How should a bank calibrate downturn LGD, and why does Basel require it instead of average-cycle LGD?

Question

AcadiFi · Accepted Answer

**Downturn LGD** is the loss given default estimate calibrated to reflect economic conditions during periods of elevated credit losses. Basel requires it because using average-cycle LGD would understate capital needs precisely when banks need capital most.

**Why Downturn LGD Matters**

During economic downturns, three factors compound to increase LGD simultaneously:

1. **Default rates rise** — more borrowers default (PD increases)
2. **Recovery rates fall** — collateral values decline, secondary markets freeze
3. **Workout periods lengthen** — courts are backlogged, buyers scarce

This positive correlation between PD and LGD creates a **double-hit** effect: banks face more defaults AND higher losses per default at the same time.

**Calibration Approaches**

```mermaid
flowchart TD
    A["Downturn LGD Estimation"] --> B["Approach 1: Historical Worst Period"]
    A --> C["Approach 2: Regression-Based"]
    A --> D["Approach 3: Haircut Method"]
    B --> E["Take LGD from worst 12-month window"]
    C --> F["Regress LGD on macro variables, forecast at stressed values"]
    D --> G["Apply supervisory haircuts to average-cycle LGD"]
```

**Approach 1: Historical Worst Period**

Identify the 12-month window with the highest average LGD in your data. For Oakmont Federal's commercial mortgage portfolio:

| Period | Average LGD | Notes |
|---|---|---|
| 2005-2006 | 18% | Pre-crisis boom, high recoveries |
| 2007-2008 | 31% | Initial crisis |
| 2009-2010 | **52%** | Worst period — property values collapsed |
| 2011-2012 | 38% | Gradual recovery |
| 2019-2020 | 29% | COVID (quick recovery in CRE) |

**Downturn LGD = 52%** (the worst 12-month period)

**Approach 2: Regression-Based**

Build a model relating LGD to macroeconomic variables:

**LGD_t = alpha + beta_1 x GDP_growth_t + beta_2 x Unemployment_t + beta_3 x Property_Index_t + epsilon**

Then forecast LGD using stressed macro values (e.g., GDP -3%, unemployment 10%, property index -25%).

This approach is more forward-looking and captures scenarios not yet observed historically.

**Approach 3: Supervisory Haircut**

Some jurisdictions prescribe a minimum uplift:

**Downturn LGD = max(Average LGD x 1.25, Average LGD + 10%)**

For average LGD of 35%:
- Method A: 35% x 1.25 = 43.75%
- Method B: 35% + 10% = 45%
- **Downturn LGD = 45%**

**Practical Considerations:**
- Data quality is critical — recovery data must be resolved to economic cycle periods
- Discount rates for recovery cash flows should also reflect stressed conditions
- Workout costs typically increase in downturns (longer legal processes, fire-sale discounts)
- The EBA (European Banking Authority) has specific guidelines on mapping economic downturns

**Exam Tip:** FRM questions may present average-cycle and downturn LGD estimates and ask you to calculate the difference in capital requirements. The downturn uplift can be 30-80% of average LGD depending on the collateral type.

For more on credit risk modeling, check our FRM Part II question bank.

How should a bank calibrate downturn LGD, and why does Basel require it instead of average-cycle LGD?

Master Part II with our FRM Course

Related Questions

Practice Questions