What's new

# Exam FeedbackFRM Part 1 (May 2014) Exam Feedback

#### plulutes

##### New Member
was there any final decision how to calculate the EWMA question where a cuple of daily returns were given?

That was a question that I just skipped and never got back to. I hadn't prepared to answer that kind of EWMA question as I'd never seen it as a practice question. I suppose I'm glad I saved the time by avoiding it.

#### Jo_

##### Member
Subscriber

I got the EL and got a selected answer just using EL=PD*AE*LGD but don't remember the number now but, UL I remember something like 118 or 180...anyway 1's and 8's. UL=AE * sqrt (variance of LGD*PD + LGD^2 * PD * (1-PD) ) However, the PD *(1-PD) was replaced with EDF^2 that was given. That's what I came up with anyway.

I don't remember the exact answers, but indeed, if you knew those two formulas, the two questions were, imo, straightforward. Just plug in the variables which were all given (did not even have to calculate PD * (1-PD) because variance for PD was simply given + LGD was given directly, instead of recovery rate). I do remember the 67 as AE indeed. Don't recall the final answers though.

Did not see the need of calculating an updated variance.
Omega ( Long run average variance rate * Gamma) was provided
Gamma as you pointed out was calculated using the 1-Alpha-Beta logic
The Long Run Average Variance Rate would then be-> Omega / Gamma
Take a square root of the long run variance rate to arrive at the long run volatility.
Compare the present volatility estimate to the long run volatility rate to arrive at whether there is going to be a +ve or -ve drift.

Fully agree.

And for everyone who starts getting more and more stressed after reading through this thread:  Last edited:

#### brian.field

##### Well-Known Member
Subscriber
Perhaps I am wrong, but I swear there were a couple questions that were simply flawed. I remember one question that stated something like the following....

If considering the mean variance framework, skewness and kurtosis do not exist. I am paraphrasing....
This is a flawed statement; the 3rd and 4th moments exist but are not considered if operating under a mean-variance framework, or to be even more precise, the third moment is 0 and the fourth is implied by the second, so it is redundant, but exists nonetheless.

Anyone remember this question?

Thanks!

Brian

#### brian.field

##### Well-Known Member
Subscriber
Overlooked it completely I might be mistaken, but I thought that the proctor announced that there was a z table in the test booklet (or the problem stated it ... or something else.) For some reason, I was aware that it was there....

#### amresh

##### Member
Subscriber
was there any final decision how to calculate the EWMA question where a cuple of daily returns were given?
I just applied EWMA formula and solved for Lambda. I was not having much time left and the mind was completely shut by that time. I got around 0.97, if I remember it correctly .

##### Member
I got smthg like 0.92 or 0.94....

#### plulutes

##### New Member
I got smthg like 0.92 or 0.94....

While I skipped this question, I mean to say that I didn't think too hard on it, but I remember that I answered .94. But here's why and also here's the question. I picked .94 just because I thought that it was a given decaying weight (Risk Metrics copyright blah blah blah)...ya know, the one we usually use in these problems. Didn't think to solve for it or even that you could. With that said, I don't remember if the question asked for a solution. I checked study materials and saw that daily returns EWMA uses .94 and Monthly uses .97. So, Amresh and Adelaide or anyone, do you recall if the daily returns were hints or did it actually say "solve" for lamba, or what "is" lamba? My head still hurts.

#### Roshan Ramdas

##### Active Member
While I skipped this question, I mean to say that I didn't think too hard on it, but I remember that I answered .94. But here's why and also here's the question. I picked .94 just because I thought that it was a given decaying weight (Risk Metrics copyright blah blah blah)...ya know, the one we usually use in these problems. Didn't think to solve for it or even that you could. With that said, I don't remember if the question asked for a solution. I checked study materials and saw that daily returns EWMA uses .94 and Monthly uses .97. So, Amresh and Adelaide or anyone, do you recall if the daily returns were hints or did it actually say "solve" for lamba, or what "is" lamba? My head still hurts.
Believe that the question asked something like what is the implied "smoothing parameter" based on the data (returns and volatility) for around 4 or 5 days......plugging the last days return and volatility numbers into the standard EWMA formula gave me a lambda of 100 % which cannot be the right answer.....I repeated the calculation twice only to get the same number. Landed up going for one of the above 95% options.....don't remember the exact value.

#### hoang.vu90

##### New Member
Believe that the question asked something like what is the implied "smoothing parameter" based on the data (returns and volatility) for around 4 or 5 days......plugging the last days return and volatility numbers into the standard EWMA formula gave me a lambda of 100 % which cannot be the right answer.....I repeated the calculation twice only to get the same number. Landed up going for one of the above 95% options.....don't remember the exact value.
I use the last day for my calculation of lambda as well. Couldnt work it out either. Did it 3 times and got some funny number from the calculator...;(

#### Roshan Ramdas

##### Active Member
Perhaps I am wrong, but I swear there were a couple questions that were simply flawed. I remember one question that stated something like the following....

If considering the mean variance framework, skewness and kurtosis do not exist. I am paraphrasing....
This is a flawed statement; the 3rd and 4th moments exist but are not considered if operating under a mean-variance framework, or to be even more precise, the third moment is 0 and the fourth is implied by the second, so it is redundant, but exists nonetheless.

Anyone remember this question?

Thanks!

Brian[/quote

Remember this question and remember selecting the answer which mentioned that the mean variance framework did not consider the skew and kurtosis or something to that effect.

#### Roshan Ramdas

##### Active Member
I use the last day for my calculation of lambda as well. Couldnt work it out either. Did it 3 times and got some funny number from the calculator...;(

Same here,....I used the last days data.....strange why we didn't get the answer,....the question really came across as being fairly straightforward.

