Exam Feedback November 2016 Part 2 Exam Feedback

[Eternity]

There was this question on improving ES metric (something of that sort) I marked the option by increasing sample size and there was this other question for which I marked that increasing sample size would reduce probability of Type II error. Hopefully I got them right.

The first one I chose lengthening the period... but I think I was wrong,

Second one should be VaR confidence level can be different from Back testing confidence level

 

[S666]

Are u sure it was a 2 tails test? VaR is always a one tail test... Well the exact same exercise was on bt video on Youtube. We had to calculate the z value. Mean=5%*252 trading days. Vol =sqrt(nqp)=sqrt(5%*95%*252) and zvalue=(20-mean)/vol which was above critical t value

Yeah I'm pretty sure...VaR is always one tailed but this is not a VaR, it is a significance test, which is usually two tailed u less otherwise specified.

I recall the BT practice question used a 1.96 t stat for a 95% confidence significance test.

 

[Trueman]

Anyone rmb on one question regarding the development of risk manage tool after 2007-2009 financial crisis?
options include:
1. shift to short time horizon to forecast any disaster or crisis
2. credit rating appears to be superior to transitional credit due diligence
3. morn concerns are given to those aspects that cannot be quantified
....

It's from FORGING BEST PRACTICES IN RISK MANAGEMENT
Every point coincides with this paper
a. Time Period / A longer-term view on risk, with regard to both the past and the future, is becoming evident (Schweser)
b. Credit Ratings / The failings of credit ratings are commonly thought to have had a significant role in the financial crisis (Schweser)
c. dont remember
d. A shift from probability to uncertainty (there's in working paper, there isn't in Schweser)

 

[hkgthebest]

Yeah I'm pretty sure...VaR is always one tailed but this is not a VaR, it is a significance test, which is usually two tailed u less otherwise specified.

I recall the BT practice question used a 1.96 t stat for a 95% confidence significance test.

Maybe one tail test or two tail test should depend on the null hypothesis testing?
H0: X=something is a two tail test while H0: X>something is a one tail test ?
For backtesting the VaR model, it is more common to use H0: exception=expected no.(which is n*p)?

 

[Trueman]

and the FRM credentials will very soon lose its shine. It should remain an elite certificate for only those who are willing to dig deep.

Let's look 3rd January at pass rate. For May16 it was 50,1% and as far as I can see the level of May exam difficulty was unchanged from previous years

 

[FRMPART2]

We all member should poll our expected scores so that we can know expcted cutt off from our sample ...just a reflection of our performances and possible cut off

 

[Johnkrause1]

Correct
1. Cybersecurity- take the systems offline and then report problem
2. Non-quantifiable risks since the financial crises
3. 90%smm
4. Self control risk assessment
5. Post high quality bonds as collateral
6. Overcollateralization
7. Pricing of a two period bond
8. Liquidity Value at Risk 6.43
9. Calculate the expected loss
10. Libor manipulation happened when banks submitted low rates
11. Reduce moral hazard by letting arranger take stake
12. Other creditors do not have recourse to the SPV
13. Risks aggregated at the granular level
14. Enterprise risk management
15. Correlation Swap receives 0.4 more
16. Expected loss
17. Smoothing only impacts volatility
18. Collapse in ABCP and CP market following Lehman
19. Increasing sample size Expected loss
20. Add stock with lowest MVAR
21. Credit Value at Risk Calculation
22. Principal mapping
23. Reject 0 because critical value above
24. Historical does not rely on normal distirbution

Unsure

1. RAROC
2. Ho Lee calculation
3. Forward Rate Libor calculation
4. CCP multilaternal netting calculation
5. CVA
6. Beta hedge

Incorrect
1. Gumbel Frechet - selected wrong answer
2. Hedge Funds
3. Operational risk
4. 6x9 FRA
5. Implied Volatility Graph
6. Libor/ OIS calculating derivatives
7. Linear/Stratification/Quadratic Programming
8. Minimum transfer amount

This is an analysis of my questions. Bold weighted ones reflect uncertain answers currently in correct but might be incorrect. I can't remember any other questions at the moment and am hoping for 40 correct which I hope will be enough to clear the exam

 

[Eternity]

Correct
1. Cybersecurity- take the systems offline and then report problem
2. Non-quantifiable risks since the financial crises
3. 90%smm
4. Self control risk assessment
5. Post high quality bonds as collateral
6. Overcollateralization
7. Pricing of a two period bond
8. Liquidity Value at Risk 6.43
9. Calculate the expected loss
10. Libor manipulation happened when banks submitted low rates
11. Reduce moral hazard by letting arranger take stake
12. Other creditors do not have recourse to the SPV
13. Risks aggregated at the granular level
14. Enterprise risk management
15. Correlation Swap receives 0.4 more
16. Expected loss
17. Smoothing only impacts volatility
18. Collapse in ABCP and CP market following Lehman
19. Increasing sample size Expected loss
20. Add stock with lowest MVAR
21. Credit Value at Risk Calculation
22. Principal mapping
23. Reject 0 because critical value above
24. Historical does not rely on normal distirbution

Unsure

1. RAROC
2. Ho Lee calculation
3. Forward Rate Libor calculation
4. CCP multilaternal netting calculation
5. CVA
6. Beta hedge

Incorrect
1. Gumbel Frechet - selected wrong answer
2. Hedge Funds
3. Operational risk
4. 6x9 FRA
5. Implied Volatility Graph
6. Libor/ OIS calculating derivatives
7. Linear/Stratification/Quadratic Programming
8. Minimum transfer amount

This is an analysis of my questions. Bold weighted ones reflect uncertain answers currently in correct but might be incorrect. I can't remember any other questions at the moment and am hoping for 40 correct which I hope will be enough to clear the exam

Shouldn't we add the stock with lowest incrementalVaR? MVaR is consider the small change only?

