Risk Tolerance, the Price of Certainty, and the Value of a Bank Charter

Doug Simons is a former investment banker, having worked in Morgan Stanley's mortgage securitization business and then leading the advisory efforts at Credit Suisse and UBS related to bank capital, asset/liability management and housing finance reform. Most recently, he served as a Senior Markets and Policy Fellow at the CFPB, where he advised Director Chopra and his team on issues related to the banking industry and represented the Bureau on FSOC’s Systemic Risk Committee.

In an incisive piece published in Open Banker last month, Tom Brown drew on the work of economist Ronald Coase to expound a framework for interpreting how new technologies can change the structure of markets. Coase’s seminal “The Nature of the Firm” argued that firms exist to organize commercial activity too expensive to conduct as a set of discreetly negotiated transactions. Tom observes this dynamic isn’t limited to firms: other institutional arrangements (e.g., corporate brands, professional licenses, or market standards such as bond ratings and FICO scores) serve as informational “proxies” that provide a trusted third-party endorsement allowing for transactions to proceed without the need for costly ongoing due diligence. He argues that as artificial intelligence drives “the marginal cost of evaluating information toward zero,” it will erode the need for these proxies and undermine the competitive position of the incumbent institutions who provide them. While Tom’s piece doesn’t name banks among these incumbents, one might assume they’d be at risk as technology makes pricing more transparent and provides customers with greater choice. However, these fears misread the source of banks’ value to their customers.

Banks’ Competitive Advantage Is Their Ability to Monetize Certainty

In a prior Open Banker piece, Tom argued the GENIUS Act (and other legal changes) made stablecoins and tokenized deposits viable alternatives for quickly moving large amounts of money. He suggests these “better, faster, cheaper” forms of payment will subvert banks’ longstanding role as “ledgers of record for money at rest” (a description that calls to mind the Grayson Moorhead skit from SNL). It’s true that banks have done little to visibly improve efficiency, with data from the FDIC’s Quarterly Banking Profile indicating that employee compensation and other expenses are unchanged (as a share of assets) since the mid-1980s. Research from NYU professor Thomas Philippon suggests this stability goes all the way back to the time of J.P. Morgan (the man not the bank), with financial services value added fluctuating between 1.5% and 2.0% of “intermediated assets” since the 1880s.

Philippon calls this enduring stability the “intermediation cost puzzle” and asks why advances in information technology haven’t led banks to reduce costs and cut pricing. He considers explanations based on market power and oligopolistic pricing, but doesn’t identify any conclusive trends. I would suggest the best explanation is offered by Tom’s “Coasian” framework. As his recent piece notes, the products that are safest from disintermediation are those “where the proxy was never standing in for hidden information at all, but is an irreducible part of what's being purchased.” In the case of a bank account, that proxy is banks’ access to a taxpayer guarantee. Unlike a rating, it doesn’t merely convey an assessment of risk. It represents the sovereign assurance that the “number” listed in one’s account can be immediately converted back into physical goods and services (that is what makes deposits “money”). My previous work on this topic suggests depositors are willing to accept a substantially (approximately 200bps) lower “convenience yield” relative to what they could earn on comparably safe Treasury securities or government money market mutual funds (but where there is a delay before sale proceeds can be spent).¹ This observed gap in yields can be interpreted as the minimum price customers are willing to pay for total certainty about their access to purchasing power.²

A willingness to accept low deposit yields anchors the banking industry’s earnings, limiting the incentive to reduce expenses and compete on price.³ Even after accounting for the costs associated with running a branch network, the yields are low enough to provide a meaningful (approximately 100bps pre-tax) spread relative to the cost of wholesale funding. In addition, access to a government backstop allows banks to make relatively long-term fixed rate loans without fully hedging their interest rate risk. Their ability to capture the term premium has been a major source of profitability, particularly over a multi-decade period of declining interest rates (indeed, it’s unclear how profitable banks’ mostly prime-grade lending would be after accounting for credit/other costs if it was fully match-funded).  

Implications for FinTechs

Despite good faith efforts to identify conditions under which payments could be a purely commercial activity distinct from taking deposits, maintaining this separation is impossible in practice. The reality is that customers lack the necessary visibility into (or control over) the timing of payments and receipts. Given the consequences of running short, they understandably want to maintain a liquidity buffer. While consumers might permit a modest balance to accumulate in their Venmo account and some are dabbling in stablecoins, they want to hold their buffers in a form considered as safe as “cash” and that almost always means at a bank. Consumer-focused fintechs have attempted to square this circle⁴ by partnering with banks, but these hybrid models are unlikely to work at the scale that wealthy households and businesses (who represent the bulk of industry deposits) expect of their primary financial relationship. It’s also worth noting that banks appear to charge relatively little to process the vast majority of their payments.⁵ Even if customers could instantly shift money back and forth between their bank and a specialized payments firm, they therefore have little reason to do so. 