#### plulutes

##### New Member
Believe that the question asked something like what is the implied "smoothing parameter" based on the data (returns and volatility) for around 4 or 5 days......plugging the last days return and volatility numbers into the standard EWMA formula gave me a lambda of 100 % which cannot be the right answer.....I repeated the calculation twice only to get the same number. Landed up going for one of the above 95% options.....don't remember the exact value.

I thought that Lambda was a given, unchanging weight and as time increases the impact on variance will decrease. So this is "solving" for variance as EWMA does, not lambda parameters. That was my understanding. Do you remember having to solve for Lambda before in practice? Just the question is making me think it was a trick, or I'm way off base on this one, which is very possible too. But the big question is do we actually change lambda with time? What do you think?

#### Roshan Ramdas

##### Active Member
Perhaps I am wrong, but I swear there were a couple questions that were simply flawed. I remember one question that stated something like the following....

If considering the mean variance framework, skewness and kurtosis do not exist. I am paraphrasing....
This is a flawed statement; the 3rd and 4th moments exist but are not considered if operating under a mean-variance framework, or to be even more precise, the third moment is 0 and the fourth is implied by the second, so it is redundant, but exists nonetheless.

Anyone remember this question?

Thanks!

Brian
Remember this question and remember selecting the answer which mentioned that the mean variance framework did not consider the skew and kurtosis or something to that effect.

#### plulutes

##### New Member
Remember this question and remember selecting the answer which mentioned that the mean variance framework did not consider the skew and kurtosis or something to that effect.

I only remember a question that described a mean and variance as being the only considerations...something to that effect, which informs us of how to look at the question given that we are supposed to consider a normal distribution to answer some other part of the question. But I really dont remembe what the specific question was. The "make the assumption that we are considering a mean and variance only" was/is a common pre-qualifiying parameter to practice questions so I didn't think anything more profound about it.

#### Roshan Ramdas

##### Active Member
I thought that Lambda was a given, unchanging weight and as time increases the impact on variance will decrease. So this is "solving" for variance as EWMA does, not lambda parameters. That was my understanding. Do you remember having to solve for Lambda before in practice? Just the question is making me think it was a trick, or I'm way off base on this one, which is very possible too. But the big question is do we actually change lambda with time? What do you think?

Really sharp memory to recollect the daily and monthly optimal lambda values.

Just browsed through the Bionic notes and here is the section that you are referring to,.....0.94 could really be a possible value,...the only reason why I can think of disputing this answer is based on the line in bold,....it makes it look like that the lambda values could vary by asset class and the value 0.94 and 0.97 are more like optimal or ideal values.

RiskMetrics™ Approach RiskMetrics™ is a branded form of the exponentially weighted moving average (EWMA) approach. The optimal (theoretical) lambda varies by asset class, but the overall optimal parameter used by RiskMetrics™ has been 0.94. In practice, RiskMetrics™ only uses one decay factor for all series:  0.94 for daily data  0.97 for monthly data (month defined as 25 trading days)

#### Roshan Ramdas

##### Active Member
I only remember a question that described a mean and variance as being the only considerations...something to that effect, which informs us of how to look at the question given that we are supposed to consider a normal distribution to answer some other part of the question. But I really dont remembe what the specific question was. The "make the assumption that we are considering a mean and variance only" was/is a common pre-qualifiying parameter to practice questions so I didn't think anything more profound about it.
My approach was more with respect to the CAPM being based on a mean variance framework and as per the CAPM,.....investors care about the first and second moment alone and the skew and kurtosis are completely disregarded.

#### plulutes

##### New Member
Really sharp memory to recollect the daily and monthly optimal lambda values.

Just browsed through the Bionic notes and here is the section that you are referring to,.....0.94 could really be a possible value,...the only reason why I can think of disputing this answer is based on the line in bold,....it makes it look like that the lambda values could vary by asset class and the value 0.94 and 0.97 are more like optimal or ideal values.

RiskMetrics™ Approach RiskMetrics™ is a branded form of the exponentially weighted moving average (EWMA) approach. The optimal (theoretical) lambda varies by asset class, but the overall optimal parameter used by RiskMetrics™ has been 0.94. In practice, RiskMetrics™ only uses one decay factor for all series:  0.94 for daily data  0.97 for monthly data (month defined as 25 trading days)

Actually, I wasn't sharp at all, I skipped over this question because I didn't fully understand it so I just defaulted in answering it by guessing .94 as it was the only parameter I remember using and I'd thought it was a given under Risk Metrics. However, the .97 I had to look up as this exam question was raising questions. So, this morning I found that .94 was daily, and the exam question was giving "daily" returns. Made me think that no solving was required, but only to recognize daily lambda.

#### plulutes

##### New Member
My approach was more with respect to the CAPM being based on a mean variance framework and as per the CAPM,.....investors care about the first and second moment alone and the skew and kurtosis are completely disregarded.

Ahhh, was that how the question was posed (CAPM)? I was trying to give Brian an answer, but didn't remember the question fully.

#### Roshan Ramdas

##### Active Member
Ahhh, was that how the question was posed (CAPM)? I was trying to give Brian an answer, but didn't remember the question fully.
There wasn't any mention of CAPM on the face of the question,.....the mean variance framework is a core CAPM assumption basis which investors care about the first 2 moments alone.

#### amresh

##### Member
Subscriber
I remembered .94 and .97 part; but as per schweser, Riskmetrics is a special case of ewma for which we use .94 n .97 for daily n monthly calculations. And question had ewma mentioned. Hence I randomly used figures of third day and got something close to .97. But I can't rely on my calculations, as I was out of gas, when I was doing this question.