Regional Banks and CRE: Where Spillover to Housing Is Most Likely

At the end of 2025, US depository institutions held $3.0 trillion of commercial real estate loans. The top four banks (JPM, BAC, C, WFC) hold just 14% of that stock. The rest — $2.6 trillion — sits at the roughly 4,000 community and regional banks below the systemic-importance threshold. For comparison, those same small banks hold only 39% of US residential mortgage debt (most of which was securitized to the agencies).

This matters because the FDIC's Supervisory Guidance on CRE Concentration Risk flags banks whose CRE loans exceed 300% of total risk-based capital, or whose construction and development (C&D) loans exceed 100% of capital. Banks above these thresholds aren't automatically in trouble — but they face heightened supervisory attention and are the ones most likely to pull back on new originations during a stress event.

The within-group dispersion is the bigger story. Averaging is deceptive: at the $1B–$10B cohort, the 90th percentile bank has a CRE ratio of 520%, the 10th percentile is at 95%. A single bad actor above 500% can blow up a state's commercial lending market for six months while it works out its book.

To get to a local estimate, we geo-locate each bank's deposit base using FDIC Summary of Deposits, then allocate its CRE portfolio proportionally across the metros where it has meaningful presence (>5% of branches or >10% of deposits). This gives us a per-metro estimate of local bank CRE exposure as a share of that metro's total CRE debt stock.

The 14 metros flagged share a profile: mid-sized, regional-bank- dominated lending market, and office vacancy at or above the national average. None of them are the big headline "CRE stress" markets (NYC, SF, LA), because those markets are funded primarily by money-center banks, CMBS, and life insurers — the regional bank channel is smaller share, so even a large office markdown is absorbed across many balance sheets.

For each flagged metro, we run a simple stress: assume a 10% portfolio markdown on all CRE loans held by the top three local banks by CRE exposure. What does that do to CET1?

CET1_stressed = CET1_current - 0.10 × CRE_loans × (1 - ALLL_coverage)
where ALLL_coverage is the bank's current allowance for loan losses
as a share of CRE book value.

The stressed scenario is not implausible. CMBS CRE loan delinquency hit 7.8% at end-2025 and office-specific delinquency is 13.2%. Bank-held CRE loans are a different population (smaller, less securitizable, often amortizing, locally relationship- managed) so their loss curves are lower — but 10% book-value markdowns on just the CRE portfolio at regional banks with 450%+ concentration ratios is within the envelope of the Fed's 2024 Supervisory Climate Scenario severely-adverse path.

When a local bank constrains new lending, the effect on the residential side of the same metro runs through three channels:

On net, we estimate a realistic regional-bank CRE stress event (10% markdown + 20% reduction in new origination for 12 months) shaves 0.2–0.5pp off forward HPI in flagged metros, primarily via the small-business-employment channel. It's a real effect but smaller than the direct mortgage-rate channel.

1. CMBS special-servicing rate for office loans. This is a leading indicator for bank-book realizations — special servicing starts 6–9 months before comparable loss recognition on bank balance sheets. Currently at 13.2% and rising, Dec 2025.
2. FDIC Q2 CRE guidance update. The agencies are expected to publish revised concentration guidance in 2026Q2 that lowers the C&D flag from 100% to 80%. That would put 42% of community banks on the flag list overnight — more supervisory attention, same books.
3. Deposit flow regime. The 2023 deposit flight (SVB, First Republic) was driven by rate differentials, not credit concerns. A 2026 equivalent driven by CRE credit concerns would look meaningfully different — sticky insured deposits leave faster, brokered replacement is more expensive. Monitor weekly H.8 small-bank deposit series.
4. Quarterly FFIEC allowance coverage. If regional banks start provisioning proactively (ALLL-to-CRE rising meaningfully), that compresses near-term earnings but buys resilience. Current median ALLL/CRE is 1.3%; stress scenarios imply loss content of 3–5%.

We don't currently apply a CRE-stress residual to metros in the forecast model — the signal is too slow-moving and the attribution too noisy to include as a systematic feature. But we flag the 14 metros above internally and watch the CMBS special- servicing rate and FFIEC ALLL coverage as risk indicators. If either moves materially in Q2 2026, we will re-run this analysis and consider adding a metro-level CRE-stress adjustment to the v1.0 model (targeted Q3 2026 release).