Hidden Vulnerabilities in Consumer Credit
Risk is no longer “in” one product. It’s in the plumbing: funding chains, data gaps, and procyclical risk transfer linking BNPL, unsecured credit, and housing finance.
Thesis
Executive Summary
OverviewConsumer credit risk is increasingly in the plumbing: data opacity, wholesale funding chains, and risk‑transfer structures that couple BNPL, unsecured credit and housing finance into a single procyclical system. The “BNPL is small” narrative misses that BNPL sits at checkout and is funding‑confidence dependent.
On scale, the CFPB BNPL Market Report (published 2025‑12‑03) shows 2023 BNPL originations of $45.2bn across 335.8m loans in its six‑lender sample Confirmed. The same report explicitly notes it cannot observe multi‑lender usage, creating measurement risk and enabling cross‑platform “phantom leverage” Confirmed.
The most under‑priced tail risk is nonbank concentration in consumer receivables. The Financial Times reports private credit firms acquired or committed approximately ~$136bn to consumer debt in 2025 Reported. This imports consumer cyclicality into vehicles marketed as stable private income—while banks often remain the short‑term liquidity providers (warehouse lines, repo, servicing advances) Modeled.
Traditional credit stress is visible in late‑cycle indicators: Fed G.19 shows total consumer credit outstanding of $5.085T (Nov 2025) Confirmed, while TransUnion reports bankcard balances of $1.11T and 90+ DPD delinquency of 2.37% in Q3 2025 Confirmed. ABI reports 565,759 total U.S. bankruptcy filings in CY2025 (+11%) Confirmed.
Housing looks stable at headline level, but stress concentrates in affordability‑sensitive cohorts: MBA reports total mortgage delinquency of 3.99% (Q3 2025), with FHA delinquency at 10.78% Confirmed. MBA explicitly cites rising taxes/insurance costs and “other personal debt obligations” as stressors—consistent with the phantom‑debt mechanism Derived.
- “Risk is migrating from bank balance sheets into less transparent, funding‑sensitive vehicles.” Modeled
- “BNPL is small in share, but large in feedback loops: checkout conversion + funding confidence.” Modeled
- “Housing stress is cohort‑specific (FHA), and it interacts with unobserved short‑term obligations.” Derived
Operating Snapshot
System anchors (latest available)
Confirmed| Metric | Value | Date | Provenance |
|---|---|---|---|
| Total consumer credit outstanding (US) | $5.0848T | Nov 2025 | Confirmed Fed G.19 |
| Revolving credit outstanding | $1.3722T | Nov 2025 | Confirmed Fed G.19 |
| Bankcard balances | $1.11T | Q3 2025 | Confirmed TransUnion |
| Bankcard delinquency (90+ DPD) | 2.37% | Q3 2025 | Confirmed TransUnion |
| Total mortgage delinquency (SA) | 3.99% | Q3 2025 | Confirmed MBA NDS |
| FHA mortgage delinquency (SA) | 10.78% | Q3 2025 | Confirmed MBA NDS |
| 30‑year fixed mortgage rate (PMMS) | 6.10% | 2026‑01‑29 | Confirmed Freddie Mac |
| Total bankruptcy filings (US) | 565,759 | CY2025 | Confirmed ABI |
Regulatory cliffs (cross‑border)
Confirmed| Jurisdiction | Change | Effective | Provenance |
|---|---|---|---|
| Australia | BNPL credit licensing + modified obligations | 10 Jun 2025 | Confirmed ASIC |
| UK | FCA regulation of Deferred Payment Credit (BNPL) | 15 Jul 2026 | Confirmed FCA CP25/23 |
| EU | CCD2 (Directive 2023/2225) applies | 20 Nov 2026 | Confirmed Eur‑Lex |
| US | CFPB BNPL Market Report published | 03 Dec 2025 | Confirmed CFPB |
| US | EWA advisory opinion: “covered” EWA not credit under Reg Z (criteria‑based) | 23 Dec 2025 | Confirmed Federal Register |
Market Topology
Where the risk actually sits
ConceptualScenario Stress Test (Illustrative)
Annualized loss mapping (BNPL + bankcards + mortgages)
Stylized| Metric | Scenario A Soft landing |
Scenario B Consumer recession |
Scenario C Credit crisis |
Expected |
|---|---|---|---|---|
| Probability Modeled | 40% | 35% | 25% | 100% |
| BNPL losses (annualized) Modeled | $1.36B | $2.71B | $4.52B | $2.62B |
| Bankcard charge‑offs (annualized) Modeled | $38.9B | $61.1B | $83.3B | $57.8B |
| Mortgage credit losses (annualized) Modeled | $9.55B | $25.6B | $61.4B | $28.1B |
| Total losses (sum) Derived | $49.8B | $89.4B | $149.2B | $88.5B |