Deposit Vulnerability & Franchise Analysis

Funding mix · uninsured exposure · deposit flight · rate sensitivity · Q1 2026 FDIC Call Report
Loading scores…

Vulnerability Tier Distribution

Franchise Tier Distribution

Most Vulnerable Banks — Top 20

Strongest Deposit Franchises — Top 20

Funding Mix & Stickiness

Non-interest demand deposits (DDA) are the cheapest, stickiest source of funding. Banks with DDA >30% of deposits have a durable franchise advantage that shows up in CoF and NIM. At the other extreme, banks where time deposits (CDs) exceed 40% of funding are exposed to hot-money flight when rates shift.

DDA (non-int)MMDA SavingsTime <$250kTime >$250k

Cheapest Funders — Top 20 DDA %

Hot-Money Risk — Top 20 Time-Deposit %

Uninsured Deposit Concentration — Post-SVB Lens

When uninsured deposits exceed 40% of the book, a bank is structurally exposed to a run. QoQ direction matters: shrinking uninsured % signals management is actively de-risking. Expanding uninsured % at banks already above 30% is the pattern that preceded the 2023 regional bank failures.

High-Risk Banks — Uninsured >30% and Trending Up

BankAssets $MUninsured % QoQ Chg ppDeposits $MVuln Tier

Deposit Flight Scoreboard

QoQ deposit outflow crossed with cost-of-funds change. Banks losing deposits AND seeing CoF rise are in a liquidity squeeze — they’re paying up to hold what remains. Banks growing deposits while CoF is flat or falling are winning share at someone else’s expense.

Biggest Losers — QoQ Deposit %

BankDeposits $MQoQ %CoF Chg bpState

Biggest Gainers — QoQ Deposit %

BankDeposits $MQoQ %CoF Chg bpState

Cost of Funds vs Peer Group

Each bank’s CoF is FFIEC’s officially-computed UBPRE013 (Cost of All Interest-Bearing Funds, annualized using average balances). Peer CoF is the median within the bank’s SPC cohort (size + geography + specialty) where the cohort has ≥5 members, otherwise the fleet median. Banks paying >50bp above peer are overpaying for funding — candidates for a treasury-services or deposit-repricing engagement. Banks paying >50bp below peer are franchise standouts worth benchmarking.

Overpayers — CoF >50bp Above Peer

BankCoF %Peer %Gap bpAssets $M

Franchise Standouts — CoF >50bp Below Peer

BankCoF %Peer %Gap bpAssets $M

Rate Sensitivity — Deposit Beta

What differentiates banks is deposit beta — how fast CoF rises with the Fed. For Q1 2026 we proxy deposit beta from deposit-mix stickiness: non-interest DDA% minus time-deposit% of total deposits. Banks at the top quartile (lots of DDA, little time) are low-beta (funding reprices slowly). Banks at the bottom quartile (little DDA, heavy time deposits) are high-beta (rate-sensitive).

Deposit beta is measured relative to the fleet distribution, not an absolute threshold. Earlier Q4 builds used modeled rate-shock CoF bp from FERMI estimates — that field is uniformly “moderate” in the Q1 SOT, so we switched to a directly-observable mix-based proxy.

Deposit Beta — Quartile Distribution

QuartileBanks Median stickiness (DDA% − time%)IQR (p25–p75)Range
Low (best)922−10.8 pp[−17.7, −0.1]−22.8 to +99.3
Mid1,842−36.6 pp[−42.5, −30.6]−48.4 to −22.8
High (worst)922−56.8 pp[−64.0, −52.0]−85 to −48.3 (excl. specialty outliers)
Negative values mean time deposits exceed non-interest DDA — the hallmark of rate-sensitive funding. Low-beta banks have time-heavy or balanced books only slightly off neutral; high-beta banks have time deposits dominating DDA by 50+ percentage points, meaning their cost-of-funds will reprice faster as rates move. Quartile cutoffs are fleet-relative.

