Here's a statement of the obvious: The opinions expressed here are those of the participants, not those of the Mutual Fund Observer. We cannot vouch for the accuracy or appropriateness of any of it, though we do encourage civility and good humor.
Support MFO
Donate through PayPal
Why The 4% Retirement Rule Is Just A Starting Point
The 4% retirement rule is more than just a random percentage selected as a starting point when making that critical retirement date. It is based on multiple studies that included using Monte Carlo analyses that explored thousands of income scenarios. I know because I did not initially trust that key percentage so I did my own Monte Carlo analyses when making my own retirement decision. Those independent calculations generated very high portfolio and income survival likelihoods for the 4% drawdown strategy.
Of course, any calculation output, regardless of its complexity, is directly dependent upon the input values. I explored a range of many reasonable performance likelihoods. As expected, survival probabilities changed. But all my analyses projected comfortable income survival rates when withdrawals were centered on the 4% annual rate. At even a 5% drawdown rate, portfolio survival rates are still high, but the degradation in survival probabilities were impacted to a level that compromised my comfort level.
There are no perfect one size fits all correct drawdown rule, but the 4% target goal is not a bad point of departure. For everyone, a final number is dependent on special circumstances that include a retiree's discipline to stay with a plan, yet be flexible enough to make moderate adjustments if circumstances change. Circumstances will most likely change so a retiree must always remain alert.
Good luck to anyone making the retirement decision soon,
I was remiss when I failed to reference a typical Monte Carlo simulator that is freely available on the Internet. Sorry about that. Here is my correction:
This Vanguard Monte Carlo calculator is not necessarily the most flexible tool, but it is fast and easy to use. Many other simulators are also easily accessed. Please give them a try. For example, calculate the difference in portfolio survival for a 4% drawdown and compare it to a 5% drawdown; a huge difference.
@mjg the vanguard calculator is a nice tool. I'm at 3.5% drawdown and more than 90% probability it will last 30 years, which considering I am 67 is wishful thinking lol.
@slick- Maybe what we need is a version of a Monte Carlo calculator to determine the chances of our longevity, which we could then match against the financial Monte Carlo results. That might radically change one of the major inputs to the financial calculator: the amount of time we are going to need any income at all.
"The 4% retirement rule ... is based on multiple studies that included using Monte Carlo analyses"
What the WSJ article says: "Based on pioneering research in the early 1990s by William Bengen, then a financial planner in California, the so-called 4% rule states that retirees can pull about 4% annually from their nest egg (a figure Mr. Bengen eventually set at 4.5%), with a high probability that their savings will last 30 years."
Bengen's research had nothing to do with Monte Carlo analyses, such as "This Vanguard Monte Carlo calculator."
Bengen said so himself: "Let me add that I am a great admirer of Vanguard and their effort to serve investors well with low-cost, well-managed funds. I use their funds in my personal portfolio. But our approach to computing "safe" withdrawal rates ... is quite different." https://www.aaii.com/journal/article/insights-on-using-the-withdrawal-rule-from-its-creator
That excellent 2018 interview with Bengen was linked to by the WSJ article. To me what is interesting is not so much the 4% figure as why he picked a thirty year target. (FWIW, in 1994, the IRS joint life table for a couple of 65 year olds was 25 years. The current Table III shows 27.4 years for a 70 year old couple, so we can assume that a 65 year old couple would have at least a 30 year life expectancy per IRS today).
Also, in his followup that I quoted from above, he addressed @slick's 3.5%
In contrast, my methods use actual historical returns and inflation rates in the order in which they occurred. Vanguard's methods create sequences of returns and inflation which probably never happened in reality. As a result, they may generate "worst case" scenarios worse than anything that has ever happened, while my methods search for the worst case that has actually occurred.
Note also that Vanguard uses different asset classes than I do in my research.
When you change things like that, numbers can change radically.
Thanks for giving Monte Carlo a try. Given that you demonstrated an interest to explore this tool, I suggest you might like to examine yet another Monte Carlo code that offers more input flexibility. Here is a Link to one such code:
This version is from Portfolio Visualizer. Give it several runs. Don't expect the results to be exactly duplicated each time. That's the nature of Monte Carlo. It's an uncertain world. The range of outcomes hopefully will add to your confidence when making the big impact retirement decision.
Best Wishes
Hi Old Joe,
All types of Monte Carlo simulators are currently available on the Internet. I never explored life expectancy ones before your question, but they too exist. Here is one from Vanguard:
All types of Monte Carlo simulators are currently available on the Internet. I never explored life expectancy ones before your question, but they too exist. Here is one from Vanguard:
It calculates the probability of at least one surviving as: 100% - (probability of person A dying) x (probability of person B dying)
This calculation has the same flaw that many investment Monte Carlo simulations have. They assume that events are independent: there are no economic cycles; the death of one spouse has no effect on the life expectancy of the survivor.
The shortcoming of much that Mr. Bengen did and reported is that he used actual historical market returns, sometimes in the precise order in which these returns were registered. That's a very limiting method. The odds of that happening again are close to zero.
