The Balance Sheet Behind the Difference
For regional banks, the balance sheet is not a passive inventory of assets and liabilities. It is the operating model that determines how the bank earns money, absorbs losses, and responds to a change in rates or property values. The specific question here is how funding, lending, revenue, and geography fit together—and how a reader can use that map to identify which shock channels deserve attention before looking at a stress projection.
Deposits, loans, and CRE define a different commercial-bank model
The Federal Reserve comparison starts with a useful contrast. In the fourth quarter of 2019, the study examined 143 regional banks and 41 large banks. At regional banks, deposits were 75.71 percent of assets and loans were 69.26 percent. At large banks, the same ratios were 60.07 percent and 50.28 percent. A regional bank therefore had more of its balance sheet tied directly to deposit funding and lending activity, while a large bank held a smaller share in each and had more room for other businesses and instruments.
The loan mix is even more revealing. Commercial real estate made up 41.64 percent of regional banks’ total loans, compared with 16.75 percent at large banks. Credit-card loans made up 2.07 percent of regional banks’ loan books but 10.76 percent of large banks’ books. These are shares of the relevant group’s total loans, not default rates. Their importance is that they identify different borrowers, contract structures, and economic drivers. A CRE-heavy lender will care more about rents, vacancies, property prices, collateral, and refinancing dates. A bank with more credit-card lending will be more exposed to household cash flow, revolving credit behavior, and consumer credit cycles.
The difference is not that one bank group has “real” loans and the other does not. It is that each group has a different concentration of claims on the economy. Regional banks’ loan-heavy balance sheets make the performance of borrowers and properties central to revenue and losses. Large banks’ broader mix can spread some exposure across consumers, capital markets, trading, and transactional services, while creating other risks that are not present to the same degree in a traditional lender. Diversification changes the map; it does not remove risk from it.
The balance-sheet percentages also need to be read with their denominators. Deposits as a share of assets describe the funding structure, while CRE as a share of total loans describes the composition of the lending book. Neither number tells us the loan’s maturity, interest rate, collateral quality, borrower strength, or location. A high CRE share is therefore a reason to ask more specific questions, not a sufficient basis for a loss estimate. The evidence becomes useful when the ratio is connected to a mechanism and a time horizon.
The balance-sheet architecture was not confined to a single pre-pandemic quarter. In the appendix’s fourth-quarter 2023 comparison, deposits were 77.16 percent of regional-bank assets and 60.98 percent of large-bank assets; loans were 66.28 percent and 46.01 percent, respectively. CRE still accounted for 44.24 percent of regional-bank loans, against 13.24 percent for large banks. The exact group counts and ratios changed, but the direction of the differences remained. That persistence supports a business-model interpretation rather than a story about an unusual quarter.
The income statement reflects the same design. In 2019Q4, regional banks’ interest income was 4.29 percent of assets, versus 3.90 percent at large banks. Their interest expense was 0.93 percent, versus 1.25 percent at large banks. A bank’s net interest income is shaped by the gap between what its assets earn and what its funding costs. A regional bank with more loans and lower-cost deposits can generate more of its revenue through that gap. A large bank’s more diversified activities produce more investment-banking, fee, other noninterest, and trading income. Diversification may reduce dependence on one spread, but it also introduces different market, trading, and wholesale-funding exposures.
The capital line belongs in this map as well. The same 2019 comparison reported total equity capital equal to 12.78 percent of regional-bank assets and 11.79 percent for large banks. In the 2023Q4 appendix, the corresponding figures were 11.19 percent and 9.84 percent. These group averages do not establish that regional banks can absorb every loss, and capital ratios vary across institutions. They do show why an analysis that jumps directly from “smaller” to “more fragile” misses another relevant dimension of the comparison: the groups’ balance-sheet proportions are different before a shock arrives.
A higher loan share and a higher CRE share should not be read as an automatic prediction of failure. Exposure answers “where could the loss come from?” It does not answer “when will it appear?” or “how large will it be?” A loan can remain current while its collateral value declines, and a property market can weaken without every borrower defaulting. The balance sheet is a map of possible transmission, not a forecast by itself.
Funding, duration, and geography turn the map into a risk profile
The first major interface is between deposits and assets. Regional banks generally finance a greater fraction of their liabilities with low-cost deposits and carry more duration on the asset side. Duration is a measure of how much an asset’s value can change when interest rates move; longer-duration assets are generally more sensitive to a rate increase. The Federal Reserve authors connect this combination to a lower short-run pass-through of short-term rate changes to both sides of the regional-bank balance sheet.
That mechanism can support earnings in one environment. If market yields decline, a bank holding longer-duration, predominantly fixed-rate loans may not see its asset yields fall as quickly as a portfolio of short-term loans. If its deposit costs also adjust slowly, the net interest margin can remain comparatively stable. But the same slow adjustment can create a different problem when rates rise. The market value of fixed-rate assets can fall, producing fair-value losses or pressure if the bank must sell or refinance around a stressed date. The bank may therefore look resilient in current income while carrying greater economic exposure to a rising-rate path.