Fleet Leaderboard

Vulnerability = how exposed the deposit base is (uninsured concentration, time-deposit reliance, deposit flight, above-peer CoF, wholesale funding).  ·  Franchise = how durable and cheap the deposit base is (DDA %, below-peer CoF, organic growth, low uninsured, low non-core).
Vulnerability: RED YELLOW GREEN
Franchise: GOLD SILVER BRONZE NONE
Bank State Assets $M Deposits $M DDA % Uninsured % CoF % CoF vs Peer bp QoQ % Vuln Vuln Driver Mix (click row for detail) Franchise

← Pick a bank from the list

Click any row in the leaderboard to see vulnerability drivers, peer comparison, funding mix, and risk breakdown for that bank — right here, no scrolling.

What's Driving Each Bank's Vulnerability? — Driver Attribution

Naming note: Previously labeled “DuPont decomposition,” which was technically incorrect. True DuPont decomposes ROE into margin × turnover × leverage. This view is a per-bank driver attribution — it shows how much each of the 5 vulnerability dimensions contributed to a bank’s composite score. Same data, more accurate label.

Total vulnerability score breaks down into 5 independent risk drivers. Two banks can both score 75 RED for very different reasons — one drowning in uninsured deposits, another paying way above peer for funding. The composition determines the remedy. Click any bar to see the bank's full breakdown.

V1 Uninsured (max 25) V2 Time-deposit reliance (max 20) V3 QoQ deposit flight (max 20) V4 CoF above peer (max 20) V5 Non-core funding (max 15)
Bank Vuln Driver mix (click for detail) Primary driver Assets $M
Bar segments are scaled to total composite vuln score for that bank, so width = driver's points contribution. Hover any segment to see exact points + the data tag (e.g. “39% time deposits”).

Geographic Heatmap — Zoom to the Bank Level

Each region is colored by the share of its banks that scored RED on the deposit-vulnerability tier. Cluster bubbles show the bank count and that share at a glance. Zoom in and state fills give way to county fills, then individual bank pins. Switch the metric to avg score for a smoothed view, or franchise for the deposit-quality lens. Hover any region or pin for details; click a pin for full analytics.

 

 
Recomputed avg vuln
Δ from full state
Toggle banks to investigate how each contributes to the state's asset-weighted aggregate. Investigation only — map coloring stays at full-state values.
State colors use the asset-weighted avg score across banks HQ'd in that state (states with <3 banks shown light). County colors use the asset-weighted avg across banks HQ'd in that county. Pin coordinates are the bank's main-office location from FDIC Summary of Deposits. Tip: use mouse wheel or +/- to zoom; drag to pan; click any state to open the bank-list investigation panel.

How this score relates to established frameworks

The Deposit Vulnerability Score adapts the deposit-stress dimension of Basel III's Liquidity Coverage Ratio (LCR) — the same conceptual framework U.S. regulators apply to banks >$50B in assets — and applies it to community banks, where LCR doesn't formally apply. The Fed's CCAR/DFAST stress tests use a similar set of inputs (depositor-stratified run-off rates, wholesale-funding mix, time-deposit reliance) to assess deposit-flight risk under adverse scenarios.

All inputs are standard regulatory measures sourced from FFIEC Call Reports and UBPR:

  • Uninsured deposit % — FFIEC RCONF051/F049 (regulatory disclosure since 2010, became the focal metric after SVB in March 2023)
  • Time-deposit % — FFIEC RC-E line 2.b (RCONF049 + RCONF050)
  • QoQ deposit change — period-over-period delta on RCON2200
  • Cost of funds — FFIEC UBPRE013 (Cost of All Interest-Bearing Funds, FFIEC-annualized)
  • Wholesale / non-core funding — FFIEC RC-M (FHLB advances + other borrowings)

What's specific to Statum is the composite synthesis: which five drivers to combine and how they're weighted. The driver weights (25/20/20/20/15 for V1–V5; 30/25/20/15/10 for F1–F5) are documented below. The Risk Drivers tab provides a per-bank driver attribution that decomposes each bank's score back to its 5 driver primitives so the synthesis is fully auditable against the bank's own call report.

