VAR
- Thread starter babu
- Start date
babu,
First I would say that our cirriculum, and recent history, warns us to be careful about assuming our VaR really does "give consideration of all the risk categories like credit risk / market risk/ counterpar party risk and arriving at cumulative VAR."
I am thinking here of Basel's guidelines for incremental risk in the trading book. The way i look at that: they realized market risk VaR didn't cover all the risks that even contribute market risk VaR, so they added on a charge (including to reflect liquidity risk). http://www.bis.org/publ/bcbs141.htm
Okay, but otherwise, my answer here would depend on (work backwards from) the assumptions that underlie VaR, including its limitations. Low VaR implies:
* lower loss threshold; i.e., we expect to lose at least this much (significance)% of the time
* low confidence
* short time horizon
* small sample
* mean reversion (in scaling over multiple periods, mean reversion is negative autocorrelation and will lower the VaR)
* Manifest model risk (omitted risk factors; contradicting assumption above)
* Ill-specified distribution (parametric)
* Inappropriate history (historical sample is too moderate, lacks black swans, etc)
David
First I would say that our cirriculum, and recent history, warns us to be careful about assuming our VaR really does "give consideration of all the risk categories like credit risk / market risk/ counterpar party risk and arriving at cumulative VAR."
I am thinking here of Basel's guidelines for incremental risk in the trading book. The way i look at that: they realized market risk VaR didn't cover all the risks that even contribute market risk VaR, so they added on a charge (including to reflect liquidity risk). http://www.bis.org/publ/bcbs141.htm
Okay, but otherwise, my answer here would depend on (work backwards from) the assumptions that underlie VaR, including its limitations. Low VaR implies:
* lower loss threshold; i.e., we expect to lose at least this much (significance)% of the time
* low confidence
* short time horizon
* small sample
* mean reversion (in scaling over multiple periods, mean reversion is negative autocorrelation and will lower the VaR)
* Manifest model risk (omitted risk factors; contradicting assumption above)
* Ill-specified distribution (parametric)
* Inappropriate history (historical sample is too moderate, lacks black swans, etc)
David
Hi babu, can you be more specific, "offset" could imply different things to me (e.g., double-counting offsets), the word doesn't have a set formula in my understanding? Do you have a context?
Hi babu - I am not sure i see how that's a different problem than portfolio VaR (aggregation), of course with currency and country-specific exposures. The nearest we get to this (in the FRM) is Jorion Chapter 7. It's not efficient for me to recast various portfolio aggregation techniques, can you please refer to Jorion Chapter 7. I am glad to help but i would be all day rephrasing source material on this one..thanks, David
Similar threads
