VAR assumption

CFA Level III
#1

VAR increase with time interval only true when R is assumed at 0%. Otherwise, math won’t hold. Confirmation? Anyone?

#2

I believe you can only exptrapolate a Daily VAR to multiple days if you assume R = 0% for daily, otherwise math doesnt hold.

#3

VAR increase with time interval. Always true, unless your portfolio is all Rf. You invest for longer, you’ll have bigger potential losses.

#4

^I think he was referrign to exptrapolating a Daily VAR estimate to a Monthly/Qrtly/Annual VAR which can only be done when Daily return is assumed to be 0%. But yes you are correct that the longer the time priod or Higher the Confidence LEvel the Higehr the VAR will be.

#5

bigwilly is right. TooOld4This, yes, but would you agree the assumption is R = 0%? VAR = R - z*sd. If R <> 0, R will increase with t Sd increase with sqrt (t) Then VAR will decrease with time interval…

#6

If R is increasing more than SD for each t, how will the VAR get smaller as t increases?

#7

Usually is VAR = 10% - 20% = -10% With R increase, such might be -5%, thus VAR decrease. But if we assume R = 0, no such issue any more…

#8

We assume R is 0 when we talk about VAR unless specified otherwise

#9

what’s going on here? VAR = R - z*sd. increase the time interval to t, assume t > 1 then new VaR = t R - z sd * sqrt(t) the VaR will decrease if t R - z sd * sqrt(t) > R - z*sd or R > z * sd / (sqrt(t) + 1) you dont need zero R, but need to be smaller than z * sd / (sqrt(t) + 1). look at ex 10 in reading 37, R is not zero but Var increase with time interval.

#10

Yes! actuaryalfred is good! 1) If R is small enough (0 is for simple example), VAR will increase with time. 2) If R is big enough, VAR will actually decrease. In Exam, go with comp_kid: R = 0

#11

t R - z sd * sqrt(t) > R - z*sd R*(t-1) > z*sd *(sqrt(t)-1) R > z * sd * ( (sqrt(t)-1) / (t-1) )

#12

comp_sci_kid Wrote: ------------------------------------------------------- > t R - z sd * sqrt(t) > R - z*sd > > R*(t-1) > z*sd *(sqrt(t)-1) > > R > z * sd * ( (sqrt(t)-1) / (t-1) ) And (t - 1) = (sqrt(t) + 1)(sqrt(t) - 1) which leads to actuaryalfreds answer.

#13

fsa-sucker Wrote: ------------------------------------------------------- > comp_sci_kid Wrote: > -------------------------------------------------- > ----- > > t R - z sd * sqrt(t) > R - z*sd > > > > R*(t-1) > z*sd *(sqrt(t)-1) > > > > R > z * sd * ( (sqrt(t)-1) / (t-1) ) > > And (t - 1) = (sqrt(t) + 1)(sqrt(t) - 1) which > leads to actuaryalfreds answer. :slight_smile: yes he just skipped the step of a^2-b^2 = (a+b)(a-b)