Firecalc -vs- Raymond James Monte Carlo

[dlgobeavs]

Greetings -

My wife and I are both 54, and seriously considering hanging things up within the next 12 months. I've been running Firecalc scenarios over the last few years, and we also work with a financial advisor from Raymond James.

I'd love some opinions on the following:

1. How do people weight the difference between Firecalc success rate -vs- Montecarlo percentages I am provided by RJ? Example is that we are at 100% with Firecalc, but range between 85% and 90% with the RJ Monte Carlo exercise.

2. Would you be comfortable with 85% on a Monte Carlo estimate? 90%??

Just looking for some groupthink advice.

Thanks in advance!!

 

[daver]

This Rob Berger video talks about success rates and difference between them.

 

[dlgobeavs]

This Rob Berger video talks about success rates and difference between them.

Thanks - great video... tons of info!

 

[Exchme]

The best way to make your retirement successful is get rid of fees charged by Raymond James. Advisory fees are a nightmare for your portfolio. Remember, you can only count on roughly 4% of your portfolio per year to live on, so if your advisor is taking 1%, that cuts your spending by 25%! Many retail brokers use funds that have high expense ratios compared to index funds, so their cut may be worse than that.

FireCalc tells you what did happen in the past, though it has limitations such as needing to estimate your taxes yourself. Monte Carlo modeling can be tweaked in lots of ways. Clearly, if your numbers are 100% in FireCalc but only 85-90% per RJ's Monte Carlo, then Raymond James is using lower than historical average returns and/or higher than historical average volatility.

Neither method represents the actual future of course, but it's hard to argue you don't have enough if FireCalc shows 100% success.

 

[Dash man]

I agree with dumping RJ. You can run Monte Carlo simulations yourself using online tools for much less than you pay RJ. If you lack confidence in what to invest in, just select a total market index fund while you start reading a lot (to learn) during your retirement. It’s really not that hard.

 

[Dtail]

Agree with @Exchme on getting rid of RJ.
In general, a Monte Carlo simulation will provide a lower result than an historical sequencing calculator. It can randomly apply a bunch of worst sequence of years that history has never experienced.
An example can be that that the losses of Years 2000 to 2002 can be followed up with the loss of 2008. History shows that this type of loss scenario has never happened and would most likely be unlikely to happen.
However there is value to using a Monte Carlo simulation in conjunction with an historical simulation.
I use Fidelity's Monte Carlo calculator and it usually 3 to 5% more conservative than Firecalc using a maximum spending concept.
Even if Fidelity provides a score of 100, it only equates to a 90% success rate, which is the highest success rate their program provides. RJ could have the same built in limitation of sorts.
It sounds like you are fine overall.

 

[momoney]

Even if Fidelity provides a score of 100, it only equates to a 90% success rate, which is the highest success rate their program provides. RJ could have the same built in limitation of sorts.
It sounds like you are fine overall.

I see scores over 100 in Fidelity when I run a retirement goal? What score in Fidelity would equate to 100% success?

 

[walkinwood]

By its nature, the monte carlo simulation results depend on the input parameters. Usually, for each asset class, you provide an average annual return and a standard deviation for the asset and an average inflation percentage. The result will also depend on the MC implementation (eg: does it vary inflation along with returns?). MC is a statistical tool & will give you slightly different results each time you run it.

Firecalc uses historical data, so you'll get the same results every time given the same inputs.

+1 on the Rob Berger video. The 50% success paper that he refers makes for very interesting reading.

Retirement planning is a continuous process. Best to revisit your calculations each year and make necessary adjustments. I've just started using Pralana Gold. For us, being ER'd for 16 years now, it is overkill, but I think it is a very well thought out tool & for about $100/yr, well worth it.

 

[Dash man]

I wouldn’t worry about scores 75 or higher. You could adjust your spending if needed to compensate if things started to look iffy. It’s not likely to have an event that will not be visible before you get into trouble. You have to monitor and make adjustments along the way. If anything catastrophic happens, we’ll all be in the same boat.

 

[Dtail]

I see scores over 100 in Fidelity when I run a retirement goal? What score in Fidelity would equate to 100% success?

