标题: 3 questions - ERP, asset allocation, Monte carlo [打印本页] 作者: DoubleDip 时间: 2011-7-11 19:51 标题: 3 questions - ERP, asset allocation, Monte carlo
1. Please remove my doubts regarding market risk premium and equity risk premium once and for all. I still get confused even in level 3 (no hope !! )
Ri = Rf + beta * (Rm - Rf)
I thought (Rm - Rf) = market risk premium
beta * (Rm - Rf) = Equity risk premium
Just saw a sample exam where (Rm - Rf) is referred to as equity risk premium. can that be right?
2. Is "investor risk tolerance is constant" for both Strategic asset allocation and Tactical asset allocation?
I have seen statements where either one of the asset allocations are being discussed and statements appear like "investor risk tolerance is constant"
3. CFAI Vol 5 page 279, Problem 12.
It is a monte carlo var problem. They are asking to calculate 5% annual var.
we are provided with 40 worst returns of 700 outcomes.
we are also provided with expected return and standard deviation.
I thought Monte Carlo method uses the same formula as variance covariance method. However CFAI solves it using similar method as Historical method.
?????
Thanks in advance for help.作者: bkballa 时间: 2011-7-11 19:51
1) I agree, market risk premium is (Rm - Rf).
But, I don't the CFA defines ERP as beta * (Rm - Rf)? I say that because this is the expected rate of return. Whereas, I thought ERP is linked to the required return.
2) It makes sense that it would be for "Strategic", because you're rebalancing to prescribed weights.
I think it is for "Tactical", even though you're shifting weights of the risky assets based on short-term expectations... because the analytical framework doesn't address a change to risk tolerance (which would change the tactical behavior). However, not sure if the readings say?
3) You would do a similar exercise as with historical VAR because it's just a statistical exercise. They have the 40 worst outcomes out of 700, so pick the 35th worst as 5% VAR.
The format's confusing, but they only have so many ways to test Monte Carlo simulation (with a non-statistical crowd anyway).作者: aidebaobao 时间: 2011-7-11 19:51
drymartini Wrote:
-------------------------------------------------------
> 1. Please remove my doubts regarding market risk
> premium and equity risk premium once and for all.
> I still get confused even in level 3 (no hope !!
> )
>
> Ri = Rf + beta * (Rm - Rf)
>
> I thought (Rm - Rf) = market risk premium
> beta * (Rm - Rf) = Equity risk premium
Equity risk premium is typically not defined for a specific stock but rather for an index. If your Rm is based on a stock index (such as SP500), that Rm-Rf would be equity premium as well as market premium.
> 2. Is "investor risk tolerance is constant" for
> both Strategic asset allocation and Tactical asset
> allocation?
constant risk tolerance is an assumption that should be specified in the question. I wouldn't assume constant risk tolerance (constant risk aversion) if it's not given. Markowtiz model though assumes constant risk tolerance.
> 3. CFAI Vol 5 page 279, Problem 12.
>
> It is a monte carlo var problem. They are asking
> to calculate 5% annual var.
> we are provided with 40 worst returns of 700
> outcomes.
> we are also provided with expected return and
> standard deviation.
>
> I thought Monte Carlo method uses the same formula
> as variance covariance method. However CFAI solves
> it using similar method as Historical method.
both methods are fine. those calculating quantiles is more accurate than estimating VaR using standard deviation as that implies normal distribution.作者: zwjy 时间: 2011-7-11 19:51
OK, I will weigh in.
drymartini, you are correct, the CAPM provides a means of estimating the equity risk premium for a spefic stock. It's as you define it. Equity Risk Premium = beta*Market Risk Premium. if you saw this: "Just saw a sample exam where (Rm - Rf) is referred to as equity risk premium. can that be right?" It is not correct. Rm - Rf is the market risk premium.
3. Monte Carlo VaR - correct, you are not imposing distributional assumptions on the process. So you just choose the worst 5% case, if this is what your confidence level is, as described by NeverUse .... the difference is in variance-covariance method, you are implicitly assuming normal distribution (let's say) so you can just use r - 1.65 standard deviation etc.
