Retirement Planning: Historical models vs Monte Carlo

[TempusFugit]

I’m about 18 months out from my target retirement date and I’ve been working on my detailed plans using various tools such as Boldin (formerly New Retirement), CFIREsim, FICalc, etc to evaluate my planned spending and portfolio longevities. 

I recently started experimenting with the Early Retirement Now SWR toolkit created by Karsten Jeske, a.k.a BigERN. This is historical back test model that incorporates the CAPE and S&P levels to see what a SWR would have been over the history of the market, much like the Trinity Study that produced the 4% rule. 

https://earlyretirementnow.com/2017/01/25/the-ultimate-guide-to-safe-withdrawal-rates-part-7-toolbox/

Jeske has written fairly exhaustively about his reasoning so I won’t try to regurgitate that here, but my more general question for the group is about the value of historical ‘backtesting’ as compared to forward projection as in a Monte Carlo analysis when evaluating our retirement plans. 

What are you using for your planning?   How do you think about Monte Carlo vs historical models?   Are you seeing meaningful differences depending on which models you use? 

[cwilkinson1ski]

I'm using Boldin too - gives me a lot of peace of mind that I won't outspend my $$$ in retirement.  I'm about 18-24 months out and can't wait.  I'm so done with work!

[scottish]

We have mandated withdrawal levels from our RSPs (I think this is like a 401-K in the US).    In our case, we can pretty well just live on the minimum withdrawals and the investment income from the non-registered portfolio is just a bonus.      So our total withdrawal rate is going to be around 1.5%.

I've setup cash reserves of about 1 year's expenses with the  idea that we'll just top that up from the RSPs when it drops down enough.   To answer your specific question, we're going to follow the "Living off your money" book with some adjustments to account for Canadian tax shelters and benefits.    So historical.

If you're going with a Monte Carlo model, have you looked at the details of the stochastic process used in the model?    I'd think this would be the "Achille's heel" of the approach...  stock markets aren't Markovian and they don't follow normal distributions.    I'm just curious, not critical.

[TempusFugit]

@scottish  Yes, I have read about some of the shortcomings of a Monte Carlo model, primarily by Jeske.   I’m not a mathematician or anything but I intuitively ‘get’ that the Monte Carlo model will have sequences that are not representative of how real world markets work.  I think perhaps the bigger issue with future-looking models is that you have to make assumptions about return ranges, and if you base those on historical returns then you’re kind of just duplicating a historical model with these weird MC sequences thrown in. 

Of course, a serious shortcoming of a historical model is the assumption that the future will look like the past, and I have doubts about that.  I’m sure everyone in the past would have said the same thing! 

@cwilkinson1ski  I’ve been happy with Boldin and it is my primary planning tool.  Of course, you have to feed it ‘good’ information in order to have any confidence in a reliable output!  Since the MC is a forward looking model, you have to make some guesses about future returns and future inflation.   

I’m using return values that average about 6.25% for my 70/30 portfolio and assuming a 3.25% inflation rate, which results in a REAL return of just over 3%.  That is of course much lower than the historical return for my AA (according to Vanguard the historical return for my AA is 9.1% - inflation so around 7% real).

What returns and inflation values are you using? 

Another nice thing about Boldin is that you can change your return rate in the future, e.g. set a low return for the first ten years and then a more ‘average’ return for the remainder of your retirement.  Vanguard released their predictions for market returns in the next ten years at a paltry 1.3%-ish for US large cap.  Obviously they can’t know the future any more than we can, but I am pretty pessimistic for the near term markets given the very high valuations.  That is where the SWR toolkit from Jeske is of value. 

[mistymoney]

my understanding of monte carlo models is that every possible combination of sequences is run. Is there any model like that that alos incorporates constraints on if the last 2 years were strongly negative, the probability of another negative year is reduced? and vice versa for positive returns?

looking at last 100 years of returns, one instance of 4 negative years (great depression), two instances of 3 negative years (2000, 1939), and one instance of 2 years (1973).

[reeshau]

There comes a point of diminishing returns with models.  The question isn't "Which is the absolute best model?" but rather "Which models do an adequate job?"

