Would historical VaR be appropriate for a portfolio that recently changed its risk exposures?
no, because the past cannot reasonably be expected to predict the future when the composition of its exposures are not the same
it would be useful if there were no material changes to the portfolio
it uses the past but the assessment is done using the current portfolio holdings. I dont understand why this cannot be used
Yes. The historical method does not use the actual returns of the historical portfolio. It looks at the distribution of returns that the _ current _ portfolio would have had.
p.57 in curriculum
To use an extremem example:
Say you have a bond portfolio that is exposed entirely to the utility sector, then you reallocate all of those bonds to start-up technology bonds.
Do you still want to use historical VAR to estimate the risk?
I say no.
Past won’t be relevant if regime change, etc.
Yes, inasmuch as VaR in general, and historical in particular, is appropriate for fixed income. The past history of the portfolio holdings is irrelevant, and it is equally true for any VaR method.
so you are saying you can use it when risk exposures recently change?
Yes. Recent changes are not relevant to the risk measurements of current portfolio.
^ if you are using the VAR to predict future losses, you are way off base
Historical VAR can be a simulation - ‘what if’
Sounds reasonable that you can use it with new risk parameters
So you two are saying that if you have a portfolio of US giovernment 1 yr t-bils, and then suddenly change the portfolio the next day to long duration Greek givernment debt, the historical VAR of the US t-bills portfolio should accureately reflect the risk of the new Greek debt portfolio?
If you ran a simulation off historical data, why couldn’t you?
^ historical VAR takes actual losses that occurred in the portfolio, ranks them, and then chooses the bottom x% worst return as your x% VAR - how can you say that the greek debt and the us debt portfolios lose at least the same amount with the same probability
Whether it is VaR in general, or historical VaR in particular, you are predicting the risk of current portfolio assuming it stays the way it is. Its past composition does not affect your result.
Appropriateness of historical method is determined by other things, like the asset classes in current portfolio (whether return distribution is close to normal), existence and availability of past data for the assets, and whether there are known regime changes during the VaR period.
LOL well for my sake, i hope this is how you answer on the exam
so ludacris i think you are trolling
Go and see the 2011 AM Mock Q8 Part A
Thanks 
I’ll quote actual CFAI text to make you comfortable. But unfortunately, they made it a little confusing by using the words “actually happened”. They mean it with respect to past prices, of course, not the past protfolio, and they explain immediately afterwards.
The historical method is also sometimes called the historical simulation method. This term is somewhat misleading because the approach involves not a simulation of the past returns but rather what actually happened in the past. In this context, note that a portfolio that an investor might have held in the past might not be the same as the one that investor will have in the future. When using the historical method, one must always keep in mind that the purpose of the exercise is to apply historical price changes to the current portfolio.39
That’s right, either type of VaR was not appropriate there based on composition of current portfolio. It is inappropriate because of a recent regime change (volatility) and non-normal return distribution in currently held assets.