Exam Feedback FRM Part 2 (May 2015) Exam Feedback

[Alberto G]

I might get around 45-50.. :/

Hope 50 then ;-) Personally I have no clue of where could I be, but with 50 I would feel quite safe.

 

[maibhihun]

i am aghast with 45-50 cutoff. I passed L-1 with mostly Q1 and so have some idea of this L2 pool, on an average. Poor guys who do not have to do calculation and happy choosing some option may think 50 can be cutoff. in my opinion, exam was super difficult and cutoff should be lower than 30...

 

[Alberto G]

i am aghast with 45-50 cutoff. I passed L-1 with mostly Q1 and so have some idea of this L2 pool, on an average. Poor guys who do not have to do calculation and happy choosing some option may think 50 can be cutoff. in my opinion, exam was super difficult and cutoff should be lower than 30...

So if I knew 10/80 and wildly guessed the other 70 and be lucky enough to be in line with the ex-ante prob of 25% right (so to sum to 28 correct answer) I would pass?? What can I say... I hope you are right, we will probably all be FRM in 6 weeks ;-)

All these certifications (CFA included) have cutoffs as percentages (around 70%) on the best tail of the total distribution of the exam takers, 30-point cutoff means the reference sample (the best, not the mean) scored between 40 and 43 out of 80... c'mon...

 

[Salonica]

It was for severity and most certainly lognormal.

Correct. Unfortunately I got this easy one wrong. Along with more easy ones!

 

[Salonica]

how many questions could we gather until now? there is also a question on hedge fund strategy. which of the strategies has the lowest correlation with its (strategies) index.
And another question on "if correlation of the of assets in the pool increases what could result" I chose the CVAR decreases. I might be wrong.

Portfolio credit VaR is a quantile of the credit loss minus the expected loss of the portfolio.
Default correlation has a tremendous impact on portfolio risk.
 Default correlation affects the volatility and extreme quantiles of loss but does not
impact the expected loss (EL).
 If default correlation in a portfolio of credits is equal to 1.0, then the portfolio behaves
as if it consisted of just one credit. No credit diversification is achieved.
 If default correlation is equal to 0, then the number of defaults in the portfolio is a
binomially distributed random variable. Significant credit diversification may be
achieved.

From BT notes

 

[vsism]

So if I knew 10/80 and wildly guessed the other 70 and be lucky enough to be in line with the ex-ante prob of 25% right (so to sum to 28 correct answer) I would pass?? What can I say... I hope you are right, we will probably all be FRM in 6 weeks ;-)

All these certifications (CFA included) have cutoffs as percentages (around 70%) on the best tail of the total distribution of the exam takers, 30-point cutoff means the reference sample (the best, not the mean) scored between 40 and 43 out of 80... c'mon...

You are right I think the cutoff will be somewhere close to 45.....Just praying to God that I get atleast those many right!!

 

[Alberto G]

You are right I think the cutoff will be somewhere close to 45.....Just praying to God that I get atleast those many right!!

this could be, that's why with 50 I said I would feel pretty comfortable...

 

[[email protected]]

Distribution required for LDA. I answered Poisson

Frequency distribution requires Poisson while severe loss requires lognormal distribution to model.

 

[mary_ful]

The higher the correlation the more useless the cda. There were two graphs that showed decreasing value of cds with correlation. I choose the one that eventually ended at zero, since the cds is worth nothing, if the correlation is one. At least thats true if the recovery rate is zero. I can't remember, i hope there was no recovery rate mentioned.

I selected the same one the CDS ended at zero because of recallying only VAR & system risk is straightforwad line

 

[mary_ful]

I don't know why not allow to take a watch on hand, which is hard to control time for me. And I felt that not be fair with non-english country. 80 questions including senario were not easy to have/ sufficient time to consider per question behind tricky. Long time will wait the testing results.....

 

[mary_ful]

Another basel related issue was with the amount requied fundin in the denominator and wether the ratio is met. I chose 55.4 and the ratio was greater that 100%

I got the answer but sorry for I forgot, required funding in assets & denominator, the logic is correct. hope both we selcet the right one

 

[mary_ful]

There was one question with the Jensen's and convexity. I had some doubt between two answers because of the concexity part in the explanation.
An other one related to scoring model. Which can help during the process: SVM, merton, Area under the Curve or ROC.
Even for the smirk and the 4 graphs I had some doubt between two.

I guess ROC due to separting good borrower from bad's

 

[mary_ful]

Longing correlation is to either
1. receive floating rate in a variance swap with the index as an underlying and receive fixed in a variance swap with a stock in that index as an underlying,
2. Buy call option on index and buy call option on an individual component in that index OR
3. receive realized correlation in a correlation swap
in the exam, only 1 was shown
<noted> I guess (1)

For the CCP, to avoid being failed
CCP need to:
1. ONLY intermediate derivative transactions
2. let the clearing member to unwind the trades in case there are defaults
3. have good practice in choosing members, valuing transactions and determining initial margins and default fund contributions
In the exam, only 2 was shown
<noted> I guess (3), i thought unwind the trades due to defaults not fix the CCP problem, may amply external default risk

Limitation of using BSM model to value bond
The answer is the volatility will go to zero at maturity as the model assumes the volatility is constant

CLN question (sth related to minimize counterpart risk)
The counterparty risk is the lowest in issuing credit link notes given the protection sellers, i.e. investors, have already paid the price to the buyers and hence no counterparty risks in case there is a default in a reference asset.