To that question, I say either Monte Carlo or historical models provide fairly similar accounts.  What works best for you is what gives you confidence to move forward.  Whichever argument / story / method lets you leap off the end of the dock and into the FIRE boat, is one that did the job.

If your situation is so marginal that some tweaks make a tremendous difference, then maybe take the message that your situation is fragile, or too sensitive to variance to be practical.

Once you take the leap, nothing speaks like experience.  With five years of FIRE under my belt, including one "unprecedented event," I am pretty confident in the future.  Not because I have seen it all, but because I didn't panic during the trials presented so far.  Not panicking, adjusting my actual approach as warranted, and seeing results better than modeled have all helped tremendously.

I still update my original 10 year plan spreadsheet as part of our annual review exercise.  But it's usefulness has greatly diminished, as we have diverged from it significantly. (In a positive way)

[TempusFugit]

...

There comes a point of diminishing returns with models. 
...

Absolutely!  I remind myself often that the natural variances in the things that we absolutely cannot predict with precision (life expectancy, market returns), but that have tremendous implications for any plan, will probably overwhelm any precision I might like to think I have.

...
If your situation is so marginal that some tweaks make a tremendous difference, then maybe take the message that your situation is fragile, or too sensitive to variance to be practical.

...

Yes and no.  An important aspect of planning for a 35-40 year retirement is that seemingly small differences add up and compound over that period to have big impacts.  We've all been taught that small increases in our savings rate or small reductions in our investment expenses over our investing lives can have outsized impacts on the size of our nest egg thirty years later.  The reverse is also true.  A small increase in spending (a WR that's 4.25% vs 4%, for example) can make all the difference between having money left over when we die at 90 or being destitute at 85.  I think some folks underestimate how powerful small differences can be if, for example they take the attitude that 4.25% is basically the same as 4%.  Especially if their SWR is really more like 3.75%. 

[reeshau]

...
If your situation is so marginal that some tweaks make a tremendous difference, then maybe take the message that your situation is fragile, or too sensitive to variance to be practical.

...

Yes and no.  An important aspect of planning for a 35-40 year retirement is that seemingly small differences add up and compound over that period to have big impacts.  We've all been taught that small increases in our savings rate or small reductions in our investment expenses over our investing lives can have outsized impacts on the size of our nest egg thirty years later.  The reverse is also true.  A small increase in spending (a WR that's 4.25% vs 4%, for example) can make all the difference between having money left over when we die at 90 or being destitute at 85.  I think some folks underestimate how powerful small differences can be if, for example they take the attitude that 4.25% is basically the same as 4%.  Especially if their SWR is really more like 3.75%.

What no (tolerable) model can tell you, for example, is that I'd there is a pandemic going on, with high inflation, then you don't travel that year.

Yes, modeling may be more important for "lean FIRE," where more of your spending is non discretionary.  But you are not pre-committed to *any* spending,  nor to foregoing work entirely.  As things come along, you can adapt.  That kind of flexible reaction is a hidden plan B that everyone has, and a general limitation of all models.

[TempusFugit]

...
If your situation is so marginal that some tweaks make a tremendous difference, then maybe take the message that your situation is fragile, or too sensitive to variance to be practical.

...

Yes and no.  An important aspect of planning for a 35-40 year retirement is that seemingly small differences add up and compound over that period to have big impacts.  We've all been taught that small increases in our savings rate or small reductions in our investment expenses over our investing lives can have outsized impacts on the size of our nest egg thirty years later.  The reverse is also true.  A small increase in spending (a WR that's 4.25% vs 4%, for example) can make all the difference between having money left over when we die at 90 or being destitute at 85.  I think some folks underestimate how powerful small differences can be if, for example they take the attitude that 4.25% is basically the same as 4%.  Especially if their SWR is really more like 3.75%.

What no (tolerable) model can tell you, for example, is that I'd there is a pandemic going on, with high inflation, then you don't travel that year.

Yes, modeling may be more important for "lean FIRE," where more of your spending is non discretionary.  But you are not pre-committed to *any* spending,  nor to foregoing work entirely.  As things come along, you can adapt.  That kind of flexible reaction is a hidden plan B that everyone has, and a general limitation of all models.

