ahem… Which of the following factors is the common weakness in historical and Monte Carlo Simulation approach to VAR estimation? A- Both assume that historical variance-covariance matrix is stable. B- A lot of data is needed for time period of interest. C- For some assets you may face model risk. It is 7:46 PM CT (T-minus 22 days)… I have a point to make after this one that some may or may not care about but I will post the answer in a couple of hours with rational and explanation. Lets go!

B

B

yah B, A is def out and C is prob true only for Monte Carlo not historical

C (because I failed it in qbank, so i remember the answer…)

B

B

B for me…

C - thinking the assumption that the past returns will repeat could be considered model risk? Historical seems like the least data needed.

Sounds like a tricky question, C as well, for SOME assets (i.e. MBS) you face model risk in both approaches.

C

ANSWER WAS C! From Qbank “The historical method uses actual returns for the position in question. An advantage of the historical method is not having to assume any particular distribution. A disadvantage is that it assumes past performance is representative of what can occur in the future, which may not be the case. The Monte Carlo simulation method for calculating VAR usually involves generating random numbers with a computer. The generated numbers… [error due either to my iphone which i doubt or due to schweser not formatting the questions correctly, but the question ends at my screen here. 50/50 chance of my error or their error either way” Total was 6 candidates 22 days away from the exam were wrong and 3 were right (4 if you count Mr. Scream). 1st. I dont think I have seen the term “a lot of data” in many CFAI readings so that threw me off, as did “historical variance-covariance matrix.” Last, the correct answer was, was for “some”, you “may?” Maybe we should consider posting random CFAI text questions instead? Either way, just an interesting question that drives me crazy sometimes. lol!

Maybe QBank is wrong? Does historical method have a “model”? What’s WRONG with B anyway? Historical method doesn’t have lots of data?

the explanation makes sense for historical you are projecting the past model into the future …time regime risk for monte carlo you have to make assumption on the distribution of the variables so there is also model risk Fawk me

Historical method has no model risk. I’m sure CFAI wouldn’t ask a question as stupid as this, or hopefully it would get thrown out by the Angoff process.

lol i love getting questions right with the wrong reasoning

Correct Answer should be A. They both assume variance covariance matrix is stable. Historical assumes past performance is representative of the future ==> Historical variance covariance matrix is the same as the one for the future (it is stable). Montle carlo simulation generates random numbers - BUT it has to be given an underlying distribution to draw from. Once you specify a distribution, you are indeed specifying a variance covariance matrix.

Skip, what’s the q#?

?

sorry deriv. couldnt find it after a cursory search. i will let you know if i come across it again, but it definitely was a scweser question