The beauty of Monte Carlo is that thousands of cases are randomly incorporated into the simulation. Potential investment outcomes and their variability can be studied in an organized manner. Indeed most haven't any chance or low odds of them ever happening. But the odds are not zero and the Monte Carl methodology represents them in a nearly correct percentage. Each Monte Carlo simulation of the exactly same input conditions will yield a slightly different, but similar, average outcome. That captures the nature of market uncertantiesl
Monte Carlo is a splendid tool to test portfolio robustness against the unpredictable swings of the marketplace over time. Luck is always a player, and Monte Carlo gives some numerical measure of a portfolio to survive when exposed to these uncertainties.
Monte Carlo simulations continue to grow in popularity. Today, it is a common practice for professional financial advisors to use Monte Carlo to stress test the survival odds of a portfolio over a long timeframe. Wise investors take advantage of this useful tool too.
You're missing a number of points and writing what it seems you want to be true: that the 4% figure "is based on multiple studies that included using Monte Carlo analyses". It wasn't.
If what you meant to say is that subsequent Monte Carlo simulations validated this figure, then there's a different problem with the narrative. Because that would also validate the use of historical data - something you say has an intrinsic shortcoming.
Of course the odds are virtually nil that the next thirty years will match a previous thirty year period. Just as the odds are virtually nil that the next thirty years will match a performance pulled out of a hat (aka a Monte Carlo iteration). This is a red herring.
In the typical Monte Carlo simulation, patterns are abstracted away. You seem to regard this as a virtue, writing disapprovingly that historical returns are used "sometimes in the precise order in which these returns were registered." (Orderings weren't preserved merely "sometimes" but always when Bengen came up with his 4% figure. See his original paper.)
Again I suggest reading the AAII piece. You'll find a concrete example of how ignoring some patterns can affect results. Bengen notes there that if one rebalances much less frequently than yearly, " you can actually add about a quarter of a percentage point to your withdrawal rate" He attributes this to persistence of performance. That's a kind of pattern that simplistic Monte Carlo simulations abstract away.
"Monte Carlo simulations continue to grow in popularity." When all else fails, cite popularity for validation. I'm sure VHS's popularity meant that it was the superior technology, that the more popular Windows is better than Mac, etc.
There really was some interesting stuff that you didn't discuss. Like how "there is an inverse relationship between the long-term valuation of the stock market and how much retirees can withdraw without running out of money."
How does that historical data fit into your Monte Carlo simulations? How do you map CAPE into means and standard deviations for large cap stocks, small cap stocks, and bonds? Those are the inputs for the simulators you're linking to.
I think Monte Carlo engines are fine tools. Just so long as they're not simplistic, matched to the right task, and employed by knowledgeable users. Used as you suggest, they have lots of issues.
There are no constraints to Monte Carlo simulation, only constraints users create in a model (or constraints that users are forced to deal with when using someone else’s model). Non-normal asset-class returns and autocorrelations can be incorporated into Monte Carlo simulations, albeit with proper care.
I'm hearing all that is wrong with Monte Carlo simulation. So, @msf, @Old_Joe, tell me a better, easy to use tool for getting an idea where you stand with retirement spending potential. Show me something else that will "estimate" the probability of your retirement money lasting as long as you do. Show me another tool that shows this type of guidance.
There are no guaranties these simulations will play-out. Nobody thinks they are infallible. But they sure as heck are better than hope and a prayer. I side with MJG on there usefulness for guidance.
@MikeM - It sounds like you're saying that you use these simplistic Monte Carlo simulation tools because "they sure as heck are better than hope and a prayer", and not because they are superior tools.
What Bengen did, you or anyone else acquainted with spreadsheets could also do. Plus, you'd get a better understanding of what that model represents and what the numbers mean.
What inputs did you use to run your simulations (mean returns, inflation, etc.)? Why did you pick those figures? How much confidence do you have in them? What cutoff did you use for an acceptable success score?
Did your choice of cutoff incorporate the fact that in the real world (unlike in the simplistic Monte Carlo models), returns have fat tails? That suggests taking with a grain of salt MJG's statement that "most [outcomes] haven't any chance or low odds of them ever happening."
Are these tools you're lauding simple to use? Sure. You can let them default instead of grappling with the questions I asked above. Are they easy to use well and interpret accurately? I'm not so sure. They're great for learning through trial and error. Not so hot for predicting.
Blanchett and Pfau (whom I cited with approval) agree with you that these simulations are better than nothing: "if the calculated success rate is low, say 15-25%, we know that a client plan generating these numbers is in danger and that the advisor and client should immediately consider plan revisions. Likewise, if the calculated probability of success is high, say 80% or greater, we know that the plan is on the right track and that the client can proceed as indicated."
MJG says these are superior tools. I say, and apparently you say, that they're okay for hitting the side of a barn. Still they seem no better, and for some reasons possibly worse, than using that spreadsheet I mentioned above.
If you want other ideas, just read the comments section of the AAII article. You don't even need to read the article itself (though it's well worth the time); Bengen offered additional thoughts in the comments section, including thoughts about other tools.
Who was first in discovering the 4% drawdown rule is not significant to me. My number one purpose in my submittals on this topic was to introduce Monte Carlo simulators to those MFOers who might not be familiar with this powerful decision enhancing tool.