I agree but I do think many people overestimate the utility of ‘flexibility’ or perhaps more precisely they underestimate the magnitude of spending cuts and duration of said cuts that might be required to ‘right the ship’ under some circumstances. 

I think it is also perhaps too easy to just think you can go back to work.  The kind of market conditions that would cause a conservative plan to fail are also coincident with the kinds of economic environment that may not lend itself to just picking up a new job, especially if you’re in your sixties. 

All of your points are valid and should be kept in mind, but I want a plan that reasonably minimizes the chances that I will ever have to make spending cuts or to look for a new job. 

Part of that, at least in my mind, is to do all the things that I can to confirm that my plan will work, based on what can reasonably be known.  I’m sure my approaching retirement target date is tending to make me a little bit over cautious.  Plus it’s kind of fun to dig into the numbers and experiment. 

[Retire-Canada]

What are you using for your planning?   How do you think about Monte Carlo vs historical models?   Are you seeing meaningful differences depending on which models you use?

I used historical data models prior to retiring in 2020. Add in a healthy dose of flexibility which I had in my plan and I felt pretty confident. I ended up retiring into the teeth of the COVID Crash. While looking back it didn't amount to much, at the time it seemed pretty daunting. I'm almost 5 years into retirement with much more money now than when I started. That's the likely outcome with so many of the very conservative retirement plans folks around here are using.

My take is people [including myself] are erring on the side of being too cautious and not appreciating that the real scarce resource is time during the prime of our lives not money. Ultimately it's each person's choice how conservative they want to be with their plan, but never lose sight of the fact you are trading free time for more money and that has a cost of its own.

[Loren Ver]

We used historic.  Going back to work has never been our plan B (or C or D really). 
Historic was so much our plan that we actually left before we hit our number knowing that our number should catch up with us eventually.  We then jumped, pulled our chutes and waited to see what happened. 

Not that we aren't planners, but looking at how the market works, the plan had a really really good chance of working.  And it did, better than expected even.

 @Retire-Canada mentioned time.  For us time was the limiting resource.  Both DH and I had parents die on the young side and we didn't see the need to spend our unaccumulatable resource accumulating more of our easily accumulating and manageable resource. 

DH and I also have a fun sense of humor and timing, so we wanted fun dates to be our forever dates.  You know, priorities.  YMMV.

Loren

[wageslave23]

OP, I have similar concerns about future returns being less than historical.  Can you use something historical like firecalc.com and then increase your investment expense to 4% to account for the fact that you think future returns will only average 3% instead of 7% historically.  That way you get actual sequences but with adjusted returns.

[TempusFugit]

OP, I have similar concerns about future returns being less than historical.  Can you use something historical like firecalc.com and then increase your investment expense to 4% to account for the fact that you think future returns will only average 3% instead of 7% historically.  That way you get actual sequences but with adjusted returns.

Interesting idea.  I’ll play around with that to see what it does. 

[VanillaGorilla]

I regularly cringe at some of the analyses posted here and elsewhere using MC methods, specifically the tendency to use MC analysis to model outlier scenarios. This is precisely what MC models are NOT supposed to do, the entire point of MC methodology is to converge to an accurate mean, not model accurate outliers.

In a FIRE context, that means that if you run a MC simulation enough times, it will pierce together 30 consecutive negative annual returns and always give you a failsafe withdrawal rate of 0. Not exactly helpful. A MC analysis will be helpful if you're curious about, say, the mean ending value of a portfolio.

Additionally, the mean and variance of the underlying distribution are generally set to....historical data. So then you're conflating historical data with a crummy model.

Historical backtests are problematic from a human perspective - can we really expect humanity to continue innovating at the pace it has for the last 150 of 10 million years? Maybe not, but consider just how many facets of our society depend on the financial markets. If the US market goes kaput then not only will the fringe FIRE folks be forced back into employment, but the entire us population's retirement plans will have gone up in smoke simultaneously. Consider that a shock like that would impact the housing and mortgage market. It would destroy vast swaths of industry. And if we all end up in Pol Pot's agrarian utopia FIRE will not be at the forefront of anybody's mind.

I don't really like Monte Carlo simulations for finance. I like backtesting with an eye for overfitting (ie be deeply suspicious of outlier portfolios, short timeframes, cherrypicked allocations, and the like), coupled with a dose of optimism for humanity's future.

