## A Quantitative Approach to Central Bank Haircuts and Counterparty Risk Management

_IMF Working Papers, October 31, 2025_

## Source details

**Canonical URL:** [A Quantitative Approach to Central Bank Haircuts and Counterparty Risk Management](https://www.imf.org/en/publications/wp/issues/2025/10/31/a-quantitative-approach-to-central-bank-haircuts-and-counterparty-risk-management-571472)

## Other formats

- [Markdown version](/en/publications/wp/issues/2025/10/31/a-quantitative-approach-to-central-bank-haircuts-and-counterparty-risk-management-571472/index.md)
- [Structured JSON version](/en/publications/wp/issues/2025/10/31/a-quantitative-approach-to-central-bank-haircuts-and-counterparty-risk-management-571472/index.json)
- [Bundle manifest](/en/publications/wp/issues/2025/10/31/a-quantitative-approach-to-central-bank-haircuts-and-counterparty-risk-management-571472/bundle-manifest.json)

## Bibliographic details
- Authors: Yuji Sakurai
- Published: October 31, 2025
- Series: IMF Working Papers
- DOI: https://doi.org/10.5089/9798229026819.001

---

### Framework for haircut setting
- Introduces four haircut model types tailored to asset characteristics—marketable or non-marketable—and data availability.
- Proposes a novel model for setting haircuts in data-limited environment using a satallite cross-country model.
- Key principles guiding haircut calibration:
  - non-procyclicality
  - data-drivenness
  - conservatism
  - avoidance of arbitrage gaps
- Model inputs detailed:
  - Value-at-Risk (VaR) percentiles
  - volatility measures
  - time to liquidation

### Quantifying residual uncollateralized exposures
- Proposes a quantitative framework for estimating expected uncollateralized exposures that remain after haircut application.
- Emphasizes the importance of expected uncollateralized exposures in stress scenarios.
- Illustrative simulations demonstrate how volatility impacts exposure using dynamic Nelson-Siegel yield curve models.

### Machine learning for validation and stress simulation
- Explores the use of Variational Autoencoders (VAEs) to simulate stress scenarios for bond yields.
- VAEs trained on U.S. Treasury data capture realistic yield curve distributions and offer an alternate tool for validating VaR-based haircuts.
- Notes and caveats:
  - Interpretability and explainability remain concerns for machine learning models.
  - Machine learning models can enhance risk assessment by uncovering potential model vulnerabilities.

---

## Content in this bundle

- **Working Paper**
  - [Working Paper (Markdown version)](/-/media/files/publications/wp/2025/english/wpiea2025225-source-pdf.pdf.md){rel="alternate" type="text/markdown"}
  - [Working Paper (PDF)](/-/media/files/publications/wp/2025/english/wpiea2025225-source-pdf.pdf){rel="external" type="application/pdf"}

---

_Source: https://www.imf.org/en/publications/wp/issues/2025/10/31/a-quantitative-approach-to-central-bank-haircuts-and-counterparty-risk-management-571472_