Even the last reference you quoted acknowledged its usefulness. In the opening statements they said: "Monte Carlo simulations will illuminate the nature of that uncertainty, but only if advisors understand how it should be applied – and its limitations". All tools have identical limitations and misusing a tool is certainly dangerous stuff. Nothing new to these constraints.
I am an experienced user of Monte Carlo analyses. In the nuclear engineering discipline (one of my masters degrees is in that field), it is a frequently used tool.
When making my retirement decision I generated my own Monte Carlo code to aid in the decision process. I recognized the shortcoming of using a Normal distribution for returns (it's not too bad but it is imperfect) so I coupled that distribution with my model for much less frequently occurring fat tails. I am not an unsophisticated user of this fine modeling concept.
There are many strategies when investing. We are different investors because of numerous critical factors including our needs, our timeframes, our knowledge, and our risk aversion. Different strokes for different folks certainly applies here.
A Monte Carlo code is one way to approach the uncertainties of market returns. From my perspective it seems like an ideal tool. You take issue which is ok by my standards. That's what makes for a vibrant marketplace.
I especially like Monte Carlo simulations because what-if scenarios can be postulated and evaluated in just short minutes. That's powerful stuff given the uncertainties of market returns. Nobody can forecast future returns with any accuracy and/or consistency. Monte Carlo helps to quantify the impact of such uncertainty on portfolio,survival odds.
Apparently you are not a fan of the Monte Carlo based approach. I am. Again, different strokes for different folks. In investing we get to choose our own poison.
@MikeM - Hi there Mike- I think that msf has pretty well covered the general response to your questions. Because Monte Carlo simulation was not widely available during our working years, I designed (took me a while, I'll tell you) a "predictive" spreadsheet which I integrated into the general financial spreadsheet which I had used for some years (and still use) to keep track of our various investments.
The "predictive" section took account of all resources which would be available to my wife and I after retirement: pensions, SS, Medicare, and income or value increase in investments of equity vehicles, bond vehicles, and real estate. Likewise we had excellent data which had been accumulated over a number of years with respect to anticipated expenses, broadly classified as "basic" (unavoidable), discretionary, and emergency.
Each of those variables was referenced to large tables which were set up to independently run compounded values over 35 years. Independent inputs for variables included inflation rate, rate of return on equities, bonds and cash (CDs and savings) accounts, and a "financial disaster" input which introduced a general meltdown variable selectable for any given year in the 35 year stretch.
By varying each of those inputs in any desired combination it was possible to see the cascading effect of various disaster scenarios occurring at different selected times. For example, I generally ran cash and bond income at 2% below the inflation rate, which was also a selected variable, and equity income at 2% above. Very conservative. Being a pessimist by nature, I generally ran setups which would cover every bad thing happening that I could imagine.
As it happened, I wasn't too far off in the predictive timing for disaster. Destruction of financial resources will be most influential the earlier that they happen in retirement, as they can set back the entire cumulative compounding effects quite seriously. Indeed, Murphy struck, in 2007/2008, just after retirement.
Nevertheless, the tables worked out pretty well. By pulling down our discretionary expenses (another independent variable input) we survived the chaos in good shape, and were able to carry on with no huge impact to our retirement mode.
Edit/add: I should also mention that we were deliberately in good shape with respect to loans and finance charges: 30 year mortgage @8% had been paid off ten years early, never any credit card or other interest expenses. (Once the mail was late with a credit card payment and it cost something like $4.21. My wife still mentions that occasionally.)
MGJ dismissed this whole effort rather casually with a reference to the "limitations" of a spreadsheet for these purposes, and endless exaltations and paeans as to the superiority and invincibility of Monte Carlo. He's welcome to his opinion: just take it with a large grain of whatever. He seems to be one of those folks who believe that whatever they do is the right and only way, that anything else is highly suspect or at best barely acceptable, and thoroughly enjoy telling you so. I'm sure that you know the type.
I often post that everyone gets to choose their own poison.
Like you, when I was preparing for retirement, Monte Carlo codes that addressed retirement issues were not readily available. That's why I assembled my own code.
It appears to me that your spreadsheets really are a subset of what Monte Carlo analyses do. Whereas you or anyone else have limited time to complete only a small number of these calculations, a Monte Carlo tool completes thousands of what could be identical computations. And it produces these countless simulations without a bias.
The bias and simplifications assumed by any hand calculation can compromise the results. For example, you reported that you assumed fixed relationships between bond and equity returns compared to inflation rates. Why you selected those fixed values puzzle me. The historical data does not support such an assumption. Actual historical data of these outcomes vary all over the map, are not constant, and not predictable. When that is the real world situation, Monte Carlo analyses shine.
I reject your assertion that I support the invincibility of Monte Carlo. I freely acknowledge its shortcomings. Also, I never tell investors what to do. In general, I merely provide references to tools that an investor might find useful when making investment decisions.
Once again, different strokes for different folks. We get to choose our own poison. You consistently misrepresent and exaggerate my true position. I do speak in a poaitive voice to support my positions. What is unexpected about that? Regardless, my......
"I generally ran cash and bond income at 2% below the inflation rate, which was also a selected variable, and equity income at 2% above. Very conservative."
