MaxiFi not yet supporting Monte Carlo

Tags: 

Hi - As a long-time ESPlanner user, I'm trying to decide when to switch over to MaxiFi, and like the sound of several of its new features. However the fact it does not yet support Monte Carlo has me concerned.

Without Monte Carlo, is Maxifi just assuming that you achieve the indicated "nominal rate of return" for each account type (regular account, retirement accounts) every single year with no deviation?

Second question: I'm assuming the non-Monte Carlo version of ESPlanner works the same way in this regard (i.e. when Monte Carlo is not being used) - is that correct?

Third question: any idea when the Monte Carlo version of MaxiFi will be out? I see your last release of MaxiFi was in February, with a note that Monte Carlo is the next major feature to add.

thanks much for your help.

John

Comments

dan royer's picture

Yes, ESPLanner in economics mode and MaxiFI ask you for the assumed average nominal rate of return. With MC enabled in either program, there is still an assumed average rate of return, but it shows you the spread or cone of variation that is possible/probable given your asset allocation. So both are 100% accurate as far as it goes; but MC answers different questions about upside and downside possibilities which you don't see without MC.

We believe we'll have MC in MaxiFi by the end of the year, and perhaps mid Fall.

We've not had a release recently only because this next release will be including so many new features.

Dan

Thanks Dan for your prompt response. Here is a related question:

I've always assumed that when running Monte Carlo, the Suggestions collection of reports is derived from the Monte Carlo analysis. Is that true?

If so, "how" is it derived? i.e. what specifically from the Monte Carlo analysis is driving what in order to set the Suggested Total Spending?

Ultimately I'm trying to understand how conservative vs. aggressive the Suggested Total Spending is...

thanks, John

dan royer's picture

Well, there's complicated math involved that I'm sure is over my head, but essentially with MC the rate of return that is assumed is derived from the asset allocation--that is, the average rate of return in the available history of the specific asset allocation. Of course the history of any asset or group of assets does not guarantee that will be the rate of return going into the future, but it's what we have to go on. In addition to the implicit rate of return in a collection of assets, there's also variance or standard deviation entailed in each asset or group of assets. This combination of rate of return and variance constitutes two important variables in an MC analysis. There's also some factors like coefficient variation which has, I believe, to do with the way different assets respond in an economic environment to each other.

But all of that is to say that what you get in an MC analysis is a calculation of the average rate of return for an asset group along with the variance of that group of assets. Then, the report shows the average but it's also able to show the probabilities of variance from that mean.

Is the historical average entailed in a group of assets and used by the program such as a 60/40 combo of international equity/long term bond, say 9%, more accurate than your stipulated assumption you might use instead in a non-MC use, say 6%? Who knows? The future will tell. It could be that the 9% will eventually show itself to be true over the next 40 years, but in the next 20 years the 6% is more correct. I'm just making these numbers up. In this is just how I think about it. I don't mind using MC but note that you can learn the average return on a combination of asset classes or specific tickers at https://portfoliovisualizer.com and use that rate of return in a non-MC run. It just won't show you the variance from the mean that you should expect. And even when I know my historical average is say 6% I still use 5% or 4.5% because I don't trust history to always prevail into the future.

That's my take! :)

Dan

OK... so it sounds like the Suggestions reports are generated in the same way regardless of whether you use Monte Carlo or not... it just uses different nominal rates of return (derived from your portfolio specification, instead of your simple entry of nominal return for each account type).

Calculating a 30+ year retirement based on average non-varying returns each year completely ignores sequence of returns risk... which is why people use Monte Carlo analysis.

So based on what you are telling me, since I have no interest in the simplistic average return-based calculations, I should ignore ALL of the Suggestions reports, and only look at the Monte Carlo reports. I should pick a percentile, probably from the portfolio trajectory data, and calculate my spending based on the living standard in that data.

Do you think I'm understanding correctly?

thanks, John

dan royer's picture

I don't think that's right. I may not be good at explaining it. You can call Kotlikoff around 5:00 ET today or later this evening or tomorrow. 617 834 2148

dan royer's picture

Larry says it actually does take account of the sequence of returns. It doesn't use an average of the returns all the way through time, but one year at a time based on a random draw, but of course there is an average implicit return when it's all said and done.

Dan, Thank you for the update on MaxiFi Planner. The Release Notes page on the MaxiFi Planner website refers to planned updates, but is dated February 2018, which makes it appear that development has stagnated and the product may not be viable. Highly recommend that the MFP website be updated to show existing and potential customers that MFP is being continuously improved. Thanks!

dan royer's picture

Right. Several people have asked if it's stagnated. It most certainly has not. These guys are full time on it every day, all week long. It's just that there are so many interruptions here in there etc. But I've seen charts built for the new monte carlo. I think it's going to be good. The next release will be a big, and then after that it will be back to a quarterly release schedule.
Dan

Dan, sorry I just now saw your reply - "Larry says it actually does take account of the sequence of returns...".

Does the "it" in this statement mean the Suggestions Reports? That would be good if true. Although this would raise another question: which portfolio trajectory (i.e. percentile) is chosen for the suggestions?

thanks, John

dan royer's picture

The antecedent for "it" was "Monte Carlo." :)

So the reports and annual suggested discretionary spending would be the mean (I'm not certain if the mean is the same as the 50th percentile in this case but it's close to it) taken from the hundreds of runs done by the MC. The MC graph will show the mean and the variation from the mean over time.

Dan

We use cookies to deliver the best user experience and improve our site.