetary policy framework. The better approach is to call out the larger subsidies at play and demand more fundamental reforms. For example, some scholars have advocated for eliminating the subsidy by imposing additional large additional assessments on banks. Absent this kind of direct reimbursement, policymakers should insist banks at least earn the subsidy they are receiving. It shows real hubris for bankers to complain about the cost of necessary regulation, cut back on the availability of branches and other services for local communities, and impose exorbitant fees on vulnerable customers, all while enjoying a large taxpayer subsidy. If the industry wants to rely on public resources for such a large fraction of their earnings, they must embrace the public mission inherent in their charters.

Deposit Franchise Channel: Evaluating Claims Made in the Report and Response

1: Funding Sources

H&Y assume banks’ reserve balances are fully funded by deposits. While Kohn and English argue it’s unclear whether this is the case, banks’ regulatory reporting confirm it’s a reasonable assumption.3 At the onset of the pandemic, inflows into transaction accounts and <$100k time deposits covered over 100% of the growth in total bank credit including the increase in reserve balances. In contrast, wholesale funding (e.g., >$100k time deposits and borrowings) actually declined on a net basis over this period and didn’t start growing on a net basis until the end of 2022 (Chart 1). Five years after the pandemic, wholesale borrowings finally caught up with the growth in reserves. However, over the horizon the report considers relevant, it’s fair to say those reserves were funded with deposits.

Chart 1: Cumulative Growth in Bank Credit and Deposit Balances (Monthly $Bn)

Source: Federal Reserve, H.8 Release, All Commercial Banks

The next question is what types of customers were providing those deposits. FDIC Call Reports provide disclosure regarding the composition of deposits by product and type of depositor. The underlying data set differs somewhat from the H.8, but the trajectory for growth in aggregate deposits is similar (Chart 2).

Chart 2: Composition of Cumulative Deposit Growth (Quarterly $Bn)

Household transaction balances rose by nearly ~$2Tn during the pandemic, but that likely reflects in part ~$750Bn of runoff from time deposits that didn’t offer meaningful yields while market rates were at zero (the pattern reversed when rates rose). In contrast, business transaction account balances spiked by nearly $4Tn. It is therefore reasonable to assume the increase in reserves was funded by business transaction account inflows.4

Sources: FDIC, Schedule RC-E and author’s calculations

2: Spreads

H&Y don’t identify an end date for the ample reserves policy, but nevertheless assume the associated deposits have a final maturity of ten years and an average life of five years (based on a straight-line amortization). Over that horizon, it assumes deposits will have a cost equal to 65% (i.e., their assumed beta) of the rate earned on reserves. Given an assumed forward IOR of 3.25%, current system reserves of ~$2.9Tn and system assets of ~$28Tn, projected net spread revenues of 1.14% on the declining balance produce a pre-tax net present value of $144Bn (Table 1). While H&Y don’t disclose their own calculations, these results are in line with those shown in Figure 4 of the report.

Table 1: Projected Interest Revenue, Expense, Spread, and NPV ($Bn unless noted)

Source: Author’s calculations

Capping the assumed subsidy based on a five-year average life is quite conservative given the Fed’s commitment to maintaining its ample reserve policy indefinitely. The most important driver of their estimated NPV is therefore the assumed 65% beta and how their methodology translates beta to an assumed spread. How reasonable are these assumptions? H&Y stipulate that reserves will likely be associated with institutional inflows that have a higher beta than other deposits. For their estimated beta, they cite a March 2024 Fed survey of bank senior financial officers that reported an average cumulative beta over the post-COVID tightening cycle of 69% for the non-operational deposits that are most rate sensitive. However, H&Y suggest there are reasons the inflows connected with reserves could have lower betas (and therefore lower costs under their methodology):

• For example, they note that when investors sell Treasuries to the Fed, they may redeploy the proceeds into corporate credit, and the seller of those bonds will “presumably” be less price-sensitive with respect to deposit yields (“presumably” is doing a lot of work here). They describe this process as “intuitive”, but it’s not at all clear why sellers of higher-yielding bonds should be any more likely to accept lower deposit yields. It seems far more likely that deposits end up being dispersed across the system but remain concentrated in the more competitively priced accounts offered to institutional investors.5

• The report also references Liquidity Coverage Ratio (“LCR”) disclosure from the four largest G-SIBs, noting that higher-beta “non-operational” deposits grew by only $420Bn between Q4’19 and Q4’21, far less than the $803Bn increase in cash balances and the $644Bn subtotal that was reserves. H&Y infer from these numbers that proceeds from Fed asset purchases must have been deposited in lower-beta operational balances. Given the way beta is used in their analysis, these operational balances are also assumed to provide higher spreads.

