Is it easy for a firm to integrate Monte Carlo VaR into its risk management process? Or does it take a lot of time and resources to set up?
Shweser notes page 107 book 4. Monte Carlo is difficult and expensive.
If you have installed a ERM system with a MC2.0 compatible docking station it will be up and running after a restart of the ERM mainframe. You just have to be online while you plug the Monte Carlo module into the dock. Remember to remove the transport screw. Monte Carlo vibrates a lot due to the high iteration-rpm, and the screw will destroy the module.
Could you tell me how long a rubber band is by the way? Just wondering…
#ironyindicatingunanswerablequestion
How long is a piece of string…
Wrt cost and benefit, one has to guess what answers CFAI want rather than what actually are in practice
On my laptop I can run a 100,000-iteration Monte Carlo simulation with 500 identified risks in less than 10 minutes.
Monte Carlo simulation being expensive is a thing of the past (though CFA Institute hasn’t yet realized that).
It can be difficult, especially when trying to determine the appropriate probability distributions and appropriate correlations for all of your risks.
I have an IOS Monte Carlo app too. This simulation provides estimates as correct as long as inputs are valid. Otherwise, you have a GIGO Monte Carlo estimations.
maybe i’m Just misremembering but I think this is a place where study materials have some inconsistencies. I think they generally say it’s expensive/hard to do MCS but not exclusively so. If a question comes up on the exam where this is important you have to guess a little bit about what CFAI wants or turn your real world brain off temporarily.
I believe that you’re correct.
lol i thought your not supposed to discuss the exam?
So then if a firm lacked said resources, would that be a case where it’s not applicable?
no one is. This is a general discussion of Monte Carlo simulation and comparing real world applicability vs CFAIs treatment of the topic.
I would think CFAI could be justified in the ‘Monte Carlo being expensive’ because a company would need a whole bunch of overpaid statisticians and macro-economists to somewhat reliably create a Monte Carlo model. Probably the company would need to keep the extra manpower to run the model check its validity, input factor weights… depending on the complexity of the scenario, a simple Monte Carlo app may not bring a comprehensive result.
what other measures would you consider appropriate if Monte Carlo was deemed expensive? I would have thought variance-covariance OK.
It depends on the asset. Var-Cov assumes a normal distribution. If the asset has a non-linear return structure (like options) then you can’t use it. You would have to use historical VaR.
Emipircal equity return has fat tail distribution
“a bond with credit risk can be viewed as a default-free bond plus an implicit short put option written by the bondholders for the stockholders.”
There are so many reasons you can argue variance corvariance method not applicable so you really need to guess under what context CFAI is testing
That’s better.
as far as I know, it’s possible to do Monte Carlo simulation on Excel (at least for simple probability distributions such as normal)
heck even I believe top softwares for MC simulations such as @risk is simply a software that automate Excel’s process or require less hassle to input.
I don’t believe that’s a “top software.” A statistical package would likely be better suited; SAS, and R come first to mind. There’s a lot of problems with using excel for statistical analysis, so it’s definitely worth looking into real stats programs.
I think the key point that hasn’t been mentioned is accuracy of Monte Carlo simulations. Monte Carlo is VERY noisy as an estimation process, and as a result, getting accurate results via Monte Carlo simulation can take a LOT of paths. That’s what I think people usually mean when they say Monte Carlo is expensive, and I think that’s a fairly accurate statement. If you need 100,000,000 simulations considering all the risk exposures on the firm to determine a VaR, you can see how that would become expensive.
That said, there are tricks to make estimates more accurate and thus less expensive, but there’s really no way around Monte Carlo being noisy.