Tier cutoffs — cross-sectional, single-quarter: Rather than fixed score thresholds (the previous approach), tiers are now defined by distance from the fleet median for the current quarter, measured in units of MAD (Median Absolute Deviation). MAD is the robust analogue of standard deviation — it's not pulled around by outliers (the very banks the score is trying to surface).

  • Vulnerability tiers (higher score = more exposed):
    • RED = composite ≥ median + 1.5·MAD — outlier band, roughly the top decile.
    • YELLOW = composite ≥ median + 0.5·MAD — above-typical.
    • GREEN = composite < median + 0.5·MAD — typical or below.
  • Franchise tiers (higher score = stronger franchise):
    • GOLD = composite ≥ median + 1.5·MAD.
    • SILVER = composite ≥ median + 0.5·MAD.
    • BRONZE = composite ≥ median − 0.5·MAD.
    • NONE = composite < median − 0.5·MAD.

The actual numeric cutoffs for this quarter are loaded with the data and printed in the page banner. They will change each quarter as the underlying distribution shifts.

State and county aggregation — asset-weighted: Each state's and each county's number on the heatmap is the asset-weighted mean of the bank vulnerability scores HQ'd there: Σ(score × assets) / Σ(assets). This answers the question "how much money is at risk in this state?" rather than "what does the average community bank in this state look like?" An equal-weighted mean (the previous approach) gave a tiny state with three small banks the same statistical voice as a state with three hundred banks of vastly different sizes — misleading at the geographic level. Asset-weighting matches the convention used by Fed surveillance and standard industry peer benchmarks.

The per-bank vulnerability and franchise scores themselves are not asset-weighted — each bank is scored on its own profile. Only the geographic aggregates (state & county fills on the heatmap, state_vuln_avg in the data) use asset weighting. Cluster bubbles on the map use the same asset-weighted aggregation across the banks within the cluster.

Audit posture: This is a cross-sectional descriptive indicator, not a predictive model. We make no claim that a high vulnerability score causes failure or even predicts it. The score identifies banks that sit at outlier levels of standard funding-risk indicators within the current FDIC Call Report cycle, using a robust statistical band (median ± k·MAD) that's defended by the underlying distribution rather than chosen by judgment. Geographic aggregates are asset-weighted (above). Driver weights are anchored to Basel III LCR run-off rates where regulatory guidance exists (V1 uninsured), otherwise equal-weighted within the residual.
Comparable composite scores in the market (Moody's deposit-funding ratings, S&P deposit scorecards) are proprietary and opaque. We chose transparency over secrecy: the formula is published below, the inputs cite their FFIEC field codes, every per-bank score can be reverse-engineered from the public call report, and the tier cutoffs derive from this quarter's actual distribution rather than fixed levels.

Deposit Vulnerability Score (0–100)

Higher = more exposed. Five weighted drivers. Each driver score is the bank's percentile rank within the fleet on that dimension, multiplied by the driver's max points. Cross-sectional, single-quarter; no fixed thresholds. The composite tier cutoffs are computed separately at each aggregation level (bank / state / county) using median ± k·MAD on that level's own distribution — see the methodology card above for the live numeric cutoffs.

V1
Uninsured concentration — uninsured % of deposits (FFIEC RCONF051/F049). Score = pct_rank × 25 within the fleet (higher % = higher score = more exposed).
0–25
V2
Time-deposit reliance — (time_lt250k + time_gt250k) / deposits. Score = pct_rank × 20 (higher % = more rate-sensitive funding).
0–20
V3
QoQ deposit flight — negative QoQ deposit change. Score = pct_rank(−qoq) × 20 (the more negative the change, the higher the score).
0–20
V4
CoF above peer — bp above the peer-cohort CoF median. Score = pct_rank(gap) × 20. Peer median uses a cohort cascade: SPC peer cohort if ≥5 banks (97% of banks land here), else state-level median if state has ≥5 banks, else no V4 contribution — no fallback to fleet median (the old approach unfairly compared small specialty banks against JPM).
0–20
V5
Non-core funding — (FHLB advances + other borrowings) / deposits. Score = pct_rank × 15.
0–15

Deposit Franchise Score (0–100)

Higher = better. Five weighted drivers, same percentile-rank scoring approach as Vulnerability but inverted for direction (low CoF, low uninsured, low non-core, high DDA, high deposit growth all score higher).

