# default rate - Snapshot
*2026-06-30 · scientific concept (credit risk) · 7 sources · light*
↪ [[default rate - Reading]]
> "Default Rate is calculated as the amount in default over the past 12 months divided by the total outstanding volume of loans not in default at period start."
Trust legend: ✅ official · ◽ independent · ⚠️ vendor · 💬 opinion
---
## What it is
**Default rate:** realized (ex-post) % of a portfolio that defaulted in a period.
Source: BIS/Fed ✅
**Don't confuse with:** Probability of Default (PD) — forward-looking, ex-ante model estimate.
### Glossary
| Term | Definition |
|------|------------|
| Default rate | Realized % defaulted in period |
| PD | Forward estimate of default likelihood |
| Default (Basel) | 90 days past due OR unlikely to pay |
---
## Where it lives
- Bank regulatory capital (Basel IRB)
- Bond/loan rating agency forecasts (Moody's, S&P, Fitch)
- ML credit-risk modeling (Lending Club, Kaggle datasets)
**How it works:** defaults ÷ total outstanding at period start. Cohort method (fixed group over time) or rolling 12-month window. Threshold: commonly 90 or 270 days past due — not fixed universally.
Source: BIS, Wall Street Oasis ✅/⚠️
---

Delinquency Rate, All Loans, All Commercial Banks (live official series)
Source: FRED / St. Louis Fed ✅
---
## Alternatives / before
| Metric | What it is | Diff from default rate |
|--------|-----------|------------------------|
| Delinquency rate | 30+ days past due | Earlier stage, loan still active |
| Charge-off rate | Written off as loss, net recoveries | Later stage, realized loss |
| PD | Model estimate | Ex-ante vs default rate's ex-post |
Progression: delinquency → default → charge-off.
---
## Market & players
| Player | What they do | Tag |
|--------|-------------|-----|
| Moody's | Issuer forecasts, private credit (spec-grade 1.7–8.3% range 2026) | ⚠️ |
| S&P Global | Trailing spec-grade rate: US 4.4%, EU 3.5% | ⚠️ |
| Fitch | Leveraged loan/HY ranges, US vs EU split | ⚠️ |
| FRED (Fed) | Official charge-off/delinquency series, free | ✅ |
Moody's/S&P/Fitch = ~95% of global rating market. Numbers diverge by methodology — no single "the" default rate.
---
## Rules
| Body | Document | Year | What it requires |
|------|----------|------|-----------------|
| BIS/Basel Committee | Basel Framework CRE36 (IRB) | Basel II→III | PD = long-run avg of 1yr default rates per grade; ≥7 grades, annual review, 3yr use test |
*not found: none*
---
## Adjacent concepts
| Concept | Distinction |
|---------|-------------|
| LGD | % of exposure lost after recovery |
| EAD | Outstanding balance at moment of default |
| EL | EL = PD × LGD × EAD — default rate feeds only PD |
Common error: confusing PD with credit spread (spread also prices liquidity + risk aversion).
---
## What to touch
| Type | Link | Notes |
|------|------|-------|
| Live series | [FRED DRALACBN](https://fred.stlouisfed.org/series/DRALACBN) | Official, free, downloadable |
| Dataset | Lending Club (Kaggle) | 30k loans, 22% default rate |
| Dataset | Give Me Some Credit (Kaggle) | Classic PD-modeling set |
---
## What research says
| Paper | Year | Key finding | Class |
|-------|------|-------------|-------|
| Alvi, Arif, Nizam — systematic review | 2024 | LightGBM/XGBoost trending above logistic regression (70–85% baseline) | ⚠️ empirical |
**Consensus:** emerging — ML edges ahead in research, but interpretable logistic regression still dominates regulatory IRB models in practice.
---
## Next: explore
- **Loss Given Default (LGD)** — other half of Expected Loss
- **Basel III IRB calibration** — long-run average → regulatory capital
- **PD modeling (logistic regression vs ML)** — turns history into forecast
*Run `/penguru:snapshot <concept>` on any of these*
---
## Sources & trust audit
| Type | Count | Examples |
|------|-------|---------|
| ✅ Official/regulator | 4 | Fed, BIS, FRED |
| ◽ Independent | 2 | Wikipedia, PMC |
| ⚠️ Vendor | 1 | Moody's |
| 💬 KOL opinion | 0 | — |
| **Total** | **7** | |
**Gaps ("not found"):** none
**Dead links dropped:** wallstreetoasis.com, numberanalytics.com (403); openriskmanual.org x2 (login-gated)