[R-SIG-Finance] VaR and ES in PerformanceAnalytics

[Brian G. Peterson]

On Mon, 2011-10-24 at 14:20 -0400, financial engineer wrote:
> I appreciate your response and the clarification. I shall ponder over
> it.
> 
> Meanwhile, rather than sending a data file, I am attaching the code I
> ran to generate R.MCO which I used in the calcs. below
> 
> MCO = get.hist.quote("MCO", start = "2010-01-04", end = "2011-10-17",
> quote = "AdjClose", compression = "d")
> R.MCO = Return.calculate(MCO, method="compound")
> R.MCO = as.xts(R.MCO)
> 
> I'd be keen to read your specific response. 

I've attached the output of chart.VaRSensitivity on the MCO data.

You can see how the modified VaR tracks to the historical VaR very well,
much better than the Gaussian approximation.  

You can also see where the modified ES breaks down, at around 98%, and
starts climbing towards zero.  The operational assumption will return
the modified VaR.

You can also see that at high probability levels modified ES will give
larger loss estimates than the historical.  I tend to think conservative
risk estimates are a good thing, but others have opinions that differ on
the desirability of this.

I previously gave my rationale for using lower p values than p=.99 with
daily or lower frequency data.

Regards,

  - Brian

-- 
Brian G. Peterson
http://braverock.com/brian/
Ph: 773-459-4973
IM: bgpbraverock
-------------- next part --------------
A non-text attachment was scrubbed...
Name: MCO.VaRSensitivity
Type: application/pdf
Size: 17595 bytes
Desc: not available
URL: <https://stat.ethz.ch/pipermail/r-sig-finance/attachments/20111024/d63b6c4e/attachment.pdf>