Skip to content
All library documents

Choosing a Parametric VaR Model to Compare with ISDA SIMM

Article Quant Q&A · Author: Randor
Read the original

Summary

The document asks how to build a simple parametric value-at-risk estimate for comparison with ISDA SIMM initial margin. It specifically raises whether the profit-and-loss distribution should be normal or asymmetric, and whether a model should include option vega or gamma. A shifted lognormal distribution is suggested as a possibility, but no model is selected or derived.

The discussion also asks how SIMM can produce different initial margin amounts for counterparties if it uses a normal-distribution assumption, and seeks a mathematical derivation or explanation of the formulas. It provides no answers, calculations, empirical comparisons, or citations, so it does not establish that SIMM uses a normal distribution or explain its margin differences. Its value is as a framing of model design questions; any practical comparison would need further research into SIMM’s sensitivities, risk weights, correlations, and calibration.

Key ideas

  • A parametric VaR comparison with SIMM must specify how option vega and gamma enter the risk estimate.
  • The choice between normal and asymmetric profit-and-loss distributions is raised but not resolved.
  • The document asks how SIMM can produce counterparty-specific initial margin under a normality assumption.
  • No derivation, evidence, or recommended VaR method is provided.

Tags

OptionsVolatilityRisk managementDerivatives pricing
Full text
# what is a simple parametric VaR approach that can be used to compare with ISDA SIMM results?


# what is a simple parametric VaR approach that can be used to compare with ISDA SIMM results?












i am looking for a simple parametric VaR approach that includes vega and/or gamma. i am not sure if to choose a non symmetric distribution for the pnl, what would people use, perhaps some shifted lognormal (shifted so it's mean is zero) ? , or to just choose normal. i think isda simm uses normal , but then, how do they achieve that the IM for each party is different ?

also if anyone has any links to maths that show the derivation for the isda simm formulas, or can offer a simple explanation of them here that would be great to see!

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.