Im starting some research into risk management tools and was wondering if there was any alternative/complimentary tools to VAR as a portfolio risk management tool.
conditional var, maximum draw down distributions, loss maximization analysis, cmetrics, scenario analysis, etc etc etc. Each has benefits and draw backs. depends on what you want to do.
the question is what you want to accomplish. VaR is designed for situation where you want to convey a message (to senior management) about a large and complex portfolio. for a smaller portfolio (let's say 50 positions) you are better off with scenario analysis and greeks because the negatives of assumptions you have to make to calculate VaR often outweigh the positives of having an aggregated metric. in fact even in major banks the day-to-day risk management is often based mainly on greeks (vis-a-vis limits) than on VaR. VaR limits usually get set by calibration to some pre-agreed risk as measured by greeks... hope this helps.
Im by no means well versed in any of these tools, so please correct me if I am way off track here, but from my limited knowledge VAR would be greater as volatility increases. But what about if you were running a strategy that took advantage of volatility spikes where most of your activity would take place place just after a spike in volatility, it just seems to me that your VaR would be distorted by this spike in volatility, which would in turn be overstating your total risk. Now I understand that any measure of risk is going to have its limitations but I was just wondering if there was a more accurate measure of risk in a situation like this. Hope this makes at least some sense.....
Assuming you are using a variance/covariance style VaR method, VaR is completely inappropriate to measure the risk of a dynamic strategy that rebalances to vol. Unless you are trying to measure very short term VaR (day or shorter) as a proxy for gap risk. If your strategy is algorithmic, it might be possible to compute VaR using a monte carlo simulations method; otherwise, I might try returns decomposition as a first attempt.
Thanks sjfan, I think thats exactly the kind of response I am after. (its completely over my head at the moment though). What if we were using options as opposed to linear products? Would that affect things at all?
You could use a delta-gamma approximation to model options are linear payoffs; However, your basic problem remains - if your strategy is vol based, your VaR model doesn't explicitly model the relationship between the change in IV and price. Therefore, you'll lose a very important (and probably governing) risk source unless you are looking at VaR for very very short term gap risks. I would argue that a monte carlo method that models the underlying prices, and IV term structure is probably the best way to go; but it's not a technically simple task.
Hi sjfan, Can you elaborate on maximum draw down distribution? Is there an analytical way to find it or do you have to rely on simulations? What methods do you use? Any reference on this topic? Thanks,