There are additional data sources H&Y could have cited to help confirm their assumptions. For example, BNY Mellon and State Street’s businesses consist almost entirely of managing investors’ ownership of securities and processing the settlement of transactions. Their approximate $600Bn in combined deposits include operational deposits related to securities settlement and nonoperational deposits from excess liquidity customers may place with them. The same LCR disclosure H&Y reference suggests roughly two thirds of these deposits were deemed operational during the pandemic.6

Consistent with the report, the custodial banks’ published rate disclosure shows a meaningful spread between IOR and rates paid to depositors (Chart 3). However, both companies source a significant fraction of their deposits in foreign jurisdictions (where rates are generally lower), so rates paid to domestic depositors are higher. State Street helpfully breaks out the rates paid on domestic deposits separately in their disclosure. The spread between IOR and these deposits has narrowed to 0.75% after exceeding 1.50% when rates first spiked. As noted above, banks will require some level of spread to cover their operating expenses. However, it wouldn’t be surprising if spreads narrowed further if forward rates are realized.

Chart 3: Custodial Bank Average Cost of Deposits vs. IOR (Quarterly %)

Sources: FRED and company quarterly financial supplements

3: A Few Words on Beta and Deposit Pricing

As noted above, H&Y’s methodology unhelpfully conflates the spread earned on deposits with their assumed beta. If true, banks’ deposit pricing decisions would be driven entirely by short-term market rates. This strains credulity - why would banks do it this way? Real-world deposit pricing reflects a complex equilibrium among banks and their customers. A key driver is banks’ desire to maintain a stable margin between assets and liabilities.7 This implies that deposit rates will be set, not at a percentage of a single short-term rate, but instead relative to a portfolio benchmark that includes a mix of underlying maturities. If deposit yields are set relative to longer maturity assets, they will move in tandem with a trailing average of market rates as those assets roll over. Historical rates may appear to exhibit relatively consistent betas over the last few rate cycles, but this is an artifact of specific starting assumptions (rates at 0.0% at the beginning of the tightening cycle) and a relatively stable term structure.

For example, if we compare 3-Month Treasury rates to an index equal to a trailing average of 1-Year Treasury rates (less a spread of 1.0% and with a floor rate set at 0.0%), we get something that bears a striking resemblance to the custodial bank pricing shown above (Chart 4).

Chart 4: 3-Month UST Rate vs. Trailing Average 1-Year UST Index

Sources: Federal Reserve, H.15 Release, trailing average calculation incorporates the most recent quarter spot rate based on the assumed maturity (e.g., 50% per quarter for a 6-Month, 25% for a 1-Year and 12.5% for a 2-Year)

While the index appears to exhibit a cumulative “beta” of 66% vs. the 3-Month rate, this is illusory. What is actually happening is that the index moves one-for-one with its actual benchmark (less a spread). Of course, absent a willingness to impose negative rates, that spread can’t be maintained when the index approaches the zero lower bound. A backward-looking cumulative beta therefore has little value in predicting how rates will move - when rates rise, observed betas may be quite low until spreads normalize. This property, as well as the lagged nature of a trailing average (causing it to rise even after spot rates have peaked), is likely why deposit rates exhibit the apparent increase in sequential betas relative to short-term rates observed in another study from Younger (i.e., the apparent “convexity” of deposits is mostly a byproduct of using the wrong benchmark).