[mistymoney]

...
If your situation is so marginal that some tweaks make a tremendous difference, then maybe take the message that your situation is fragile, or too sensitive to variance to be practical.

...

Yes and no.  An important aspect of planning for a 35-40 year retirement is that seemingly small differences add up and compound over that period to have big impacts.  We've all been taught that small increases in our savings rate or small reductions in our investment expenses over our investing lives can have outsized impacts on the size of our nest egg thirty years later.  The reverse is also true.  A small increase in spending (a WR that's 4.25% vs 4%, for example) can make all the difference between having money left over when we die at 90 or being destitute at 85.  I think some folks underestimate how powerful small differences can be if, for example they take the attitude that 4.25% is basically the same as 4%.  Especially if their SWR is really more like 3.75%.

What no (tolerable) model can tell you, for example, is that I'd there is a pandemic going on, with high inflation, then you don't travel that year.

I thought that was what rich broke dead had with the % spending flexibility and % of stash, inflation adjusted, when you applied the flexibility.

[Ron Scott]

I would be wary of using models that assume historical data can be predictive of future results.

[Retire-Canada]

I would be wary of using models that assume historical data can be predictive of future results.

The models that I am aware of that use historical data like cFIREsim don't assume anything about future results. They tell you how your proposed portfolio would have done based on historical data. You have to decide how you incorporate that into your retirement planning.

Personally I feel pretty good about a plan that had a a high success rate against historical data. If a time machine is not available I'll take that and a healthy dose of flexibility in my plan and sleep well.

[reeshau]

It's all about specifics vs. general conclusions.  @VanillaGorilla said it well: a Monte Carlo analysis is a statistical method, and so works best in its averages.

When considering the applicability of the 4% rule and historical data, consider that the market crash of 1929, followed by the Great Depression, is the second worst historic scenario.  (Bengen called it "the little dipper.")  The stagflation of the 1970's is actually the scenario that sets 4% as SAFEMAX, in Bengen's terminology.

That doesn't match to how those events stick in our memories, but it's the math.  And I'm confident that the 2008 financial crisis, and the Covid pandemic, as bad as the were, recovered quickly enough that they won't come close to redefining it.

Outside of the scenarios, the safe withdrawal rate is as much as 6-8%.  You just don't know what scenario you are living, until the end.  So, this is already very conservative.  Following it is quite likely to leave you with a significant surplus..

[TempusFugit]

Re: the safety of the 4% rule of thumb, I find Jeske’s argument persuasive.  He makes the analogy of traffic jams, and picking a random time of day and using the probability that for most of every day, there are no traffic jams.  If, however, your journey is going to be during the rush hours, your odds of encountering a traffic jam are much higher than a full-day probability would suggest. 

In this metaphor, the rush hours are the times of very high valuations (CAPE ratios) and market highs. 

I suspect that most people retire during these periods, because market highs are where folks hit their numbers and they can look at something like the 4% rule and declare victory.  My parents, for example, retired in the late 90’s and then took a major hit in the dotcom bust.   And not enough time has yet passed for us to know whether retirees in that period will see another failure of the ‘rule’. 

Not to really open the debate again about the 4% rule I know there is an entire exhaustive thread on that topic.  Just explaining a little bit of my doubts about the applicability of that simple rule given that we are currently at extreme market valuations and because of the massive and growing government debt we may be in a new inflationary period that could last a long time.

[reeshau]

The problem with the rush hour analogy is that, even as a commuter, you aren't *always* driving at rush hour.  Some times you slow down, and some days flow just fine.  It's very hard to picture timeframes this long.

As far as the 90's--yeah, they are 30 years ago.  There are already some interesting charts of that time.  This one is from A Wealth of Common Sense.

https://awealthofcommonsense.com/2024/09/31-years-of-stock-market-returns/

It reads as the x axis (across the top) being the starting year, and the y axis (going down the left) is now many years since then.  The number is the compound market return.

If you retired before 1997, you're fine, even going through the dot bomb.  You are at the 10% long-term average.  Even retiring in 2000, at the beginning of "the lost decade" (see -1% return for the 10 year row) you are, at worst, at 7%; lagging the long-term average, but catching up.

