标题: *Not a Question about CFA* [打印本页] 作者: Beatnik 时间: 2013-8-11 19:21 标题: *Not a Question about CFA*
Hi All - I am working for a firm managing retirement accounts. Our portfolio construction is largely ETFs and mutual funds. I was wondering if there was a way to perform value at risk and stress test analyses on these portfolios, i.e. is there an algorithm (I can program fairly well) or a reasonably-priced program that will accomplish this? Anyone know if this sort of analysis is helpful for these types of accounts?作者: Kapie 时间: 2013-8-11 19:23
I don’t think VAR or stress testing is relevant to your business. It’s only relevant to financial firms that need a minimum level of solvency, like banks.
If you’re talking about your own financial strength, you need to focus more on your P&L than your VAR.
If you’re talking about client balances, then these measures are less applicable than regular beta/stdevs.作者: tango_gs 时间: 2013-8-11 19:24
I’m probably using the wrong phrasing.
What I’d like to see us doing (and we’ve started doing this weekly) is to review all client accounts and to address any issues. For example, if we have client’s who are in a short bond position because of anticipated rate increases - we may want to liquidate that position if we think rates have maxed out.
IN ADDITION, I was hoping to do some sort of “worse case” scenario. For example, to say “how would these portfolios do if the 10-year treasury goes to 3.5%” - etc. I guess each security would need to be evaluated for duration, beta, etc.? I don’t even know how possible this is with mutual funds, as it would require constantly updating information that may not be immediately available (i.e. some funds update positions only quarterly).
It was just a thought, really.作者: dandman 时间: 2013-8-11 19:25
Yes you can do this, limited usefulness perhaps but it is possible. There is software to accomplish most of what you are probably after, ie-RiskMetrics, perhaps Factset. You can do stress testing, scenario analysis, and see VaR statistics.
A lot of this stuff is going to be based on historical return data, which is why I say it may not be as useful. You can also use bloomberg. I am not going to comment on how useful VaR and such items are, thats a different debate, but you can get at the info…
Greenman-disagree.VaR can provide data on expected fluctuations in a portfolio (I am 95% sure I shouldnt lose more than 2% in a day)….usefulness beyond banks. Same for stress testing (what should I expect my portfolio to do if there was a repeat of Sept 11th, or rates moved 100bps, etc)…useful beyond banks and not directly related to only solvency.作者: Zestt 时间: 2013-8-11 19:26
VaR can be useful for managing personal portfolios, but it is hard to explain what it means to clients who are not financial professionals, because saying “it means that you won’t lose more than 1.5%, 95% of the time” takes practice to understand (and because 5% of 250 trading days is still 12-13 days/year, or about once per month). It’s more useful as a way to try to convert your qualitative estimate of risk tolerance to some kind of quantitative threshold that you use for your own internal management, rather than a number you tell your client you are holding yourself to.
It shouldn’t be to difficult to do parametric var (where you assume something like a normal distribution of returns - or other distribution, but that does get more complex), because you just compute the standard deviation of the expected portfolio return and figure out where the 5% cutoff is (or whatever cutoff you are using). You need to know the mathematics, but the calculations aren’t all that difficult, particularly if you have some experience with matrix multiplication or a relatively small number of ETFs. You can also do historical VaR, which simply looks at the bottom X% of historically realized returns and figures out where the cutoff is. The challenge here is that some ETFs have very short histories, but you can often use the indexes that they track as a substitute.作者: BC_MBA_student 时间: 2013-8-11 19:27
Look at products/services offered by Factset and Moody’s Analytics… I’m sure there are others, but those two immediately came to mind.
