Forums  > Risk Management  > Component (Historical) VaR ??  
     
Page 1 of 1
Display using:  

jcarterdm


Total Posts: 12
Joined: Mar 2014
 
Posted: 2015-04-23 15:46

Hello everyone,

Is there any way to compute component VaR (how a portfolio's VaR is affected by the marginal risk added by certain transaction/asset class/sector/unit/whatever) using a historical simulation approach of VaR?

Any guidance will be appreciated

Thanks!


mmport80


Total Posts: 85
Joined: Jul 2010
 
Posted: 2015-04-24 05:53
IIRC, do a numerical calculation.

I.e. bump up your holding in bonds for example, note the change. Do the same for equities, and so on.

Recalculate your historical VaR after each bump.

Then take the numerical derivative: D_Risk/D_Holding_Change.


--- http://johnorford.blogspot.com http://blog.johnorford.com

jcarterdm


Total Posts: 12
Joined: Mar 2014
 
Posted: 2015-04-24 13:49
Thank you. That was my first idea, and sounds fine to me, but I wanted to be sure that it's conceptually "correct" (or at least coherent). I also thought of fitting a parametric VaR scaled by some factor "a". "a" would have a value so this parametric VaR is equal to the previously computed historical VaR. Then use the analytical formula of Component VaR. How about that?

mmport80


Total Posts: 85
Joined: Jul 2010
 
Posted: 2015-04-24 15:58
Your suggestion is mixing apples and oranges, prob not best to do that.


--- http://johnorford.blogspot.com http://blog.johnorford.com

pj


Total Posts: 3343
Joined: Jun 2004
 
Posted: 2015-04-24 16:16
please explain me how comparing apples and oranges...

OFFENDERS WILL BE TERMINATED

agentq


Total Posts: 25
Joined: Jul 2008
 
Posted: 2015-05-08 05:07
This is a decent reference for a kernel approach


Cracking VaR with kernels
Previous Thread :: Next Thread 
Page 1 of 1