@MJG: Did you perhaps overlook the "very conservative" qualifier?
I could care less that the "Actual historical data of these outcomes vary all over the map, are not constant, and not predictable." I deliberately projected a very pessimistic set of values which would emulate long periods of minimal returns, and threw in a meltdown scenario for good measure. Our plan survived.
You seem to be one of those folks who believe that whatever they do is the right and only way, that anything else is highly suspect or at best barely acceptable, and thoroughly enjoy telling us so.
@ MFO Members: Hard to convince someone of your position when you dealing with not only a clown but an id--- !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! Regards, Ted
"I am an experienced user of Monte Carlo analyses. In the nuclear engineering discipline (one of my masters degrees is in that field), it is a frequently used tool."
@MJG: I expect that the nuclear engineers responsible for 3-mile Island, and Chernobyl were also "experienced user[s] of Monte Carlo analyses".
Hi Joe. I built a very similar spreadsheet, probably about 15-18 years ago, and have upgraded it over the years with a bunch of what-if statements, like what if I pulled from the nest egg and bought an annuity in such and such year, what if I take my pension as a lump sum or in monthly life-long payments to compare scenarios for when to retire, what if inflation is 2%, 4% or more in different decades, ect, ect... 3 different compare tables showing the value of your nest egg in 3 comparable charts. You guys are right on this. Doing it yourself and understand each calculation is a huge benefit, if not greater reassurance in it's ability to predict.
Not bragging, but most people cannot do the same with Excel. That is why I think the Monte Carlo simulation is a great tool for most people, not as an exact answer but a ball park. That's all I was trying to say.
Take away MJG's annoyingness, (which we agree on, hmm, spell check says annoyingness is not a word) in my opinion people could and should be using something, and if you aren't a spreadsheet builder yourself I'm not sure there is anything better than Monte Carlo to get a ball park for probability to sustain.
Your most recent post clearly implies that somehow the sad facility failures that you identified were caused by shortfalls in their design as completed by nuclear engineers with a propensity for Monte Carlo analyses. That's a huge and unsupported extrapolation even for you. Evidence does not justify your claim. Systems erode over time.
If you don't recognize the benefits of Monte Carlo analyses (and there are instances when it is not appropriate) that's ok for me. I do and still use that tool. As I have consistently said: we each get to choose our own poison.
Mistakes happen and I am surely not immune to making my fair share of them. Monte Carlo tools have become a common tool these days in the investment community, They have the potential to reduce investment decision errors. My reference to Vanguard is just one of many serious advocates for deploying Monte Carlo as part of the investment decision making process.
@MikeM- Thanks, Mike. I completely agree that Monte Carlo surely can't hurt, and if that's all that you can use, it's much better than doing nothing.
@MJG: "facility failures that you identified were caused by shortfalls in their design as completed by nuclear engineers with a propensity for Monte Carlo analyses."
No sir, not at all. Merely indicating that a masters in nuclear engineering is a guarantee of exactly nothing. Incidentally, I may be expecting too much here, but I would at least hope that a nuclear engineer would anticipate "erosion" and design against that possibility.
@Ted: Not sure what you mean. I've certainly never suggested that MJG is either a clown or an "id--- !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!" The only MFO poster that I've ever called a clown is that guy who thinks that the rules are for everyone but him, cheats so as to kick undeserving posts into the "Comments +" category, and then leaves footprints that a kid could detect. But you already knew that.
@MFO Members: My message was not directed at MJG, sorry for the confusion . It was directed at the individual who's yet to contribute anything positive to the discussion board but finding fault with others. Regards, Ted
Who was first in discovering the 4% drawdown rule is not significant to me. My number one purpose in my submittals on this topic was to introduce Monte Carlo simulators
Obviously not significant to you, as you didn't introduce Bengen. Rather, you injected a claim that the 4% rule was based in part on Monte Carlo analyses. As you stated above, your purpose in submitting this false statement was to bridge from the WSJ article to what you wanted to write about.
Even the last reference you quoted acknowledged its usefulness. Already addressed in my post prior to the one I'm responding to here. Useful, yes, it's better than a poke in the eye. . All tools have identical limitations If all tools have identical limitations, then why push just a single tool? Previously you claim to have identified a shortfall in tools that use historical data. Shortfall, limitation. What's the difference? It seems tools don't all have identical limitations.
I am not an unsophisticated user of this fine modeling concept. ... I am an experienced user of Monte Carlo analyses. In the nuclear engineering discipline (one of my masters degrees is in that field), it is a frequently used tool.
How does the fact that a tool is used in an engineering discipline built on physical rules make it an effective tool in a field (financial planning) that lacks a similar foundation? The observation that you and some engineers frequently use this tool (regardless of the domain) is not is a good argument that it is well suited for the unsophisticated, inexperienced user.
From my perspective it seems like an ideal tool.
Ideal, yet something with faults that could be eliminated: "I generated my own Monte Carlo code to aid in the decision process. I recognized the shortcoming of ..."
To summarize, starting with a false statement, you suggested that people use tools with model shortcomings that you found unacceptable for your own use. Tools that make sense to you, because you have a masters of nuclear engineering.
