Allen, Understanding Market, Credit and Operational Risk: The Value at Risk Approach

RenInvs

New Member
Subscriber
Hi i have the next questions-reading about Chapter 2&3 de Allen:

1. Could you explain the difference between forward based and model based? adn why MA not model heavy tail? pg 11
2. In GARCH model, why it said unconditional heavy tails?

3.- What's the relationship between GBM-> BSM and MA?

4. What is the difference between GARCH and ARCH?

5.When MA supposed mean r=0 for short time, MA could be OK if i compute return with more space?

6.What is the difference between regressive and contiional??

7. Why drawback is non linearity?

8.why the decay of rismetrics are inconsistent? and robus at the same time?

9.Why GARCH and EWMA are recursive?

10 .How do you get the formulae to compute the weight from HYBRID model??

11.Could you explain more the Kernel Function by factor? i understand that is associted to economic factor or event, but i can't see in the formulae?

12.The graph in HYBRID method. The x axis should be 7,9,16, ... and not 1,2,3..right? I want to be sure that is a typo.

13./Comparing Dowd and Jorion, why Jorion choose an weight more conservative and Dowd is more relaxed? exit an explanation?

14.When the reading present the volatility approach, i could see the historical simultaion and hybrid approach more related with how to take the tail and not the volatility as we could see in GARCH, MA and EWMA?

15. In the page 20, we compare the acum weight between HS and Hybrid Weight. The first one is more conservative than Hybrid weight, so why i shoul take the second one? If you coul explain it. Ill appreciate.

Thanks a lot!!
 

brian.field

Well-Known Member
Subscriber
These questions are a bit too non-specific to adress in my opinion. I would suggest trying to ask more specific questions that are more aligned with the AIMS. I doubt any of the above questions would be asked on part 1 (or part 2). Just my initial thoughts.....
 

RenInvs

New Member
Subscriber
Maybe i should be more specific,but if you prepare just for the exam you won't be prepared for the reality, just to solve questions from a test.
 

ShaktiRathore

Well-Known Member
Subscriber
May be i can answer a few,
Your Q 13 is very interesting i like it. See you are trying to fit discrete values with continous,Var is based on continous distribution aapproach while we are dealing with discrete numbets here.if we assume p/l data instead of continous data a discrete distn like binomial comp. Apples to apples instead of oranges,we see that jorion var as var inside the tail instead of to it(95.5% insteaad of 95%) whereas dowd would licate the var adjacent to 95% var we desire(it locates 94.5% var). Both are adjacent to actual var but Dowd Var seems more appropriate first its not licating var inside the 5% tail which seems more conservative,second we round Dowd Var of 94.5% to get 95% var seems more closer to 95% var as compared to jorion which rounds to 96% var far from actual var we desire. Acc. To garp both are valid Vars.
9)both garch and ewma take into account past volatility putting more weight on most recent,in this way finding future volatility based on past values in a recursive way,both models are based on historical volatilities with a recursive formula

4)Arch is estimating volatility with a param lambda while garch is a more generalised version of arch introducing w the long term variance component which is absent in arch along with alpha and beta
6)regressive means tending towards mean,regressive models give mean values of variables based on past values while onditional models like arch derive their values from conditional on valu of some variable e e.g. Conditional heteroskedasticity leds to variable dependent on some other variable like in arch current volatility is conditioned to previous volatility value,hence its name conditional
3) Gbm has constant drift rate and a random shock component while ma assumes variable movg average not a constant average with no shock.
Thanks
 
Last edited:
Top