In short, given the premium placed on liquidity, operating efficiency isn’t sufficient on its own (outside of a few consumer niches) to make stablecoin issuers and money transmitters competitive with banks. In order to be truly competitive, they would need to offer higher-balance customers the same ironclad liquidity guarantee possessed by bank deposits. Stablecoin advocates claim their reserve portfolios can be reliably converted to cash, but this sort of contingent claim is not the same thing as a guarantee. A Coasian analysis underscores that one expresses a forecast about the ease of turning assets into purchasing power, whereas the other eliminates the risk entirely. No amount of other useful features (e.g., speed, programmability and the micro-payments suited to agentic commerce) are likely to make up for that discrepancy. As Tom notes, when the proxy is the “actual product,” “no amount of cheaper information changes that.” Little wonder the crypto industry saw the ability to pay a market yield as a red line in the failed CLARITY Act process. They are well aware these products aren’t “money” in the same sense as bank deposits and understand future growth depends on the ability to pay rates that can compete with short-term investments that lack a convenience yield. 

Is It Unfair for the Government to Give Banks a Competitive Advantage?

Do deposit guarantees provide banks with an unearned subsidy? Yes! No doubt the approximate $150Bn per annum taxpayer subsidy banks are receiving should be reinvested in their communities (if not taxed away altogether) rather than captured by shareholders. However, this wouldn’t impact the competitive advantage subsidized pricing provides, one that isn’t going away. Policymakers understand the public has liquidity preferences (they want a long-term maturity on their debts while retaining immediate access to their savings) that are impossible to reconcile in a closed system. Through long and painful experience, they’ve also learned that allowing households and businesses to enjoy a surplus liquidity position is better for long-term economic growth. Someone else must therefore be in a deficit position, and in our system that role is played by banks (operating with a taxpayer backstop). 

Critics who complain banks are a “cartel” are therefore making a category error. The banking oligopoly isn’t some sort of market failure: banks are quasi-sovereign entities and restricting access to taxpayer support ensures the rest of the economy operates more squarely on a free-market footing (the oligopoly isn’t a bug, it’s a feature). Companies that want to compete in the payments business (and therefore take responsibility for customers’ liquidity) should do the work necessary to obtain a bank charter. In return, regulators should offer a realistic path for applications to be approved (including considering specialized charters better suited to a more limited business model). Once depositors’ liquidity risk is equalized, competition can shift to areas (e.g., cost efficiency and service quality) where superior technology can be decisive.

Will AI Reduce Intermediation Costs?

While technology offers clear opportunities to reduce physical costs (personnel, equipment, energy, time, etc.), it’s less clear what impact it will have on the risk-based financial costs (liquidity buffers/spreads, credit spreads, and term premia) that constitute most of a bank’s revenues. As noted above, Philippon’s work shows intermediation costs have been stable for over a century. Similar stability can be seen in U.S. Treasury term premia, stock market volatility, and corporate bond spreads. My own work on J.P. Morgan Chase’s deposit pricing suggests the convenience yield premium on deposits hasn’t changed meaningfully in over twenty years. One way to interpret these trends is that markets’ ability to forecast outcomes hasn’t really improved over the years, despite the explosion in available data and sharply reduced cost of turning that data into useful analysis.

Will this structural stability continue? Maybe humans were the bottleneck and AI will permit a sea change in forecasting as vast amounts of data can now be analyzed in detail. Former Treasury Under Secretary Jonathan McKernan recently advanced the view that an AI agent could do a better job of modeling household and business cash flows, permitting them to operate with smaller liquidity buffers and move excess deposits into higher yielding alternatives. However, this ignores the risk aversion inherent in liquidity planning (i.e., the convenience premium applied to yields is necessarily matched by similar conservatism in deciding how much to hold). Buffers aren’t sized to expected needs, but rather to a perceived worst-case outcome. Better data indicating just how unlikely that scenario may be isn’t all that relevant. 

