Backtesting of VAR and condityional Var

[amal]

hi david , hi everyone
I'm new here and i need some help,
i'm working on risk mesures ( VAR and ES )
1/ i want to know please how to calculate thes values using models such as ARCH/GARCH...
2/ Is there any way to backtest VAR an ES??

pleaaaase help me
thank you in advance

 

[amal]

pleaaaaaaaaaaaaaaaase any one here to help me ((

 

[David Harper CFA FRM]

Hi amal, your question is (very) broad and encompasses, in my opinion, just so much stuff i'm not sure where to begin. My favorite text on this is
http://www.amazon.com/Market-Analysis-Value-Models-Finance/dp/0470997885
... much of this enitire 4th volume is devoted exactly to these questions and their many answers.

i'm not just throwing you a random link, i've read her entire series and many other books in the categories, this book has several chapters, with easily accessible corresponding worksheets, that show exactly how to perform these tasks.

re: backtest VaR: exceeding the VaR is typically characterized as a Bernoulli (eg., Basel)--i.e., the daily loss either exceeds VaR or not--such that if we assume i.i.d., a binomial distribution can be used to backtest a VaR. Very straightforward. See Jorion VaR 3rd Chapter 6

If you have something more specific, feel free to ask for help or share something too, thanks,

 

[amal]

Thank you david very much for your answer, i'll read the link you sent to me,
and just more thing please can we backtest the ES??
thanks again

 

[David Harper CFA FRM]

Hi amal, sure thing. Yes, according to Carol Alexander (i'm pretty sure her ETL = ES = conditional VaR for continuous distributions) in 6.4.5. of same book above, backtest of ETL is possible:

The test is based on the observation that, if the process dynamics are correct and ETL is an unbiased estimate of the expectation in the tail below the VaR, the standardized exceedance residuals should behave as a sample from an i.i.d. zero mean process. The null hypothesis is that εt has zero mean, against the alternative that the mean is positive, since it is a positive mean that suggests that the ETL is too low, and underestimation of the ETL is what we want to guard against. So the test statistic is ... page 344, MRA vol IV, CA

 

[amal]

thank you David so much i'll try to backtest it with MATLAB and tell if it's OK or not,
any way good luck in youre job and thanks again