>> How does the fact that a tool is used in an engineering discipline built on physical rules make it an effective tool in a field (financial planning) that lacks a similar foundation?
Many engineering disciplines are more like financial planning, and vice-versa, than you may realize. An awful lot of judgment and choice and sketchy causality entailed.
Oh-oh. I just lost a whole lot of faith in engineering. If the guy who engineered that bridge in Genoa is still around maybe he can find a new career as a financial planner.
haha, it is just that in systems it becomes so multivariate so quickly, and the physical rules interact in subtle ways (in addition to the clear ones), and always with the unforeseen and unintended consequences and interactions. I should add experience to my list of entailments.
I just checked Merriam-Webster for their take on "high-falutin", and danged if that don't describe ol' MJG pretty good. "pretentious, fancy; expressed in or marked by the use of language that is elaborated or heightened by artificial or empty means : pompous…"
You clearly don't think much of the "way" I suggested that investors might find Monte Carlo analyses tools useful when making investment decisions. But what do you think of Monte Carlo for that purpose independent of the way I presented the suggestion? That's the important investment point of my submittal, not the way I write.
You are welcome to whatever you decide on any subject. I will never challenge that. I like Monte Carlo methods when examining highly uncertain outcomes. That's just my opinion based on both engineering and investing experiences. It is not shocking that others do not share the same opinions. That's partly what makes for a divergent and challenging investing marketplace.
Thank you for reading my posts with such attention, but not with such a picayune inclination. Sorry, I already said that.
Comments
The 4% retirement rule is more than just a random percentage selected as a starting point when making that critical retirement date. It is based on multiple studies that included using Monte Carlo analyses that explored thousands of income scenarios. I know because I did not initially trust that key percentage so I did my own Monte Carlo analyses when making my own retirement decision. Those independent calculations generated very high portfolio and income survival likelihoods for the 4% drawdown strategy.
Of course, any calculation output, regardless of its complexity, is directly dependent upon the input values. I explored a range of many reasonable performance likelihoods. As expected, survival probabilities changed. But all my analyses projected comfortable income survival rates when withdrawals were centered on the 4% annual rate. At even a 5% drawdown rate, portfolio survival rates are still high, but the degradation in survival probabilities were impacted to a level that compromised my comfort level.
There are no perfect one size fits all correct drawdown rule, but the 4% target goal is not a bad point of departure. For everyone, a final number is dependent on special circumstances that include a retiree's discipline to stay with a plan, yet be flexible enough to make moderate adjustments if circumstances change. Circumstances will most likely change so a retiree must always remain alert.
Good luck to anyone making the retirement decision soon,
Best Wshes
I was remiss when I failed to reference a typical Monte Carlo simulator that is freely available on the Internet. Sorry about that. Here is my correction:
https://retirementplans.vanguard.com/VGApp/pe/pubeducation/calculators/RetirementNestEggCalc.jsf
This Vanguard Monte Carlo calculator is not necessarily the most flexible tool, but it is fast and easy to use. Many other simulators are also easily accessed. Please give them a try. For example, calculate the difference in portfolio survival for a 4% drawdown and compare it to a 5% drawdown; a huge difference.
Best Wishes
What the WSJ article says: "Based on pioneering research in the early 1990s by William Bengen, then a financial planner in California, the so-called 4% rule states that retirees can pull about 4% annually from their nest egg (a figure Mr. Bengen eventually set at 4.5%), with a high probability that their savings will last 30 years."
Bengen's research had nothing to do with Monte Carlo analyses, such as "This Vanguard Monte Carlo calculator."
Bengen said so himself: "Let me add that I am a great admirer of Vanguard and their effort to serve investors well with low-cost, well-managed funds. I use their funds in my personal portfolio. But our approach to computing "safe" withdrawal rates ... is quite different."
https://www.aaii.com/journal/article/insights-on-using-the-withdrawal-rule-from-its-creator
That excellent 2018 interview with Bengen was linked to by the WSJ article. To me what is interesting is not so much the 4% figure as why he picked a thirty year target. (FWIW, in 1994, the IRS joint life table for a couple of 65 year olds was 25 years. The current Table III shows 27.4 years for a 70 year old couple, so we can assume that a 65 year old couple would have at least a 30 year life expectancy per IRS today).
Also, in his followup that I quoted from above, he addressed @slick's 3.5% When you change things like that, numbers can change radically.
Thanks for giving Monte Carlo a try. Given that you demonstrated an interest to explore this tool, I suggest you might like to examine yet another Monte Carlo code that offers more input flexibility. Here is a Link to one such code:
https://www.portfoliovisualizer.com/monte-carlo-simulation#analysisResults
This version is from Portfolio Visualizer. Give it several runs. Don't expect the results to be exactly duplicated each time. That's the nature of Monte Carlo. It's an uncertain world. The range of outcomes hopefully will add to your confidence when making the big impact retirement decision.
Best Wishes
Hi Old Joe,
All types of Monte Carlo simulators are currently available on the Internet. I never explored life expectancy ones before your question, but they too exist. Here is one from Vanguard:
https://personal.vanguard.com/us/insights/retirement/plan-for-a-long-retirement-tool
I have no idea how good or reliable it is, but it is probably a respectable starting point. Good luck and a long life expectancy.
