Attributing Expected Loss Changes to Risk Metric Movements

Summary

The document addresses how changes in a portfolio or risk measure might be traced to movements in its underlying risk inputs. It uses expected loss as a simplified stand-in for risk-weighted assets and represents expected loss as the sum of exposure at default (EAD) multiplied by probability of default (PD) and loss given default (LGD) across granular observations. The cited paper describes a way to apportion changes in the aggregate measure among changes in those components.

This offers a framework for investigating whether shifts in exposure, default likelihood, or loss severity drove an observed change. The source does not spell out the attribution formulas or provide empirical results, and it cautions by implication that expected loss is only a simple proxy for RWA. It therefore introduces a useful attribution concept rather than a full account of RWA behavior across market conditions.

Key ideas

• Expected loss can be represented as the granular sum of exposure times default probability times loss severity.
• Changes in the aggregate can be apportioned among changes in its component risk metrics.
• Expected loss is presented as a simplified proxy for risk-weighted assets, not as an identical measure.

Tags

# Want to understand the links and relationship between all the risk metrics?


# Want to understand the links and relationship between all the risk metrics?












For Example : if Risk weighted asset (RWA) increased or decreased this month, which other risk metrics could have influenced RWA to increase or decrease. Also in different situations like, upward trend, sudden increase, downward trend, sudden decrease, neutral or stable RWA. I want to understand and trying to create a document for my reference and learning purpose.

## Answer by Clive (score 1)

https://quant.stackexchange.com/a/49466

Our paper answers this question in the admittedly simple context of Expected Loss (EL) calculated as the straightforward summation of EAD * PD * LGD from the granular level.

EL here is the surrogate for your RWA, and the risk metrics are the granular data (EAD, PD, LGD).

The paper proposes how changes in EL can then be attributed / apportioned to changes in each of the risk metrics.

Hunt, C.; Taplin, R. Aggregation of Incidence and Intensity Risk Variables to Achieve Reconciliation. Risks 2019, 7, 107. https://www.mdpi.com/2227-9091/7/4/107

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.