Serving the Quantitative Finance Community

 
User avatar
felix123321

Parametric Value at risk for savings plan

June 17th, 2020, 2:19 pm

Hi all!

I have a question on how to calculate the parametric 95% VaR for a savings plan that is investing in a risky asset.
We have an initial amount and in addition monthly recurring investments. 
Without monthly payments the answer is easy (Standard delta norm. dist. VaR):

Var = ((1+r)^t - 1,65*v*(t)^0,5) * x

x = Initial investment
v = Expected (monthly) volatility of asset
r = Expected (monthly) return of asset 
t = Total period in months

How is that calculated with regular payments (of equal size).
I have not found any closed end parametric solution for this problem so far and run out of time : (
I am thankful for any indication or link to a solution.

Best regards,
Felix
 
User avatar
Alan
Posts: 2958
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Re: Parametric Value at risk for savings plan

June 17th, 2020, 3:22 pm

I will guess there is no closed form -- as it looks like you need the distribution of a sum of correlated lognormal variates. My advice is to just spend a day and develop a nice Monte Carlo with errors below your required accuracy. It will run in a few seconds and then you'll be done. 
 
User avatar
felix123321

Re: Parametric Value at risk for savings plan

June 18th, 2020, 1:33 pm

ok understood .. that would be the alternative
 
User avatar
DavidJN
Posts: 242
Joined: July 14th, 2002, 3:00 am

Re: Parametric Value at risk for savings plan

June 19th, 2020, 10:00 am

Wasn't the usual solution strategy for parametric VaR to map the portfolio cash flows to standardized tenors that one can find a covariance matrix for? This was explained with examples in the original Riskmetrics technical documentation.
 
User avatar
Alan
Posts: 2958
Joined: December 19th, 2001, 4:01 am
Location: California
Contact:

Re: Parametric Value at risk for savings plan

June 19th, 2020, 2:16 pm

Sounds like an approximation worth exploring.  

Another approach -- likely related -- would be to look at approximate Asian option valuation (w/ monthly averaging). You could split the terminal distribution into the sum of two correlated variates: one from the initial investment and a second from the sum of the (presumably equal) monthly investments. The terminal distribution of the latter can perhaps be well-approximated by the Asian option arguments.

The advantage of first developing an accurate Monte Carlo is that, once you have it, you can then go on to test approximations.