Monte Carlo Simulations - What is considered a good result?
![PeterM](https://thebigtent.paraplannersassembly.co.uk/applications/dashboard/design/images/defaulticon.png)
Morning all
What would you consider a 'good' result for a Monte Carlo simulation for drawdown sustainability?
I've been asked to look at our drawdown review process and we currently have Cashcalc which has a useful Drawdown Monte Carlo simulator which I was thinking of using to provide a guide on sustainability.
However while the results are colour coded red, amber and green I can't find any guidance on what is considered 'good'.
Obviously a score of 80% is better than 60% but for clients a score of 80% might end up being a concern as they would be thinking about the 20% chance of the plan failing.
How does everybody deal with this?
Comments
Hi,
In our firm we also use CashCalc and their Monte Carlo simulation. You are right there is no rule or guidance as to what is good. I think we use it more to show the client how sustainable their desired income and generally if its around 65% then the advisers use this to advise the client on potential actions they can take now to make that income more achievable.
I guess there is not a really a 'good' result? Maybe it's more of discussion tool. But you do raise a good point.
Hi, this is a really good question.
Just my thoughts on it are
...
Monte Carlo uses normal distribution underlying the model so any distribution that sits at the 16% to 84% percentile (i.e 1 standard dev) you should expect circa 68% of outcomes, to fall within, i..e the average range.
Outcomes above these that that fall between 2% and 98%, are expected c95% of all outcomes to fall within, i.e those that fall between 2% to 16% are +84 and 98%. above average outomes.
Anything above this I would ignore.
However, don't stop here. You should then over lay your result with a confidence interval, which you adjust as follows:
Is the simulation running 100's or 1000's of simulations, if 100's reduce confidence.
Is the market forecast for next 2 or 3 years, good, poor or neutral, again if poor, reduce confidence,
Does the client have a good capacity for loss - again, adjust accordingly.
I prefer to use a mix of two concepts personally, the 5 number summary and a weaker powered probability model (chebyshevs inequality) as a guide, as these cash flow models for me are all to rosey cheeked, so use the range at the 25% percentile to 75%, as this is technically the interquartile range and where you would find at least 50% of the observations.
In short, the model is a simplification of the real world, and you should check with a market correction, based on clients risk profile as well and also check does the clients the income strategy sustain the clients income to beyond their average life expectancy until the age they have a 1 in 4 chance of living to.
Overall, if you dont have time for all the above the 25% and 75% percentile helps provide a cautious approach which you can adjust based on your confidence of their capacity for loss.