The compression in effective spreads at the zero lower bound encourages banks to reduce reserves (i.e., by imposing negative interest rates or surcharge fees on their “excess” deposits). This has the effect of pushing these deposits and their associated reserve balances toward banks that are more tolerant of zero spread positions (e.g., those for whom leverage capital requirements aren’t the binding constraint) and willing to accept these deposits as an accommodation for customers. Nevertheless, the variability in effective spreads should be factored into any forward-looking estimate of their expected value (e.g., if spreads are 1.00% in 80% of outcomes and 0.00% in the remaining 20%, the expected spread would be only 0.80%)

4: Cost of Capital

H&Y’s discussion of the cost of capital is more ambiguous. They correctly note that reserves don’t add to a bank’s risk-weighted assets and leverage capital requirements are not the binding constraint for most institutions. For banks where risk-based capital is the binding constraint, any spread obtained from holding reserves just inflates aggregate returns. However, they go further and argue all returns are an unnecessary subsidy, even for banks where leverage requirements are the binding constraint. If one’s starting position is that banks should not be paid to hold reserves, any return to shareholders obviously represents a subsidy. If one instead believes the ample reserves policy is part of the wider monetary policy framework and doesn’t serve banks’ narrow commercial interests (e.g., because even a positive spread dilutes Net-Interest Margin and ROA), it would be reasonable to let them cover the cost of capital that could otherwise be returned to shareholders. This argues for calculating estimated subsidies based on excess returns over the cost of capital for those banks where leverage capital is the binding constraint.

Additional Thoughts on the Maturity Transformation Channel

There is not much to add to Kohn and English’s critique of this part of the report. However, one point deserves additional comment. H&Y argue “there are reasons to believe commercial banks benefit in an asymmetric way” from the Fed’s QE purchases. They claim this is because regulators could direct banks to purchase securities if this was deemed necessary and point to the introduction of the LCR rule as an example of this power. However, the LCR does not require banks to stockpile duration, and this is the fundamental point of QE. A bank could meet its high-quality liquid asset (“HQLA”) requirements by holding long-duration U.S. Treasuries or GSE MBS while offsetting that additional duration elsewhere in its balance sheet (e.g., by extending its debt maturity profile, shifting a portion of its loan production from fixed to floating, or through derivatives transactions). As Kohn and English note, the only way for bank purchases to substitute for QE is if the banks are also directed not to hedge their interest rate risk.8 Given banks’ relative lack of financial resources compared with the Fed, this seems ill-advised.

Indeed, one of the best justifications for a large Fed balance sheet is that it facilitates the post-GFC regulatory model where banks stockpile liquidity on balance sheet. Reserves act as a key lubricant for that model given the pitfalls (e.g., negative mark-to-market, transmitting liquidity stress to the Treasury market, etc.) of maintaining liquidity entirely in the form of securities. However, in a scenario where the LCR was amended to allow HQLAs to be replaced by committed secured liquidity from the Fed, banks would have less need to hold reserves.9 This is the scenario where the Fed could consider returning to the corridor system. On the other hand, attempting to reduce the balance sheet while maintaining current liquidity requirements would be asking for a repeat of the 2019 repo market dislocation. In the worst case, where banks didn’t hedge their HQLA portfolios, it runs the risk of encouraging a repeat of the 2023 failures at Silicon Valley and First Republic. Banks don’t typically take the duration risk the Fed does on its balance sheet. When they do, it can end badly.

Evaluating Recommendations from the Report

Having spent the first part of the report asserting that paying interest on reserves is a bad deal for taxpayers, H&Y finish by laying out a series of proposed alternatives. These range from attempting to reduce banks’ demand for reserves so that the Fed can run a smaller balance sheet to eliminating the payment of interest on some or all of banks’ reserve balances.10 Of course, a floor system based on the provision of ample reserves was established for a reason. Any of these alternatives would require the Fed to more frequently intervene to provide emergency liquidity and conduct more of its business through nonbank counterparties. Most importantly, unless reserve balances were significantly contracted, it would put the Fed’s ability to transmit monetary policy at risk. Even if the Fed retained its reverse-repo facility in an attempt to set a floor on money market rates, the lack of interest on reserve balances would risk a sluggish transmission of policy rates to sectors that rely primarily on bank lending for funding (e.g., real estate construction, small business credit and credit cards). If policymakers are in earnest about eliminating interest on reserves, the best approach (as noted above) is to return to the corridor system while revising the LCR and other liquidity rules to reduce on-balance sheet liquidity requirements. However, given the downsides of this approach, a better reason is needed then to eliminate a subsidy that is only tangentially connected to the Fed’s balance sheet and a small part of a much larger benefit banks gain from their privileged role in money creation.