Best Wishes
The Vanguard page doesn't do simulations. It basically just does table lookups: "Calculations are based on mortality data from the Society of Actuaries Retirement Participant 2000 Table."
https://www.soa.org/experience-studies/2000-2004/research-rp-2000-mortality-tables/
It calculates the probability of at least one surviving as:
100% - (probability of person A dying) x (probability of person B dying)
This calculation has the same flaw that many investment Monte Carlo simulations have. They assume that events are independent: there are no economic cycles; the death of one spouse has no effect on the life expectancy of the survivor.
The Science of Longtime Couples Who Die Close Together
https://www.thecut.com/2015/11/science-of-longtime-couples-who-die-together.html
The shortcoming of much that Mr. Bengen did and reported is that he used actual historical market returns, sometimes in the precise order in which these returns were registered. That's a very limiting method. The odds of that happening again are close to zero.
The beauty of Monte Carlo is that thousands of cases are randomly incorporated into the simulation. Potential investment outcomes and their variability can be studied in an organized manner. Indeed most haven't any chance or low odds of them ever happening. But the odds are not zero and the Monte Carl methodology represents them in a nearly correct percentage. Each Monte Carlo simulation of the exactly same input conditions will yield a slightly different, but similar, average outcome. That captures the nature of market uncertantiesl
Monte Carlo is a splendid tool to test portfolio robustness against the unpredictable swings of the marketplace over time. Luck is always a player, and Monte Carlo gives some numerical measure of a portfolio to survive when exposed to these uncertainties.
Monte Carlo simulations continue to grow in popularity. Today, it is a common practice for professional financial advisors to use Monte Carlo to stress test the survival odds of a portfolio over a long timeframe. Wise investors take advantage of this useful tool too.
Best Wishes
If what you meant to say is that subsequent Monte Carlo simulations validated this figure, then there's a different problem with the narrative. Because that would also validate the use of historical data - something you say has an intrinsic shortcoming.
Of course the odds are virtually nil that the next thirty years will match a previous thirty year period. Just as the odds are virtually nil that the next thirty years will match a performance pulled out of a hat (aka a Monte Carlo iteration). This is a red herring.
In the typical Monte Carlo simulation, patterns are abstracted away. You seem to regard this as a virtue, writing disapprovingly that historical returns are used "sometimes in the precise order in which these returns were registered." (Orderings weren't preserved merely "sometimes" but always when Bengen came up with his 4% figure. See his original paper.)
Again I suggest reading the AAII piece. You'll find a concrete example of how ignoring some patterns can affect results. Bengen notes there that if one rebalances much less frequently than yearly, " you can actually add about a quarter of a percentage point to your withdrawal rate" He attributes this to persistence of performance. That's a kind of pattern that simplistic Monte Carlo simulations abstract away.
"Monte Carlo simulations continue to grow in popularity." When all else fails, cite popularity for validation. I'm sure VHS's popularity meant that it was the superior technology, that the more popular Windows is better than Mac, etc.
There really was some interesting stuff that you didn't discuss. Like how "there is an inverse relationship between the long-term valuation of the stock market and how much retirees can withdraw without running out of money."
How does that historical data fit into your Monte Carlo simulations? How do you map CAPE into means and standard deviations for large cap stocks, small cap stocks, and bonds? Those are the inputs for the simulators you're linking to.
I think Monte Carlo engines are fine tools. Just so long as they're not simplistic, matched to the right task, and employed by knowledgeable users. Used as you suggest, they have lots of issues. David Blanchett and Wade Pfau,The Power and Limitations of Monte Carlo Simulations, 2014.
https://www.advisorperspectives.com/articles/2014/08/26/the-power-and-limitations-of-monte-carlo-simulations
Been telling MJG that for more years than I can count. Good luck getting him to listen.
There are no guaranties these simulations will play-out. Nobody thinks they are infallible. But they sure as heck are better than hope and a prayer. I side with MJG on there usefulness for guidance.
What Bengen did, you or anyone else acquainted with spreadsheets could also do. Plus, you'd get a better understanding of what that model represents and what the numbers mean.
What inputs did you use to run your simulations (mean returns, inflation, etc.)? Why did you pick those figures? How much confidence do you have in them? What cutoff did you use for an acceptable success score?
Did your choice of cutoff incorporate the fact that in the real world (unlike in the simplistic Monte Carlo models), returns have fat tails? That suggests taking with a grain of salt MJG's statement that "most [outcomes] haven't any chance or low odds of them ever happening."
Are these tools you're lauding simple to use? Sure. You can let them default instead of grappling with the questions I asked above. Are they easy to use well and interpret accurately? I'm not so sure. They're great for learning through trial and error. Not so hot for predicting.
Blanchett and Pfau (whom I cited with approval) agree with you that these simulations are better than nothing: "if the calculated success rate is low, say 15-25%, we know that a client plan generating these numbers is in danger and that the advisor and client should immediately consider plan revisions. Likewise, if the calculated probability of success is high, say 80% or greater, we know that the plan is on the right track and that the client can proceed as indicated."