This is a trade-off, not a contradiction. A bank can report stable current income while holding assets whose economic value is more sensitive to rates. It can have strong capital and liquidity at the system level while a particular balance sheet has a difficult maturity or deposit profile. It can have a high CRE share without immediate delinquencies, while still facing a future refinancing wall. Reading the bank requires keeping these time horizons separate rather than asking whether the institution is simply “safe” or “unsafe.”
The distinction between exposure, sensitivity, and capacity is useful here. Exposure tells us how much of a balance sheet is connected to a sector or funding source. Sensitivity tells us how an economic variable has historically moved an income or loss measure. Capacity tells us how much capital, liquidity, earnings, and management flexibility can absorb the movement. The Federal Reserve’s comparison supplies strong evidence about the first two at the group level; it does not turn them into a complete estimate of the third for any one institution.
This is why earnings and market value should not be treated as interchangeable. Interest income describes the return being recognized from lending; fair value describes what an asset may be worth under current market conditions. A fair-value decline does not automatically become a realized credit loss. It can still matter because it affects financing choices, investor confidence, collateral, and the bank’s ability to respond if deposits leave or borrowers need new terms. The relevant judgment depends on the accounting treatment, the asset’s maturity, the funding schedule, and the bank’s ability to hold the position—not on one ratio alone.
The second interface is between the loan book and the property market. A regional bank with a large local CRE portfolio may know its borrowers and properties in more detail than a national aggregate index can express. That relationship information can make its reported losses less responsive to a noisy national price measure. It can also make the bank more dependent on the specific local market it knows well. If office vacancies, rents, construction costs, or refinancing conditions deteriorate in that market, local information does not diversify the exposure; it may simply improve the bank’s ability to identify it.
The third interface is between local knowledge and geographic diversification. In 2025, the median large bank operated in approximately 14 states, while the median regional bank operated in about four. About 20 percent of regional banks operated in only one state. For an individual institution, that footprint can make local employment, industry, housing, or CRE conditions unusually important. A bank may understand those conditions deeply and still be concentrated in them.
Yet individual concentration should not be confused with sector-wide isolation. The regional-bank sector’s deposits were broadly distributed across the United States. The five largest states accounted for around 34 percent of regional-bank deposits, a share close to their contribution to national output, and the Federal Reserve reports a Herfindahl–Hirschman Index well below 1,000. That means a local shock can be severe for one regional bank without necessarily becoming a systemic shock. It also means that a broad national downturn can affect many regional banks at once even if no single bank has a nationwide footprint.
The OCC’s Spring 2026 risk assessment provides a useful current backdrop, but it must be attached to the right population. It reports that net interest margins improved across the banking system in 2025 as funding costs declined. Banks with less than $30 billion in assets saw larger improvements, partly because their asset yields were more favorable, while loans at larger banks repriced downward more quickly. The report also says unrealized securities losses fell to their lowest levels since 2021. These observations are consistent with the importance of funding costs and asset duration, but they are not regional-bank statistics: the OCC’s under-$30-billion group is narrower and differently defined than the Federal Reserve’s $10-billion-to-under-$100-billion regional-bank group. The distinction is part of reading the evidence correctly.
The relationships in this diagram are analytical rather than a claim that every regional bank has identical contracts. They show the questions that should be asked in sequence. A reader should first identify the funding base, then the asset mix, then the geography and revenue sources. Next comes the timing question: which items reprice immediately, which mature later, and which losses become visible only after a borrower or property market deteriorates? Finally, the reader should ask how the bank’s capital and liquidity position can absorb the resulting change.
A practical example shows why the sequence matters. Suppose rates rise while a regional bank holds many fixed-rate CRE loans and funds itself with deposits that initially remain inexpensive. Its current interest margin may not deteriorate immediately; in fact, the slower repricing of assets and liabilities can make the income statement look calm. But the market value of fixed-rate assets may fall, and a later CRE refinancing wave may force borrowers to confront higher debt-service costs. If deposits become more expensive at the same time, the original funding advantage can narrow. The risk is not captured by the CRE share, the margin, or the capital ratio considered in isolation. It emerges from their interaction over time.
The same reasoning applies to geography. A bank that operates mainly in one state may have better information about local borrowers than a national model, but a local recession can affect many of its loans together. A sector whose deposits span many states may be diversified against one region, yet still share a common sensitivity to rates, national property financing, or broad unemployment. The reader should therefore ask whether the evidence describes a bank, a peer group, or the system before drawing a conclusion about systemic risk.
That interaction also explains why the bank’s revenue mix matters. Large banks’ greater fee, trading, and investment-banking income can provide sources of revenue that are less directly tied to traditional loan spreads, but those activities respond to markets and capital-market conditions. Regional banks’ more concentrated interest-income model may be steadier under some broad conditions and more exposed to a small number of lending and funding relationships under others. The evidence does not support a universal winner. It supports a conditional profile.
The balance-sheet map is therefore a screening tool, not a verdict. It can tell a reader which shock variables deserve a closer look: short- and long-term rates, deposit pricing, CRE prices, rents and vacancies, refinancing volumes, household credit, and local economic conditions. It cannot, by itself, say how a particular bank will perform. To answer that, the next question is whether the stress model uses relationships estimated from banks with this same architecture—or whether large-bank behavior is doing too much of the explaining.