VaR Methods & Filtered Historical Estimation

[RomanS]

Hi,
i just encountered the following question from 2010 Schweser Practive Exams E1 L2 Q 47

Question
Jim Johannsen has collected a large data set of daily market returns for three emerging markets.
He is concerned about the non-normal skew in the data and is considering non-parametric estimation methods.
Johannsen is not familiar with these techniques and he discusses the prodcedure with his colleague Lily Tong.
During the course of their discussion, Lily makes the following statements:

i) "Age-weighted historical simulation redcues the impact of older observations only after surpassing a user-defined threshold."
ii) "Volatility-weighted historical simulation augments historic returns with an additive volatility adjustment."
iii) "Filtered historical estimation incorporates sophisticated parametric as well as non-parametric techniques."

How many of Ms. Tong's statements are correct

a) Zero
b) One
c) Two
d) Three

Answer
a) Zero

While I agree that i) and ii) are wrong, I doubt iii) is wrong since Filtered historic simulation [FHS] makes use of a (sophisticated) paramteric technique, e.g. (G)ARCH and it also uses a non-parametric technique, namely basic historical simulation.

Now, the sample answer provided says: "...Statement III is incorrect because filtered historical simulation combines sophisticated non-parametric estimation with traditional historical simulation."

I think this one too, is wrong as FHS combines parametric estimation (e.g. GARCH) with traditional historical simulation.

My answer would be b) One

What would your answer be and why?

Best Roman

 

[David Harper CFA FRM]

Hi Roman,

I agree with you that Statement III is correct. Dowd call FSH "semi-parametric." The underlying HS is obviously non-parametric but the only filter described (GARHCH) is clearly parametric (it may be that a non-parametric filter is also possible but that would not render III as false!).

From Dowd (emphasis mine):
4.5.1. "Non-parametric approaches are capable of considerable refinement and potential improvement if we combine them with parametric ‘add-ons’ to make them semi-parametric: such refinements include age-weighting (as in BRW), volatility-weighting (as in HW and FHS), and correlation-weighting.

Also, I checked the underyling paper (Non-parametric VaR techniques. Myths and Realities. By Giovanni Barone-Adesi and Kostas Giannopoulos):
"The problems of earlier models spurred the search for better estimates of VaR. A number of recent VaR techniques are based on non-parametric or mixture of parametric and non-parametric statistical methods. The family of Historical Simulation (HS) models belongs to the former group. The Filtered Historical Simulation (FHS) as developed by Barone-Adesi et al (1998) and Barone-Adesi et al (1999, 2000) belongs to the second group."

Thanks, David

 

[RomanS]

David,
There is nothing to add from my side except saying THANK YOU.
Best Roman