If you retired in 2009 (not a good year!  market low! layoffs! bankrupcies!) you have a 14% compound annual return, and have to think a lot more about RMD's and estate planning than running out of cash.

To each their own--some people are meticulous planners, and some are pessimists.  But if you are maximizing joy or utility over your lifetime, given some amount of cash, then one-sided planning risks being way out of whack, either way.

[VanillaGorilla]

Re: the safety of the 4% rule of thumb, I find Jeske’s argument persuasive.  He makes the analogy of traffic jams, and picking a random time of day and using the probability that for most of every day, there are no traffic jams.  If, however, your journey is going to be during the rush hours, your odds of encountering a traffic jam are much higher than a full-day probability would suggest. 

In this metaphor, the rush hours are the times of very high valuations (CAPE ratios) and market highs. 

I suspect that most people retire during these periods, because market highs are where folks hit their numbers and they can look at something like the 4% rule and declare victory.  My parents, for example, retired in the late 90’s and then took a major hit in the dotcom bust.   And not enough time has yet passed for us to know whether retirees in that period will see another failure of the ‘rule’. 

All discussions here quickly devolve into arguments about the 4% rule. It's the Godwin's Law of the MMM forums. This forum is equally predictable when discussing market conditions: when the market is up people complain about valuations and wring their hands over the imminent crash. When the market is down people complain about the bear market and wring their hands that nothing will ever get better. :) It's just human nature.

CAPE is overwrought to me - I think it's one of the few flaws with ERN's arguments (the other is his analysis of mortgage rates and retirement). Industry and finance has changed far too much to compare modern CAPE to historical values. How are you going to compare high tech software/internet/AI companies to blue chip stocks of yesteryear? How can you argue that OpenAI or Google should be fairly valued with a similar CAPE to Coca Cola or GM? If you only bought stocks at or below the historical mean then you'd have about two years over my entire life to buy and you'd be quite poor. If you bought stocks regularly during that period you'd be quite rich.

If you're that wigged out about things work an extra year or two and bring your 4% down to 3.3%. Then you're good to go.

Even retiring in 2000, at the beginning of "the lost decade" (see -1% return for the 10 year row) you are, at worst, at 7%; lagging the long-term average, but catching up.

Y2K is the closest thing we have to a failure of the 4% rule in modern decades, it's hardly a success. That cohort will very likely make it to 30 years but depending on their asset allocation, their portfolio is modestly to severely depleted, and pushing for a 50 year withdrawal seems very dangerous.

Back to Monte Carlo simulations.

[Ron Scott]

I would be wary of using models that assume historical data can be predictive of future results.

The models that I am aware of that use historical data like cFIREsim don't assume anything about future results. They tell you how your proposed portfolio would have done based on historical data. You have to decide how you incorporate that into your retirement planning.

Personally I feel pretty good about a plan that had a a high success rate against historical data. If a time machine is not available I'll take that and a healthy dose of flexibility in my plan and sleep well.

Using big 20th Century catastrophes to stress test a 4% approach is pretty good. The depression, world war, and stagflations were nasty. Having a portfolio hold up for the Trinity study’s 30 year period when faced with all that is helpful for planning.

We have no idea how the future will play out and what big events lie ahead. Some however might have a worse affect on markets and returns. Here are some examples that could make the 4% rule fail, especially for retirement periods of 40 years or more. You may or may not find them worthwhile to factor into your SWR strategy.

1. A truly “worldwide” war, involving all the players in the previous wars but with China/Russia/NK/Iran teamed up, AI military robots on the loose, prolonged decimation of supply chains, in which upsetting the American-led world economic order is the endgame and creating an economic mess that takes a few decades to sort out is considered acceptable.

2. An environmental catastrophe that plays out over a decade in which there is massive migration of the world’s population and the means of production for 5% of world GDP is lost and needs to be rebuilt.

3. Several BIGNEWTECH companies that are privately owned upset the market positions of companies that are publically traded, causing a significant shift in ownership of capital.

4. Two of the above occur while you are in retirement.

5. At least one of the events occur during your retirement. It is managed very well by governments, but still results in a major SORR downside. You need to withdraw a major portion of your equity portfolio during the first 10 years of retirement.