MJG says these are superior tools. I say, and apparently you say, that they're okay for hitting the side of a barn. Still they seem no better, and for some reasons possibly worse, than using that spreadsheet I mentioned above.
If you want other ideas, just read the comments section of the AAII article. You don't even need to read the article itself (though it's well worth the time); Bengen offered additional thoughts in the comments section, including thoughts about other tools.
Who was first in discovering the 4% drawdown rule is not significant to me. My number one purpose in my submittals on this topic was to introduce Monte Carlo simulators to those MFOers who might not be familiar with this powerful decision enhancing tool.
Even the last reference you quoted acknowledged its usefulness. In the opening statements they said: "Monte Carlo simulations will illuminate the nature of that uncertainty, but only if advisors understand how it should be applied – and its limitations". All tools have identical limitations and misusing a tool is certainly dangerous stuff. Nothing new to these constraints.
I am an experienced user of Monte Carlo analyses. In the nuclear engineering discipline (one of my masters degrees is in that field), it is a frequently used tool.
When making my retirement decision I generated my own Monte Carlo code to aid in the decision process. I recognized the shortcoming of using a Normal distribution for returns (it's not too bad but it is imperfect) so I coupled that distribution with my model for much less frequently occurring fat tails. I am not an unsophisticated user of this fine modeling concept.
There are many strategies when investing. We are different investors because of numerous critical factors including our needs, our timeframes, our knowledge, and our risk aversion. Different strokes for different folks certainly applies here.
A Monte Carlo code is one way to approach the uncertainties of market returns. From my perspective it seems like an ideal tool. You take issue which is ok by my standards. That's what makes for a vibrant marketplace.
I especially like Monte Carlo simulations because what-if scenarios can be postulated and evaluated in just short minutes. That's powerful stuff given the uncertainties of market returns. Nobody can forecast future returns with any accuracy and/or consistency. Monte Carlo helps to quantify the impact of such uncertainty on portfolio,survival odds.
Apparently you are not a fan of the Monte Carlo based approach. I am. Again, different strokes for different folks. In investing we get to choose our own poison.
Best Wishes
The "predictive" section took account of all resources which would be available to my wife and I after retirement: pensions, SS, Medicare, and income or value increase in investments of equity vehicles, bond vehicles, and real estate. Likewise we had excellent data which had been accumulated over a number of years with respect to anticipated expenses, broadly classified as "basic" (unavoidable), discretionary, and emergency.
Each of those variables was referenced to large tables which were set up to independently run compounded values over 35 years. Independent inputs for variables included inflation rate, rate of return on equities, bonds and cash (CDs and savings) accounts, and a "financial disaster" input which introduced a general meltdown variable selectable for any given year in the 35 year stretch.
By varying each of those inputs in any desired combination it was possible to see the cascading effect of various disaster scenarios occurring at different selected times. For example, I generally ran cash and bond income at 2% below the inflation rate, which was also a selected variable, and equity income at 2% above. Very conservative. Being a pessimist by nature, I generally ran setups which would cover every bad thing happening that I could imagine.
As it happened, I wasn't too far off in the predictive timing for disaster. Destruction of financial resources will be most influential the earlier that they happen in retirement, as they can set back the entire cumulative compounding effects quite seriously. Indeed, Murphy struck, in 2007/2008, just after retirement.
Nevertheless, the tables worked out pretty well. By pulling down our discretionary expenses (another independent variable input) we survived the chaos in good shape, and were able to carry on with no huge impact to our retirement mode.
Edit/add: I should also mention that we were deliberately in good shape with respect to loans and finance charges: 30 year mortgage @8% had been paid off ten years early, never any credit card or other interest expenses. (Once the mail was late with a credit card payment and it cost something like $4.21. My wife still mentions that occasionally.)
MGJ dismissed this whole effort rather casually with a reference to the "limitations" of a spreadsheet for these purposes, and endless exaltations and paeans as to the superiority and invincibility of Monte Carlo. He's welcome to his opinion: just take it with a large grain of whatever. He seems to be one of those folks who believe that whatever they do is the right and only way, that anything else is highly suspect or at best barely acceptable, and thoroughly enjoy telling you so. I'm sure that you know the type.
Regards
OJ
I often post that everyone gets to choose their own poison.
Like you, when I was preparing for retirement, Monte Carlo codes that addressed retirement issues were not readily available. That's why I assembled my own code.
It appears to me that your spreadsheets really are a subset of what Monte Carlo analyses do. Whereas you or anyone else have limited time to complete only a small number of these calculations, a Monte Carlo tool completes thousands of what could be identical computations. And it produces these countless simulations without a bias.
The bias and simplifications assumed by any hand calculation can compromise the results. For example, you reported that you assumed fixed relationships between bond and equity returns compared to inflation rates. Why you selected those fixed values puzzle me. The historical data does not support such an assumption. Actual historical data of these outcomes vary all over the map, are not constant, and not predictable. When that is the real world situation, Monte Carlo analyses shine.
I reject your assertion that I support the invincibility of Monte Carlo. I freely acknowledge its shortcomings. Also, I never tell investors what to do. In general, I merely provide references to tools that an investor might find useful when making investment decisions.