At a more macro level, Tom points to Jim Simons (no relation) and Renaissance Technologies as an example of where technology enabled better forecasting. If known unknowns can be reduced, the implication is that households and businesses wouldn’t need to pay as much for financial intermediaries to absorb risk. While this scenario is possible, I suspect a different dynamic is at play. Access to better data and analytical tools (including machine learning) has already improved forecasting across a range of domains (e.g., weather, logistics, epidemiology, etc.) and the impact of the latest generation of AI will allow additional progress. However, rather than using these improvements to stabilize returns (and hope they are rewarded by creditors and shareholders), companies (often at the direction of their private equity owners) have taken on additional risk to boost returns. Corporate debt has increased by roughly 70% as a share of net value added over the last fifty years.⁶ Risk in companies’ underlying businesses may have declined, but higher leverage has left both debt and equity looking as risky as they did historically. Similarly, investors’ increasing comfort with funding nascent industries through venture capital and leveraged finance has meant markets are funding a riskier set of companies than they did previously. 

Putting all this together, risk tolerance appears to exhibit a “barbell” distribution: while emergency reserves should ideally have no risk at all, the remaining risk “budget” is essentially fixed and spent on maximizing returns. Improved technology is therefore unlikely to lessen the demand for the financial sector to absorb those risks (e.g., liquidity) households and businesses don’t want to bear themselves. Banks have a unique ability to absorb liquidity risk given their government backstop, and they should be able to continue extracting a high price for providing customers with certainty (not just guaranteed access to deposits, but also a long-term fixed rate on loans). A bank charter therefore remains essential for companies hoping to compete in managing customer liquidity.

The opinions shared in this article are the author’s own and do not reflect the views of any organization they are affiliated with.

^[1]  This is a critical point: the low yield on deposits is not about the government guaranteeing their ultimate repayment at par (as that guarantee is available from higher-yielding Treasury securities). It is instead about the additional guarantee of immediate availability in a form that can be spent. Some researchers have ascribed depositors’ willingness to accept a below-market yield to inattention (or “sleepiness”), the implication being that greater transparency and active surveillance (e.g., by an AI agent) will increase account switching. This defies common sense: most deposits are from wealthy households and businesses, deposits are a small fraction of those parties’ total assets (which likely also include short-term government debt), they are sophisticated enough to understand what they are earning, and it would be inexpensive to move funds to obtain a higher rate. The more reasonable interpretation is that they accept a lower rate because they value the liquidity their deposits provide.

^[2]   I describe it as a minimum as it’s not at all clear that depositors would willingly move funds into an unguaranteed account merely to earn an additional 200bps. Obviously, they will accept whatever yield banks are willing to pay on a fully guaranteed account, and the 200bps is best interpreted as the competitive equilibrium among banks (based on their own costs) for where these accounts can be supplied. However, because of the convenience cash provides, depositors don’t really expect a return and this limits the competitive pressure to cut costs and boost yields.

^[3]  Of course, savings could be used to boost profitability, and the industry claims to be intensely focused on costs. The apparent stability in overall expenses may therefore reflect efficiency gains being reinvested into upskilling and improved service quality (sacrificing profitability to defend market share may be a sensible strategy in the context of returns already exceeding the cost of capital).

^[4]  Editor’s note: Or, if you will, “Dorsey this Allaire.”

^[5] According to McKinsey’s 2025 Global Payments Report, North American payments revenue totaled $700Bn in 2024. However, they estimate 51% of this amount is attributable to credit cards, where most of the cost is borne by merchants (and only indirectly by customers through higher prices). Another 19% is attributed to deposit net-interest income, but this conflates the benefit from making specific payments with the value obtained from safely storing funds. Only the remaining 30% (i.e., $210Bn) is directly attributable to non-credit card fees, and $56Bn of that amount relates to cross-border transactions. The remaining $154Bn of fees represent only 0.12% of the $128.5Tn in 2024 non-credit card domestic payments reported by the Federal Reserve.

^[6]  Per the Federal Reserve Z.1, corporate borrowings were 33% of GDP in 1976 while net value added was 48% (a ratio of 69%). By 2025, borrowings had risen to 50% of GDP (after peaking at 59% in 2020) while net value added had fallen to 43% (a ratio of 117%). Corporate value added has risen more slowly than GDP in recent decades as activity expanded in business not organized as corporations. The increase in debt has been matched by an increase in profit margins (profits defined as the sum of retained earnings, dividends paid and net buybacks), so leverage looks basically unchanged on that basis (reinforcing the point that leverage has been used to maintain risk across different parts of the capital structure).

Open Banker curates and shares policy perspectives in the evolving landscape of financial services for free.

If an idea matters, you’ll find it here. If you find an idea here, it matters. 

Interested in contributing to Open Banker? Send us an email at [email protected].