F1
DDA % — non-interest demand deposits / deposits. Score = pct_rank × 30 (higher DDA = stickier, cheaper funding).
0–30
F2
CoF below peer — bp BELOW the peer-cohort CoF median (same cohort cascade as V4). Score = pct_rank(peer_cof − bank_cof) × 25.
0–25
F3
QoQ deposit growth — positive QoQ deposit change. Score = pct_rank × 20.
0–20
F4
Low uninsured reliance — inverse of uninsured %. Score = pct_rank(−uninsured) × 15.
0–15
F5
Low non-core funding — inverse of (FHLB + other) / deposits. Score = pct_rank(−non_core) × 10.
0–10

Inclusion & Data Sources

Scope: US FDIC-insured banks with assets ≥ $100M AND deposits ≥ $50M (Q1 2026 FDIC Call Report). At smaller scale, deposit ratios reflect single-account effects rather than systemic risk. 648 banks excluded by this floor.

Data sources:

  • Uninsured deposit % — FFIEC RCONF051/F049 (regulatory disclosure since 2010, became focal after SVB in March 2023)
  • Time-deposit % — FFIEC RC-E line 2.b (RCONF049 + RCONF050)
  • QoQ deposit change — period-over-period delta on RCON2200 (Q4 2025 → Q1 2026)
  • Cost of funds — FFIEC UBPRE013 (Cost of All Interest-Bearing Funds, FFIEC-annualized). The Q1 2026 SOT extracts round UBPRE013 to 2 decimal places (~1pp precision), so the scorer side-loads a higher-precision derived value computed from raw interest expense / interest-bearing liabilities.
  • SPC peer cohort for V4/F2 — side-loaded from BoardCommand fleet index (112 distinct cohorts, 69 with ≥5 banks). 97% of banks land in their SPC cohort; 3% fall to state-level peer; 0% to fleet-median fallback.
  • Wholesale / non-core funding — FFIEC RC-M (FHLB advances + other borrowings)
  • Bank HQ lat/lon — FDIC Summary of Deposits (June 2024), first-listed branch convention for main office
  • County FIPS — FCC Area API spatial join from HQ lat/lon

Deposit-beta quadrants on the rate-sensitivity tab are derived from fleet quartiles of deposit-mix stickiness (DDA% minus time-deposit%); the irr_deposit_beta_inference field in the Q1 SOT extracts is uniformly "MODERATE_BETA" so cannot be used directly.

Interactive investigation

On the State Heatmap tab, click any state to open a side panel listing every bank HQ'd there (sorted by assets). Toggle individual banks on/off to see how each contributes to the state's asset-weighted aggregate. Live "Δ from full state" shows the recomputed mean. Excluding individual banks is an investigative tool — map coloring stays at full-state values. Useful for stress-testing claims like "Hawaii's number is dominated by First Hawaiian Bank" or "Utah's average reflects Goldman Sachs Bank USA's profile".

What this score is — and isn't

  • Is: A cross-sectional, single-quarter, empirically-derived composite that identifies banks at outlier levels of standard regulatory funding-risk indicators within the current FDIC Call Report cycle.
  • Is not: A predictive model. We make no claim that a high vulnerability score causes failure or even predicts it. No back-testing is published.
  • Anchored to: Basel III LCR deposit-stress framework. Uninsured-deposit weight (25%) matches the Basel III LCR run-off rate for uninsured non-operational deposits. Other driver weights are equal-weighted within the residual.
  • Comparable products: Moody's deposit-funding ratings and S&P deposit scorecards are proprietary and opaque. The Statum approach trades secrecy for transparency — every per-bank score reverse-engineers from public call reports, and every tier cutoff is derived from the current quarter's distribution rather than asserted.