Net Stable Funding Ratio
The bank passes the test given the ratio, ASF divided by RSF, is greater than 1 (in fact, 1.3XX)
RSF is 5X.X & ASF is 72

Standardized vs Basic Indicator
The bank is of course paying 37million more under basic indicator approach because, in the latest 3 years, there is a negative gross income and hence you will divide the capital by 2 under basic indicator approach vs by 3 under standardized approach

QQ plot
Thinner tails for sure

95% Credit Var of 100 CDS with $1000, 2% PD and LGD 100% each
The answer is 3000
At 95% level, there will be five defaults so the value is 5*1000
the expected loss is 100*0.02*1000*1

Convexity
Convexity will of course increase the bond price comparing with simply using the expected value of interest rate

Square root rule
For time varying volatility, e.g. volatility estimated via GARCH, using square root rule will overestimate the VaR whereas underestimate in case the underlying process has a jump

Policy needs to be corrected
The senior bond is only subordinated to preference share

Rule of thumb
Correlation of zero does not imply independence

 

[saran11]

Portfolio credit VaR is a quantile of the credit loss minus the expected loss of the portfolio.
Default correlation has a tremendous impact on portfolio risk.
 Default correlation affects the volatility and extreme quantiles of loss but does not
impact the expected loss (EL).
 If default correlation in a portfolio of credits is equal to 1.0, then the portfolio behaves
as if it consisted of just one credit. No credit diversification is achieved.
 If default correlation is equal to 0, then the number of defaults in the portfolio is a
binomially distributed random variable. Significant credit diversification may be
achieved.

From BT notes

Thanks for the reference!

Can you remember the answer. Just curious what could be the answer.

 

[mary_ful]

Unfortunately don't agree with your answer
If you are to allocate $100 million VaR limit among 3 managers with UNCORRELATED portfolios, the VaR Limit on each manager should be
Sq root (VaR ^2 + VaR ^ 2 + VaR ^2 ) = 100
VaR Limit on each manager must be < 100

I gues sthe same answer , VAR limit on each manager was <100, I guess 84 or 8X, hope both we get the right answer.

 

[xenatr]

I guess ROC due to separting good borrower from bad's

ROC is used to determine the model accuracy instead of being used to predict the default possibility. Therefore, ROC must be ruled out.

I gues sthe same answer , VAR limit on each manager was <100, I guess 84 or 8X, hope both we get the right answer.

We have discussion on whether the question is asking to determine VaR limit or the portfolio amount and cannot reach a consensus on the question details.

I got:
1 - (1 - 0.042 / 0,6 ) ^ 2 = 0.1351

Which uses not the continous but discrete PD(!)
I have to say that, despite the statement in the cover about all interest rates and the like to be continously compounded, some were definitely not, and not marked as such. I am positive that the question about two ways to average forward rates also used discretely compounded rates. I think GARP wasn't consistent on this.

Thanks for your explanation. I dont recall this formula To be precise, this one is an approximation of the continuous formula.
Discounting on a yearly basis is not equal to discounting on a continuous basis.
Approximation is approximation.

 

[mary_ful]

Some general questions I remember that may not have been mentioned yet here:

• Asking question regarding payment vs close-out netting and the difference
• 3 question set on XYZ's SPE of mortgage bonds, one question was the most pressing concerns regarding the consulting companies proposals for the SPE…
• Choosing between "Gamma" and "Theta" companies based upon balance sheets as the better counterparty, i.e. longer duration assets, shorter duration liabilities, less non-USD exposure, and something else were answer choices
• Repo with Hedge Fund question – some answers included special rate b/w General and Special, I chose an increase in bond price will reduce the HF’s exposure
• Question on Exposure of $9ook with a $1m threshold and $250k mta, exposure increases to $1.2m, and the outcome. I did not like this question as I thought this should trigger a $1.2m margin call, but that was not an answer. I chose increase $300k but the answer may have been zero.
• Question on Cybersecurity and the next step after discovering some sort of issue (forget exact details)

regarding to "Question on Exposure of $9ook with a $1m threshold and $250k mta, exposure increases to $1.2m, and the outcome. I did not like this question as I thought this should trigger a $1.2m margin call, but that was not an answer. I chose increase $300k but the answer may have been zero."---- I guessed zero , the reason is threshold +mini transfer > exposure

 

[JDGutzmann]

...

Thanks for your explanation. I dont recall this formula To be precise, this one is an approximation of the continuous formula.
Discounting on a yearly basis is not equal to discounting on a continuous basis.
Approximation is approximation.

I agree - only picked up this formula in one of their old practice exams. As I said, I think its strange for GARP to be using different methods without clarification.

 

[MK55]

Hi,
There was a question asking 95% Credit Var of CDS with 2% PD and LGD 100% each

I don't know if my mind is playing tricks. But if i remember correctly we were given a cumulative probability distribution table with associated number of defaults. I am not sure but I guess it was mentioned 3 defaults besides a cum prob slightly less than 95%.
This question was previously discussed here where the answer was suggested to be 5k-2k = 3k. So to convince myself I'm thinking 5 defaults was mentioned below 3 with a cum prob greater then 95%. So shouldn't it be 3k -2k =1k?

 

[saran11]

Hi,
There was a question asking 95% Credit Var of CDS with 2% PD and LGD 100% each

I don't know if my mind is playing tricks. But if i remember correctly we were given a cumulative probability distribution table with associated number of defaults. I am not sure but I guess it was mentioned 3 defaults besides a cum prob slightly less than 95%.
This question was previously discussed here where the answer was suggested to be 5k-2k = 3k. So to convince myself I'm thinking 5 defaults was mentioned below 3 with a cum prob greater then 95%. So shouldn't it be 3k -2k =1k?

I thought its was not cumulative default probabilities. I think I got 3k (5k - 2k). I got 5 defaults to reach 95%. Might be wrong though. :/