Exam Feedback November 2016 Part 2 Exam Feedback

[frmpart2dan]

i think 48 is a fair guess (60%) but the lower bound can be 40 if it is a really hard paper...

60% is a bit stretch for such a crazily hard test!

 

[SA210]

I am sure notional amount is 300

Realized Correlation was .4133333.
So the difference when multiplied by 300, gives 0.4 million

 

[Diep Anh Ngo]

Realized Correlation was .4133333.
So the difference when multiplied by 300, gives 0.4 million

Do you remember notional amount is 300 as well?
[0.4133-0.4]*300 < 0.4million

 

[frmpart2dan]

My expected score is 43 correct with a range of 40 to 45 possibly with a small upwards or downward bias

I think I am got 32 qns I know for sure are right, the rest I think 8 qns with high certainty, so about 40 +- 3, very scared indeed as I am borderline 50% only

 

[mghm]

I consider 15 questions to be with 100% accuracy. So,

15+65*0.65 = 54 (remaining attempt with 65% accuracy)
15+65*0.6 = approx 51
15+65*0.5 = approx 47
15+65*0.4 = approx 41
15+65*0.3 = approx 35

This is how I think about my result.

 

[Johnkrause1]

I think I am got 32 qns I know for sure are right, the rest I think 8 qns with high certainty, so about 40 +- 3, very scared indeed as I am borderline 50% only

With me it is similar I have 35 qns I believe to have correct, 13 incorrect and 32 remaining which I cant remember, partly guessed so 25 % expected score on 32 qns

 

[Johnkrause1]

Whats giving me comfort is that I read a post in last years feedback forum and there was someone who was 100% convinced he only got 41 correct and scored 1's in most quantiles and a couple of 2s so hopefully the scoring will be something similar this year

 

[Moey]

Do you remember notional amount is 300 as well?
[0.4133-0.4]*300 < 0.4million

The crux of that question wasn't the plug n' play correlation swap value, but answering what would happen re the receiver if the actual correlations were higher than the expected correlations resulting in 0.4m pnl, so the answer was receive greater than 0.4m.

 

[Diep Anh Ngo]

The crux of that question wasn't the plug n' play correlation swap value, but answering what would happen re the receiver if the actual correlations were higher than the expected correlations resulting in 0.4m pnl, so the answer was receive greater than 0.4m.

Don't agree
The unit which the bank can receive/pay in the question in million not in % (of correlation). It is importance that what is notional amount? (300 or 400, i remember 300)

 

[devdutt.21]

Don't agree
The unit which the bank can receive/pay in the question in million not in % (of correlation). It is importance that what is notional amount? (300 or 400, i remember 300)

it states that the Correlations were miscalculated and lower. Actuals were higher. so the Realized correlation * VAR will be higher than 0.40 million

 

[FRMPART2]

One question was about using equation st _st_1=a(u__st_1) where a was to be taken from the regression equation and value of st was to be calculated for forward month.it was straightforward question

 

[devdutt.21]

One question was about using equation st _st_1=a(u__st_1) where a was to be taken from the regression equation and value of st was to be calculated for forward month.it was straightforward question

Yes i think the value was 0.20 and the forcasted worked out to be 0.28

 

[FRMPART2]

Two questions were asked when u have to use the equation where in lamda~spread/1_rr

 

[FRMPART2]

One question in which different types of portfolio construxtion options were given like screen startification linear prograaming and quadratic programming in which the correct answer was linear programming is not suitable for predefined no of stocks.

 

[FRMPART2]

Fot ho lee model they have asked about lower node in second step.hence it was like ro +(lamda 1+lamda2)_2std sqrt time

 

[frmpart2dan]

One question was about using equation st _st_1=a(u__st_1) where a was to be taken from the regression equation and value of st was to be calculated for forward month.it was straightforward question

That is one of the very few straight forward qns, just plug in 2 times that eqn. Another easy one is the ES where just average 4 numbers. Unfortunately not many qns like that in the exam!

 

[FRMPART2]

Is backtesting abd var should be done at same confidence level or it can be different. As per basel if u have exceptions more than Ci .then capital multiplier is increased to match the VAR at 99 % CI var.hence backtesting can be done at lower confidence interval and increase in exceptions can be managed by capital multiplier

 

[farahm]

Is backtesting abd var should be done at same confidence level or it can be different. As per basel if u have exceptions more than Ci .then capital multiplier is increased to match the VAR at 99 % CI var.hence backtesting can be done at lower confidence interval and increase in exceptions can be managed by capital multiplier

Yes I agree with you. Confidence levels can be different - backtesting does not have to coincide with VaR confidence level.

 

[FRMPART2]

Yes I agree with you. Confidence levels can be different - backtesting does not have to coincide with VaR confidence level.

Hence answer should be backtesting and Var CI should not be same

 

[farahm]

Hence answer should be backtesting and Var CI should not be same

I did choose that yes.Was this part of case studies questions? I forgot.