Once again, different strokes for different folks. We get to choose our own poison. You consistently misrepresent and exaggerate my true position. I do speak in a poaitive voice to support my positions. What is unexpected about that? Regardless, my......
Best Wishes
"I generally ran cash and bond income at 2% below the inflation rate, which was also a selected variable, and equity income at 2% above. Very conservative."
@MJG: Did you perhaps overlook the "very conservative" qualifier?
I could care less that the "Actual historical data of these outcomes vary all over the map, are not constant, and not predictable." I deliberately projected a very pessimistic set of values which would emulate long periods of minimal returns, and threw in a meltdown scenario for good measure. Our plan survived.
You seem to be one of those folks who believe that whatever they do is the right and only way, that anything else is highly suspect or at best barely acceptable, and thoroughly enjoy telling us so.
QED.
Regards,
Ted
@MJG: I expect that the nuclear engineers responsible for 3-mile Island, and Chernobyl were also "experienced user[s] of Monte Carlo analyses".
Not bragging, but most people cannot do the same with Excel. That is why I think the Monte Carlo simulation is a great tool for most people, not as an exact answer but a ball park. That's all I was trying to say.
Take away MJG's annoyingness, (which we agree on, hmm, spell check says annoyingness is not a word) in my opinion people could and should be using something, and if you aren't a spreadsheet builder yourself I'm not sure there is anything better than Monte Carlo to get a ball park for probability to sustain.
Thanks for the feedback msf and old_joe.
Your most recent post clearly implies that somehow the sad facility failures that you identified were caused by shortfalls in their design as completed by nuclear engineers with a propensity for Monte Carlo analyses. That's a huge and unsupported extrapolation even for you. Evidence does not justify your claim. Systems erode over time.
If you don't recognize the benefits of Monte Carlo analyses (and there are instances when it is not appropriate) that's ok for me. I do and still use that tool. As I have consistently said: we each get to choose our own poison.
Mistakes happen and I am surely not immune to making my fair share of them. Monte Carlo tools have become a common tool these days in the investment community, They have the potential to reduce investment decision errors. My reference to Vanguard is just one of many serious advocates for deploying Monte Carlo as part of the investment decision making process.
Best Wishes
@MJG: "facility failures that you identified were caused by shortfalls in their design as completed by nuclear engineers with a propensity for Monte Carlo analyses."
No sir, not at all. Merely indicating that a masters in nuclear engineering is a guarantee of exactly nothing. Incidentally, I may be expecting too much here, but I would at least hope that a nuclear engineer would anticipate "erosion" and design against that possibility.
@Ted: Not sure what you mean. I've certainly never suggested that MJG is either a clown or an "id--- !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!" The only MFO poster that I've ever called a clown is that guy who thinks that the rules are for everyone but him, cheats so as to kick undeserving posts into the "Comments +" category, and then leaves footprints that a kid could detect. But you already knew that.
Regards,
OJ
Regards,
Ted
Obviously not significant to you, as you didn't introduce Bengen. Rather, you injected a claim that the 4% rule was based in part on Monte Carlo analyses. As you stated above, your purpose in submitting this false statement was to bridge from the WSJ article to what you wanted to write about.
Even the last reference you quoted acknowledged its usefulness.
Already addressed in my post prior to the one I'm responding to here. Useful, yes, it's better than a poke in the eye.
.
All tools have identical limitations
If all tools have identical limitations, then why push just a single tool? Previously you claim to have identified a shortfall in tools that use historical data. Shortfall, limitation. What's the difference? It seems tools don't all have identical limitations.
I am not an unsophisticated user of this fine modeling concept. ...
I am an experienced user of Monte Carlo analyses. In the nuclear engineering discipline (one of my masters degrees is in that field), it is a frequently used tool.
How does the fact that a tool is used in an engineering discipline built on physical rules make it an effective tool in a field (financial planning) that lacks a similar foundation? The observation that you and some engineers frequently use this tool (regardless of the domain) is not is a good argument that it is well suited for the unsophisticated, inexperienced user.
From my perspective it seems like an ideal tool.
Ideal, yet something with faults that could be eliminated: "I generated my own Monte Carlo code to aid in the decision process. I recognized the shortcoming of ..."
To summarize, starting with a false statement, you suggested that people use tools with model shortcomings that you found unacceptable for your own use. Tools that make sense to you, because you have a masters of nuclear engineering.
Many engineering disciplines are more like financial planning, and vice-versa, than you may realize. An awful lot of judgment and choice and sketchy causality entailed.
Picayune, picayune, and more picayune.
You clearly don't think much of the "way" I suggested that investors might find Monte Carlo analyses tools useful when making investment decisions. But what do you think of Monte Carlo for that purpose independent of the way I presented the suggestion? That's the important investment point of my submittal, not the way I write.
You are welcome to whatever you decide on any subject. I will never challenge that. I like Monte Carlo methods when examining highly uncertain outcomes. That's just my opinion based on both engineering and investing experiences. It is not shocking that others do not share the same opinions. That's partly what makes for a divergent and challenging investing marketplace.
Thank you for reading my posts with such attention, but not with such a picayune inclination. Sorry, I already said that.
Best Wishes.