 

[S666]

I chose incremental VaR

 

[mghm]

This time GARP crossed the threshold of BT. In FRM language, the questions were at Block Maxima (GARP) level given the BT the Mu, μ (the threshold).

Earlier BT use to stay above the GARP level but now things have changed.

Hi David .... You are on job now!
@David Harper CFA FRM

 

[frmfrm]

The first one I chose lengthening the period... but I think I was wrong,

Second one should be VaR confidence level can be different from Back testing confidence level

1. Lengthening the period may not be as practical, because the nature of markets, portfolios and risk changes over time (GARP BOOK)

2. Basel Committee requires that market VAR confidence level and Backtesting confidence level both should be calculated at 99% confidence level (GARP BOOK)

Basel Penalty Zones (99% VAR)

Green : 0 to 4
Yellow : 5~ 10
Red : 10 or more

Increasing sample size would reduce probability of Type II error

 

[VAD]

Back to the CVA question. I know you must use the marginal PD in the formula, yet the question was written (imo) in such way that it asked for a CVA over 2 years not specifically in year 2, implying we should use a a cumulative PD instead (1-exp(lambda*2)). What did you guys answer?

 

[raya]

Back to the CVA question. I know you must use the marginal PD in the formula, yet the question was written (imo) in such way that it asked for a CVA over 2 years not specifically in year 2, implying we should use a a cumulative PD instead (1-exp(lambda*2)). What did you guys answer?

Well, I've always seen CVA expressed as ∑ EA*PD*LGD*discount_factor.
So I've calculated
CVA1 = EA1*PD1*LGD1*df1 with PD1 = 1-exp(-hazard rate * 1)
CVA2 = EA2 *PD2*LG2*df2 with PD2 = (1-PD1)*1-exp(-hazard rate * 1) = (1-PD1) * PD1 because it was said that hazard rate is constant.
Then CVA = CVA1 + CVA2

 

[raya]

I've really disliked the long texts with 3 questions associated. I've skipped them and kept them for the end of the exam. But at the end I was so tired I don't even remember those questions ... I count all of them as failed :'(

 

[Rent23]

Shouldn't we add the stock with lowest incrementalVaR? MVaR is consider the small change only?

I choose incremental VaR as well, being the change in VaR owing to a new position = VaRp+a - VaRp

MVaR is the increase in portfolio VaR following an additional dollar of exposure to that component so not the case.

 

[CAN]

.

 

[devdutt.21]

I choose incremental VaR as well, being the change in VaR owing to a new position = VaRp+a - VaRp

MVaR is the increase in portfolio VaR following an additional dollar of exposure to that component so not the case.

MVaR is linear approximation (first partial derivative). Incremental VaR is a full revaluation.
Not sure which one to use as both will give the similar result. However MVaR will be a very fast way to derive so I chose MVAR.

 

[devdutt.21]

Are u sure it was a 2 tails test? VaR is always a one tail test... Well the exact same exercise was on bt video on Youtube. We had to calculate the z value. Mean=5%*252 trading days. Vol =sqrt(nqp)=sqrt(5%*95%*252) and zvalue=(20-mean)/vol which was above critical t value

My understanding
Var Back test is 2 tail test with test static x- pT/Sqrt(p(1-p)T > 1.96

 

[Johnkrause1]

MVaR is linear approximation (first partial derivative). Incremental VaR is a full revaluation.
Not sure which one to use as both will give the similar result. However MVaR will be a very fast way to derive so I chose MVAR.

Its another grey area this one. I can see how both MVaR and Incremental VAR are correct.

 

[devdutt.21]

Correct
1. Cybersecurity- take the systems offline and then report problem
2. Non-quantifiable risks since the financial crises
3. 90%smm
4. Self control risk assessment
5. Post high quality bonds as collateral
6. Overcollateralization
7. Pricing of a two period bond
8. Liquidity Value at Risk 6.43
9. Calculate the expected loss
10. Libor manipulation happened when banks submitted low rates
11. Reduce moral hazard by letting arranger take stake
12. Other creditors do not have recourse to the SPV
13. Risks aggregated at the granular level
14. Enterprise risk management
15. Correlation Swap receives 0.4 more
16. Expected loss
17. Smoothing only impacts volatility
18. Collapse in ABCP and CP market following Lehman
19. Increasing sample size Expected loss
20. Add stock with lowest MVAR
21. Credit Value at Risk Calculation
22. Principal mapping
23. Reject 0 because critical value above
24. Historical does not rely on normal distirbution

Unsure

1. RAROC
2. Ho Lee calculation
3. Forward Rate Libor calculation
4. CCP multilaternal netting calculation
5. CVA
6. Beta hedge

Incorrect
1. Gumbel Frechet - selected wrong answer
2. Hedge Funds
3. Operational risk
4. 6x9 FRA
5. Implied Volatility Graph
6. Libor/ OIS calculating derivatives
7. Linear/Stratification/Quadratic Programming
8. Minimum transfer amount

This is an analysis of my questions. Bold weighted ones reflect uncertain answers currently in correct but might be incorrect. I can't remember any other questions at the moment and am hoping for 40 correct which I hope will be enough to clear the exam

15. Correlation Swap receives 0.4 more
My understanding . Formula is a multiplier of Sum of Correlations. with 2/(n2-n) . Since correlations were understated the sum of correlations will be higher.
hence Correlation Swap receives 0.4 more is correct as per me.