***** question about monte carlo versus historicals

B

bulbar

Recycles dryer sheets
Joined
May 6, 2014
Messages
109
I ran ***** with about $3MM in assets, SS for me and the spouse out a ways, $130k/yr spending, 36 year span, 50/50 stocks/bonds. I chose to run it using Historical Data - All Years. It came back at a 92.6% success rate.

Then I ran it using Monte Carlo. I just used the default values (6.4% equity gain/1% st dev, 1.8% bond gains/5.1% st dev). The success rate dropped all the way down to 40%.

I'm not smart enough to know what that means and if that means my plan is a bust. Can anybody help?
 
monte carlo simulation probably doesn't have mean reversion programmed - you can get some wacky results with monte carlos - even just going 10 years out
 
Be sure to run any Monte Carlo sims several times as you can get a range of results.


If using MC, get an idea of the range of % success results - you can get a false sense of security if, by chance, you run it once and it shows 100% success, cuz the next time you run it, the results may be shockingly low.
 
It's back to 97% success under MC. I had some large "other incomes" coming in one year later later in the simulation that I think screwed it up. I simplified it to just one portfolio in year 1 and MC always comes back over 90%
 
I'd use Flexible Retirement Planner for MC simulations. It will run 2000 (I think) and give you the 10/90% bands as well as the number of times money actually ran out in those simulations and the %age in each year.
 
bulbar said:
I ran ***** with about $3MM in assets, SS for me and the spouse out a ways, $130k/yr spending, 36 year span, 50/50 stocks/bonds. I chose to run it using Historical Data - All Years. It came back at a 92.6% success rate.

Then I ran it using Monte Carlo. I just used the default values (6.4% equity gain/1% st dev, 1.8% bond gains/5.1% st dev). The success rate dropped all the way down to 40%.

I'm not smart enough to know what that means and if that means my plan is a bust. Can anybody help?
Click to expand...

the problem with historical is alot of it depends on time frames playing out in the same order.

each 30 year period is really dependent on exactly how the time frames before it and intersecting with it behaved.

as an example 1987 to 2003 saw the greates bull market ever . 17 years of almost 14% returns.

as pfau said "

Between 1926 and 2010, there are only 56 rolling 30-year periods. And as is about to be discussed, these 56 periods are not independent of one another. Meanwhile, it is not uncommon to see a Monte Carlo simulation study based on 10,000 simulated paths of financial market returns"


"because of the way that overlapping periods are formed with historical simulations, the middle part of the historical record plays an overly important role in the analysis. With data since 1926 and for 30-year retirement durations, 1926 appears in one rolling historical simulation, while 1927 appears in two (for the 1926 and 1927 retirees). This pattern continues until 1955, which appears in 30 simulations (the last year for the 1926 retiree through the first year for the 1955 retiree). The years 1955 through 1981 all appear in 30 simulated retirements"


wade pfau looked into this in one of his studies


http://wpfau.blogspot.com/2012/05/monte-carlo-simulations-vs-historical.html
 
Last edited:
mathjak107 said:
the problem with historical is alot of it depends on time frames playing out in the same order.

each 30 year period is really dependent on exactly how the time frames before it and intersecting with it behaved.

as an example 1987 to 2003 saw the greates bull market ever . 17 years of almost 14% returns.

as pfau said "

Between 1926 and 2010, there are only 56 rolling 30-year periods. And as is about to be discussed, these 56 periods are not independent of one another. ...
Click to expand...

IMO, this is the value of historical analysis, not a problem at all. I don't think market cycles are independent, that's why they are called cycles, so why use a model that treats them as independent?

We get bubbles, bubbles burst. Bad periods followed by recovery. Recovery followed by a slow down. Of course, a 'near-end-of-the-world' type scenario could occur, and there would be no recovery (in our lifetimes at least), but you can only plan for so much before you decide to work until you die.

I think MonteCarlo is the problem - it does not account for these cycles, or does so artificially in some way often only known to the programmer (and maybe not even to the programmer).

An ironic justification of MonteCarlo is when it is mentioned that the results are in-line with historical analysis! Then why not just use the historical analysis?

-ERD50
 
ERD50 said:
...

An ironic justification of MonteCarlo is when it is mentioned that the results are in-line with historical analysis! Then why not just use the historical analysis?

-ERD50
Click to expand...

