How to Judge a Regional Bank Without Shrinking It
The answer to the title’s question is conditional: regional banks are not simply smaller large banks because their business model changes how shocks travel through the balance sheet, but that difference does not make them uniformly safer or more fragile. The useful judgment is to match the bank, the shock, the analytical tool, and the evidence. This final synthesis brings those elements together so a reader can evaluate a regional bank without mistaking a group average for an institution-specific conclusion.
The answer is a conditional one
The strongest evidence is structural. In the Federal Reserve comparison for 2019Q4, regional banks held deposits equal to 75.71 percent of assets and loans equal to 69.26 percent, compared with 60.07 percent and 50.28 percent at large banks. Commercial real estate made up 41.64 percent of regional-bank loans, versus 16.75 percent for large banks, while large banks had much more credit-card exposure. The 2023Q4 appendix showed the same broad direction after the pandemic and the 2023 banking stress. These are not merely different sizes of the same balance sheet. They are different concentrations of claims on borrowers, property markets, depositors, and sources of revenue.
The revenue and funding mix gives the structural difference a mechanism. Regional banks rely more on interest income from loans and on deposits as a funding base. In 2019Q4, their interest income was 4.29 percent of assets, versus 3.90 percent for large banks, while interest expense was 0.93 percent, versus 1.25 percent. Large banks had more investment-banking, fee, other noninterest, and trading income, consistent with a more diversified and capital-markets-oriented model. Each design has a different risk bearer. In a traditional lending model, borrowers and properties determine more of the bank’s income and credit outcomes, while the bank’s capital absorbs recognized losses. In a diversified model, market activity and wholesale funding introduce other ways for earnings and valuation to change.
Regional banks’ reliance on low-cost deposits and longer-duration, often fixed-rate assets can also produce a two-sided rate profile. When yields fall, asset yields may adjust slowly and funding costs can remain favorable, supporting the margin. When yields rise, the market value of longer-duration assets can fall, and deposits may become more expensive later. A fair-value decline is not automatically a realized credit loss, but it can affect collateral, financing choices, confidence, and the room available to respond. The same balance-sheet choice can therefore support current income in one rate environment and create a vulnerability in another.
The Federal Reserve’s FLARE results show that the distinction reaches beyond a descriptive balance-sheet table. Regional banks were generally less sensitive than large banks to aggregate macroeconomic and financial variables, and many PPNR components and NCO rates were more persistent. In the illustrative equations, the three-month Treasury-yield coefficient for interest income was 0.184 in the regional model and 0.255 in the all-bank model; for interest expense, it was 0.156 and 0.207. A lower coefficient means a smaller modeled short-run response to the same aggregate variable. Higher persistence makes the projected path flatter and carries recent conditions further into the forecast.
That evidence supports a narrower conclusion than “regional banks are resilient.” It says that a regional-bank model can be a better description of how the group responds to broad macro-financial stress. It does not say that every regional bank reacts less to every shock. Construction and multifamily CRE were important exceptions, with regional-bank NCO rates more sensitive to aggregate CRE prices or unemployment in those segments. A large CRE share combined with higher sensitivity in a selected segment can matter more than a lower average sensitivity across the rest of the loan book.
The scenario results make model choice consequential. In the Federal Reserve’s counterfactual exercise, both models were trained on 1997Q1–2023Q4 and projected nine quarters under the 2024 severely adverse scenario. Using regional-bank coefficients for regional banks produced higher PPNR, smaller aggregate NCO increases, and a CET1 decline that was nearly one percentage point less at the largest difference by the end of the horizon than the path generated with all-bank coefficients. About two-thirds of the CET1 difference came from PPNR projections and about one-third from NCO projections.
That nearly one-percentage-point gap is not a universal safety margin. It is the result of a specified scenario, historical sample, asset-weighted regional-bank aggregate, and fixed-balance-sheet framework.