These may seem completely outlandish to you with a 0% probability of happening, or you may have faith in government officials to work things out for you if they do. And we obviously can’t prepare for nuclear winters and asteroid extinctions. But all I’m really saying here is the next 40+ years could be worse for retirees than the past 120. if any of these scenarios seem somewhat plausible and would have the potential nip a third of your portfolio, it might be worthwhile to save a little more than 25X to see you through 45 to 90.

The value in planning for retirement lies more in what you don’t know about the future than what you do know about the past. IMO

[TempusFugit]

Well, obviously we all have faith in our government leaders to guide us through any combination of these things.  Right?

[Villanelle]

...
If your situation is so marginal that some tweaks make a tremendous difference, then maybe take the message that your situation is fragile, or too sensitive to variance to be practical.

...

Yes and no.  An important aspect of planning for a 35-40 year retirement is that seemingly small differences add up and compound over that period to have big impacts.  We've all been taught that small increases in our savings rate or small reductions in our investment expenses over our investing lives can have outsized impacts on the size of our nest egg thirty years later.  The reverse is also true.  A small increase in spending (a WR that's 4.25% vs 4%, for example) can make all the difference between having money left over when we die at 90 or being destitute at 85.  I think some folks underestimate how powerful small differences can be if, for example they take the attitude that 4.25% is basically the same as 4%.  Especially if their SWR is really more like 3.75%.

What no (tolerable) model can tell you, for example, is that I'd there is a pandemic going on, with high inflation, then you don't travel that year.

Yes, modeling may be more important for "lean FIRE," where more of your spending is non discretionary.  But you are not pre-committed to *any* spending,  nor to foregoing work entirely.  As things come along, you can adapt.  That kind of flexible reaction is a hidden plan B that everyone has, and a general limitation of all models.

I agree but I do think many people overestimate the utility of ‘flexibility’ or perhaps more precisely they underestimate the magnitude of spending cuts and duration of said cuts that might be required to ‘right the ship’ under some circumstances. 

I think it is also perhaps too easy to just think you can go back to work.  The kind of market conditions that would cause a conservative plan to fail are also coincident with the kinds of economic environment that may not lend itself to just picking up a new job, especially if you’re in your sixties. 

All of your points are valid and should be kept in mind, but I want a plan that reasonably minimizes the chances that I will ever have to make spending cuts or to look for a new job. 

Part of that, at least in my mind, is to do all the things that I can to confirm that my plan will work, based on what can reasonably be known.  I’m sure my approaching retirement target date is tending to make me a little bit over cautious.  Plus it’s kind of fun to dig into the numbers and experiment.

Let's say the market is down 22% and inflation is high.  Even if the models say you are good, are you really not going to think, "hey, maybe we should aim to spend $20 less per week on groceries by relying more heavily on sales and adding a meatless meal" and "sure, the elderly neighbor asked me to make a grocery run for them for $15 a week and I can just do theirs while I shop for mine, but nah, I don't want the extra 2o minutes added to my week's obligations."  IOW, when the SHTF, most people are going to make cuts (or seek income, or both) when the market is rough, whether they've planned for it or not.  Because looking at models is one thing and looking at the bank account balance is another.  Maybe you really are a stalwart, intrepid type who will put all your faith in the models and power ahead at full steam in the face of a recession.  But most people aren't, so trying to plan to avoid something you'd probably do anyway may end up costing you several more years you could have been FIREd.  So it makes sense to do some soul-searching to see if you really think you have the stomach to not make cuts or bring in income in the seanrios you are seeking to avoid. 

[Ron Scott]

I would be wary of using models that assume historical data can be predictive of future results.

The models that I am aware of that use historical data like cFIREsim don't assume anything about future results. They tell you how your proposed portfolio would have done based on historical data. You have to decide how you incorporate that into your retirement planning.

Personally I feel pretty good about a plan that had a a high success rate against historical data. If a time machine is not available I'll take that and a healthy dose of flexibility in my plan and sleep well.

Using big 20th Century catastrophes to stress test a 4% approach is pretty good. The depression, world war, and stagflations were nasty. Having a portfolio hold up for the Trinity study’s 30 year period when faced with all that is helpful for planning.