As I understand it, and someone correct me if I'm wrong, the reason for using MC with historical is to account for sequence of returns. Pfau, Cotton, or Kitces (or someone) talks about this somewhere but I'm too lazy to search for it :p. I think it was Pfau that talks about fat tails accounted for in MC simulations, but I'm not sure. Dirk Cotton addresses strategies for overcoming SOR risk here:

The Retirement Cafe: RIIA Webinar -- Sequence of Returns Risk

To all of this, I personally would never be comfortable with a mere 100% success rate in any calculator. All calculators I use provide for generous heaping of $$ left over at end of plan, aka, "padding" (in addition to the padding in my budget).

Remember, we are engineering for the absolute worst case scenario. RIP provides results accounting or average, below average, and signficantly below average (which I use), as does Financial Engines, which I use free through VG. ESPlanner, if you use the most conservative mode, can provide results for earning returns just keeping with inflation in Monte Carlo, or complete loss of risk assets in upside planning (I model both of these modes as they are the most conservative in ESP). Firecalc uses historical, while you can fiddle with ******** to produce all kinds of juicy results.

In all probability, if we model conservatively, we will probably have more money than we know what to do with in later years. Visited the folks over the holidays and this is precisely the "predicament" they are in.
 
Last edited:
I find it intriguing how revered these probablaistic calculators seem to be. These are mere models of what we might expect. They are flawed from the outset in several important way:


1) They are based on historical data (either directly in historical methods or through estimating the paramets of a MC simulation)

2) That they don't model rational retiree behaviour. I don't think any reasonable person will continue to spend themselves into poverty as is depicted in these models.

These tools are rough measuring instruments. The give you some confidence that you are in the ballpark. Seeking 100% success in the model and lamenting anything less is a bit over the top IMHO. I rather like the idea of insuring your basic needs through pension/SS. etc as much as possible as the "padding" in the retirement plan and following a variable spending approach..
 
Last edited:
you do not want to introduce the variables of spending patterns into the equation. then you cannot compare or identify mathamatically what any of the results mean.

it is spending patterns that can make things better while the numbers themselves are based ON WORST CASE SCENARIOS.

that is kind of like having a built in buffer for those big unexpected expenses.

the fact we do not generally need anywhere near all that inflation adusting every year as we age on top of human spending patterns that have us cuting back by comfort level when markets are down help imorove success rates.


if anything if we have average markets and inflation it will be a pleasent upside surprice instead of a big disapointment if we don't and possibly catastrophic consequences.
 
Last edited:
the traditional studies did not take human spending into consideration.
many planners today use lifestyle planning . after taking the results from studies like the trinity software adjusts thing to fit your lifestyle.

want more early on for travel and less later? the sofware can give you a road map.
 
mathjak107 said:
the traditional studies did not take human spending into consideration.

...
Click to expand...

This is known. In fact the author of either the original Trinity study or Bengan's follow up study (forget which) later advised that one should be flexible and not take inflation adjusted withdrawals blindly. And yes, ******** can model this and other spending approaches (I particularly like this feature). I agree with your excellent way (IMO) of viewing spending in your above post:

it is spending patterns that can make things better while the numbers themselves are based ON WORST CASE SCENARIOS.

that is kind of like having a built in buffer for those big unexpected expenses.

the fact we do not generally need anywhere near all that inflation adusting every year as we age on top of human spending patterns that have us cuting back by comfort level when markets are down help imorove success rates.


if anything if we have average markets and inflation it will be a pleasent upside surprice instead of a big disapointment if we don't and possibly catastrophic consequences.
Click to expand...
 
I had done a Google search but did not see it so thanks for the link.

I have used FIRECalc in the past, but ***** appears to be different.
 
Thanks for the post, REW. I hadn't used ***** in a while, and when I went back to it, noticed it had improved a great deal. I believe the improvements in modelling were added after the 3/14 thread as I posted there and would have mentioned the improvements (which I really like), but did not at the time.
 
Raymond01 said:
I had done a Google search but did not see it so thanks for the link.

I have used FIRECalc in the past, but ***** appears to be different.
Click to expand...

:facepalm: Sorry Raymond , Crowdsourced Financial Independence and Early Retirement Simulator/Calculator , hope this is what you were looking for.
 
You must log in or register to reply here.

Similar threads

S
Hello from the sucker state! Looking for some kind comments; am I in the ballpark?
2
Replies
47
Views
4K
ivinsfan
ivinsfan
B
Monte Carlo or Historical in ***** and Firecalc?
2
Replies
37
Views
5K
samclem
samclem
M
using actual data or monte carlo sims-4% rule
2
Replies
43
Views
7K
mathjak107
M
Samz
FIRECalc v Fidelity RIP v Financial Mentor BRC
Replies
13
Views
3K
pb4uski
pb4uski
T
utility of monte carlo simulations
2
Replies
28
Views
4K
REWahoo
REWahoo

Latest posts

Back
Top Bottom