So it works like this.
The score indicates how much maximum expenses one can spend.
So if your expenses are 50k yearly and your score is 150, then your maximum spending is 50*1.50 equals 75k yearly and still achieve a 90% success rate.
Their module is designed (and it is explained in their methodology) that using for example the "Significantly Below Average" module provides a 90% comfort or effectively success rate.
They do not have a module which provides a 100% success rate. So the maximum spending i.e. "Score" should be used in conjunction with their most conservative module (mentioned above) for one's analysis.

 

[Dtail]

By its nature, the monte carlo simulation results depend on the input parameters. Usually, for each asset class, you provide an average annual return and a standard deviation for the asset and an average inflation percentage. The result will also depend on the MC implementation (eg: does it vary inflation along with returns?). MC is a statistical tool & will give you slightly different results each time you run it.

Firecalc uses historical data, so you'll get the same results every time given the same inputs.

+1 on the Rob Berger video. The 50% success paper that he refers makes for very interesting reading.

Retirement planning is a continuous process. Best to revisit your calculations each year and make necessary adjustments. I've just started using Pralana Gold. For us, being ER'd for 16 years now, it is overkill, but I think it is a very well thought out tool & for about $100/yr, well worth it.

Just a slight note on the Fidelity MC module. If you run the same data 10 times, it will provide the same exact results each time.
Conceptually, that is not how a true MC module works. I did raise it with them and they acknowledged as such, but will not change it. However when I have used other MC modules, it does change each time, but a small insignificant amount.

 

[MuirWannabe]

As said above. Dump RJ. They are not helpful. Your chances of retirement success only go up if you do. Pay whatever fees you must to wrench yourself away from their clutches.

 

[Dtail]

By its nature, the monte carlo simulation results depend on the input parameters. Usually, for each asset class, you provide an average annual return and a standard deviation for the asset and an average inflation percentage. The result will also depend on the MC implementation (eg: does it vary inflation along with returns?). MC is a statistical tool & will give you slightly different results each time you run it.

Firecalc uses historical data, so you'll get the same results every time given the same inputs.

+1 on the Rob Berger video. The 50% success paper that he refers makes for very interesting reading.

Retirement planning is a continuous process. Best to revisit your calculations each year and make necessary adjustments. I've just started using Pralana Gold. For us, being ER'd for 16 years now, it is overkill, but I think it is a very well thought out tool & for about $100/yr, well worth it.

I have heard very good things on Pralana Gold. Super detailed but particularly better on the Roth conversion module than most.

 

[NgineER]

Greetings -

...
1. How do people weight the difference between Firecalc success rate -vs- Montecarlo percentages I am provided by RJ? Example is that we are at 100% with Firecalc, but range between 85% and 90% with the RJ Monte Carlo exercise.

2. Would you be comfortable with 85% on a Monte Carlo estimate? 90%??

I wouldn't be surprised if the 85-90% is due to RJ skimming off the top. I.e. they assume that you will pay them 1% in AUM and therefore you will have a lower success rate than you would with Firecalc.

 

[obsessed]

I wouldn't be surprised if the 85-90% is due to RJ skimming off the top. I.e. they assume that you will pay them 1% in AUM and therefore you will have a lower success rate than you would with Firecalc.

Sounds about right. I added an extra 1% in fees to the standard Firecalc parameters and the success rate dropped from 95% to 84% and their fees are probably higher.

 

[Koolau]

Sounds about right. I added an extra 1% in fees to the standard Firecalc parameters and the success rate dropped from 95% to 84% and their fees are probably higher.

Interesting - I would have thought it would have been even worse than that!

 

[FLSUnFIRE]

With FIRECalc, your probability is based off historical results. With Monte Carlo, what are you basing your input assumptions on? I'd probably use historical statistics myself but then the Monte Carlo isn't going to factor in all the correlations/interactions between PYs performance, inflation, business cycles, human psychology, etc that are baked into the historical record/the dataset FIRECalc uses. IMO, calculators based off history are the best option and MC is mostly a fun toy to play with.

The plan is just the plan and once excecuted (ER) one will adapt as the truth reveals itself. As time goes on I accept that I cannot predict or control the macro environment and my philosophy continues to shift towards Alfred E. Newman's.

 

[bltkmt]

I am in the exact same situation. 80% of my assets are at RJ, with guys I used to work with during my career. They charge me 40bps and I am happy to pay it. Their models show about 85% success rate while Firecalc is at 100%. It is just a different methodology.