Loss Given Default and Recovery Rates
When a borrower defaults, the lender rarely loses everything. Loss Given Default (LGD) measures the fraction of exposure that is actually lost after accounting for recoveries through collateral liquidation, guarantees, and workout processes. As a complement, the recovery rate (RR) = 1 − LGD.
LGD is a fundamental input to credit risk models and, alongside Probability of Default (PD) and Exposure at Default (EAD), determines the Expected Loss (EL = PD × LGD × EAD) that anchors regulatory capital calculations.
Types of LGD
Workout LGD: Based on the actual cash flows recovered during the workout process, discounted back to the default date:
LGD = 1 − [Σ (Recoveries_t − Costs_t) / (1 + r)^t] / EAD
Where recovery cash flows are discounted at an appropriate discount rate (often the contract rate or a risk-adjusted rate) and costs include legal, administrative, and collection expenses.
Market LGD: Based on the market price of defaulted debt shortly after default. If a defaulted bond trades at 40 cents on the dollar, the implied LGD is 60%.
Implied Market LGD: Derived from credit spreads or CDS prices of performing debt:
CDS spread ≈ PD × LGD → LGD ≈ CDS spread / PD
Key Drivers of Recovery Rates
| Factor | Impact on Recovery |
|---|---|
| Seniority | Senior secured > Senior unsecured > Subordinated > Equity |
| Collateral quality | High-quality, liquid collateral improves recovery |
| Industry | Utilities and telecoms recover more than tech startups |
| Economic conditions | Recovery rates fall during recessions (downturn LGD) |
| Jurisdiction | Creditor-friendly legal systems yield higher recoveries |
| Time in default | Longer workouts erode recovery through costs and asset deterioration |
| Debt cushion | More subordinated debt below you = higher recovery |
Historical Recovery Rate Data
Moody's and S&P publish extensive recovery rate statistics:
| Instrument Class | Average Recovery Rate | LGD |
|---|---|---|
| Senior Secured Bank Loans | ~70-80% | 20-30% |
| Senior Secured Bonds | ~55-65% | 35-45% |
| Senior Unsecured Bonds | ~35-45% | 55-65% |
| Subordinated Bonds | ~25-35% | 65-75% |
| Junior Subordinated | ~15-25% | 75-85% |
These averages mask significant variation — recovery rates have a bimodal distribution, clustering around either very high (near full recovery) or very low values.
Downturn LGD
Basel II/III requires banks using the Advanced IRB approach to estimate downturn LGD — the LGD that would prevail during economic stress. This requirement exists because:
- Default rates and LGD are positively correlated — more borrowers default when economic conditions are poor, and recovery rates simultaneously fall
- Using through-the-cycle average LGD would understate risk during downturns
- The correlation between PD and LGD amplifies portfolio losses beyond what independent models predict
Methods for estimating downturn LGD:
- Mapping to macroeconomic variables — Regress historical LGD on GDP growth, unemployment, asset prices
- Conservative add-on — Add a fixed buffer to long-run average LGD
- Worst-observed period — Use LGDs from the most severe historical downturn
LGD in Basel Capital Framework
Under the IRB approach:
Foundation IRB (F-IRB): Supervisory LGD values prescribed by regulators:
- Senior secured: 45% (with eligible collateral adjustments)
- Senior unsecured: 45%
- Subordinated: 75%
Advanced IRB (A-IRB): Banks estimate their own LGD using internal models, subject to:
- Downturn LGD requirements
- Minimum data history (typically 7+ years covering a downturn)
- Regular backtesting and validation
LGD Modeling Approaches
Parametric Models:
- Beta distribution fitting (bounded between 0 and 1)
- Regression models using collateral type, seniority, industry, and macro variables
- Tobit models to handle boundary observations
Non-Parametric Approaches:
- Look-up tables by segment (seniority × collateral × industry)
- Machine learning methods (random forests, gradient boosting)
- Nearest-neighbor approaches based on comparable defaults
Practical Challenges
Data Scarcity: Defaults are relatively rare events, and recovery data takes years to collect (workouts average 2-3 years). Small sample sizes make statistical estimation challenging.
Definition Consistency: When does "default" start? What discount rate is appropriate? How are costs allocated? Different definitions produce different LGD estimates.
Collateral Valuation: Collateral values are uncertain and may decline precisely when they are needed most (downturn correlation). Stress testing collateral values is essential.
FRM Exam Relevance
LGD is tested heavily in FRM Part 1 (foundations of risk) and Part 2 (credit risk):
- Expected Loss formula (EL = PD × LGD × EAD)
- Workout LGD calculation with discounted cash flows
- Drivers of recovery rates (seniority, collateral, economic cycle)
- Downturn LGD requirements under Basel IRB
- Foundation vs Advanced IRB LGD treatment
- Relationship between LGD and credit spreads