Exam Feedback May 2018 Part 2 Exam Feedback

kiufrm

New Member
hope someone can help clarify my understanding of VaR.

to find 10 days VaR of portfolio, do I find VaR of portfolio then find 10-day VaR or do I find 10-day VaR of individual assets and get 10-day VaR of portfolio?

thank you in advance.
 

terrance00232

New Member
hope someone can help clarify my understanding of VaR.

to find 10 days VaR of portfolio, do I find VaR of portfolio then find 10-day VaR or do I find 10-day VaR of individual assets and get 10-day VaR of portfolio?

thank you in advance.


I remember that practice exams have similar questions and the method might be the first one. To calculate the 1-day portfolio VaR first and convert it to 10-day portfolio VaR.
 

Bernardo

New Member
I remember that practice exams have similar questions and the method might be the first one. To calculate the 1-day portfolio VaR first and convert it to 10-day portfolio VaR.
For normal parametric VaR, the two approaches are equivalent. For lognormal (also parametric) VaR, I would apply the square root law on the vol.
 
For normal Parametirc Var, both methods will result in the same values. However for Log Normal its different. I remember there was question in the paper asking for calculating 10 Day Var through Log Normal method. We were given annual values. 1. I tried to calculate annual Var and convert it into 10 days by Dividing by Sqrt250 * Sqrt10 but i didnt get the answer.2 I converted annual values to daily one and calculate one day var and convert into 10 days by Multiplying by Sqrt10....I got the answer.
 

oldfed

Member
For normal Parametirc Var, both methods will result in the same values. However for Log Normal its different. I remember there was question in the paper asking for calculating 10 Day Var through Log Normal method. We were given annual values. 1. I tried to calculate annual Var and convert it into 10 days by Dividing by Sqrt250 * Sqrt10 but i didnt get the answer.2 I converted annual values to daily one and calculate one day var and convert into 10 days by Multiplying by Sqrt10....I got the answer.[/QUO

Hi bilal, Terrance, Kiufrm, Bernardo

I struggled with that one too but for another reason! ;-)

For your issue, I agree with Bernardo. In my opinion, for the order of computation, when facing a lognormal VaR, which is non-linear function [With a computation of e(kx) different from ke(x)], you always have to scale the volatility inside the function and in that case you get a 10d VaR directly by adjusting volatility in the function [dividing annual volatility by SQRT (250) / SQRT (10) = SQRT (25) = 5].

However, my problem, when facing a single Lognormal VaR in a question (no comparison with Normal VaR, just one Lognormal VaR to compute) is the use of the mean. I know, that in GARP papers, computation of a single Normal VaR always ignores the mean of the distribution. But in the case of a single Lognormal VaR computation, I have to admit that I still do not have the answer! I did not look for it enough in fact ;-).

And good future week-end to everyone.
 

Karim_B

Active Member
Subscriber
Hi bilal, Terrance, Kiufrm, Bernardo

I struggled with that one too but for another reason! ;-)

For your issue, I agree with Bernardo. In my opinion, for the order of computation, when facing a lognormal VaR, which is non-linear function [With a computation of e(kx) different from ke(x)], you always have to scale the volatility inside the function and in that case you get a 10d VaR directly by adjusting volatility in the function [dividing annual volatility by SQRT (250) / SQRT (10) = SQRT (25) = 5].

However, my problem, when facing a single Lognormal VaR in a question (no comparison with Normal VaR, just one Lognormal VaR to compute) is the use of the mean. I know, that in GARP papers, computation of a single Normal VaR always ignores the mean of the distribution. But in the case of a single Lognormal VaR computation, I have to admit that I still do not have the answer! I did not look for it enough in fact ;-).

And good future week-end to everyone.
Hi @oldfed
One of the Schweser guys mentioned the following rules in terms of the GARP FRM questions and the use of the mean in VaR calculations:

1) For Lognormal VaR always use the mean.

2) For Normal VaR don't use the mean unless it's one of the following scenarios:
a) Comparing Normal VaR to Lognormal VaR
b) Calculating Surplus at Risk

Let's see what @David Harper CFA FRM thinks when he's back from content planning duty.

Best
Karim
 

oldfed

Member
Hi @oldfed
One of the Schweser guys mentioned the following rules in terms of the GARP FRM questions and the use of the mean in VaR calculations:

1) For Lognormal VaR always use the mean.

2) For Normal VaR don't use the mean unless it's one of the following scenarios:
a) Comparing Normal VaR to Lognormal VaR
b) Calculating Surplus at Risk

Let's see what @David Harper CFA FRM thinks when he's back from content planning duty.

Best
Karim

Ok! Thank you very much Karim.

You're right, let's wait for feedback.

Have a good week-end
 

kaval

Member
Hi bilal, Terrance, Kiufrm, Bernardo

I struggled with that one too but for another reason! ;-)

For your issue, I agree with Bernardo. In my opinion, for the order of computation, when facing a lognormal VaR, which is non-linear function [With a computation of e(kx) different from ke(x)], you always have to scale the volatility inside the function and in that case you get a 10d VaR directly by adjusting volatility in the function [dividing annual volatility by SQRT (250) / SQRT (10) = SQRT (25) = 5].

However, my problem, when facing a single Lognormal VaR in a question (no comparison with Normal VaR, just one Lognormal VaR to compute) is the use of the mean. I know, that in GARP papers, computation of a single Normal VaR always ignores the mean of the distribution. But in the case of a single Lognormal VaR computation, I have to admit that I still do not have the answer! I did not look for it enough in fact ;-).

And good future week-end to everyone.
..........................................................................................

I differ with this. and agree with Bilal. first, we need to find one day var by calculating one-day volatility. which is annual vol/ 250 (not divided by sqrt of 250) and then calculate 10day var by this formula: one day var * Sqrt(10)/Sqrt(250). i got the ans.

thanks
Kaval
 

kaval

Member
guys what's the message you are getting if you try to sign in on the registration page? r we all getting the same message or there is a difference
 

kaval

Member
it's the same, somehow I entered the ERP registration page. so relax. I know next few days are going to be restless. let's see what's there in future. :)
 

terrance00232

New Member
by the way, the message was "There are no exams available to register for. Please check back soon".

Hi, Kaval. I don't really understand what this message means.
Is it possible that anyone else will get different messages from the registration page?
 

geotheox

New Member
Mine states that "You will need to pass the FRM Exam Part II by May 2021. Otherwise you will have to re-enroll in the FRM Program as a new candidate."
 
Top