We have no idea how the future will play out and what big events lie ahead. Some however might have a worse affect on markets and returns. Here are some examples that could make the 4% rule fail, especially for retirement periods of 40 years or more. You may or may not find them worthwhile to factor into your SWR strategy.

1. A truly “worldwide” war, involving all the players in the previous wars but with China/Russia/NK/Iran teamed up, AI military robots on the loose, prolonged decimation of supply chains, in which upsetting the American-led world economic order is the endgame and creating an economic mess that takes a few decades to sort out is considered acceptable.

2. An environmental catastrophe that plays out over a decade in which there is massive migration of the world’s population and the means of production for 5% of world GDP is lost and needs to be rebuilt.

3. Several BIGNEWTECH companies that are privately owned upset the market positions of companies that are publically traded, causing a significant shift in ownership of capital.

4. Two of the above occur while you are in retirement.

5. At least one of the events occur during your retirement. It is managed very well by governments, but still results in a major SORR downside. You need to withdraw a major portion of your equity portfolio during the first 10 years of retirement.

These may seem completely outlandish to you with a 0% probability of happening, or you may have faith in government officials to work things out for you if they do. And we obviously can’t prepare for nuclear winters and asteroid extinctions. But all I’m really saying here is the next 40+ years could be worse for retirees than the past 120. if any of these scenarios seem somewhat plausible and would have the potential nip a third of your portfolio, it might be worthwhile to save a little more than 25X to see you through 45 to 90.

The value in planning for retirement lies more in what you don’t know about the future than what you do know about the past. IMO

Well, obviously we all have faith in our government leaders to guide us through any combination of these things.  Right?

LOL, which is part of the point here.

You don’t need to trash the past to be wary of models that use it to assess a SWR. You just need to ask “what if the future isn’t as rosy to investors”? That seems like a perfectly reasonable question to me.

[TempusFugit]

Regarding the 4% rule, I think the biggest issue is simply that it isn't the way people really spend in retirement. Not accounting for the reality of spending changes over a multi-decade period is something that might trip up lots of people, in one direction or another.

For example, I plan to begin my retirement with something closer to 5%, even a smidge over.  That's because I have other income that will start a few years later, and then SS will enter the picture (I am assuming an 18% haircut, per current law).   My mortgage will also be paid off ten years in. 

I also model my retirement spending to drop at some point as I get into my middle 70's (God willing). 

I try to model the lumpy expenses like buying new cars every 10 years or replacing the roof on the house, etc. Those years can result in higher withdrawal rates, sometimes higher than 6%.  In fact I started another thread to ask how people planned to handle withdrawals for that type of expense to minimize tax hits. 

It's pretty straightforward to enter these things into the Boldin tool, which then uses Monte Carlo models to evaluate the plan.  I think it would be great if it also back tested against historical conditions, but alas, it does not. 

I've been experimenting with FICalc to do that part.  And that's really what prompted this thread.   I was curious what other people were doing and if there were some tools available for historical testing that I'm not ware of. 

So, for my plans I am doing these things:

1) Boldin is my primary tool which uses MC analysis based on the return ranges that I input, and it imputes a standard deviation based on the returns (higher deviation for higher return values).  I used portfolio visualizer to find historic returns for my assets, along with Vanguard's data for a 70/30 AA. I reduced these because I'm a pessimist.

2) FICalc to see historical analysis, and I input additional income for SS, pension, and additional withdrawals for cars, major repairs, etc. 

3) Jeske's SWR toolkit to get a historical evaluation that considers todays CAPE and market level (all time highs).  BTW, for my parameters (my age, my future income, life expectancy, etc) the (constant) SWR is actually 5.12%.   Now, his tool is actually evaluating your spending level, not your *withdrawal* level, meaning that the 5.12% is saying a constant spending level of 5.12%, irrespective of where that money comes from, SS, pension, or actual withdrawals from the portfolio, is (was) sustainable.  The WR from the portfolio would actually drop over time as other sources of income come online.  His tool also provides for a dynamic spending model that starts even higher but of course is dependent on your willingness to cut spending, perhaps significantly below 4%, if the market changes dramatically.

4) I did also pay a CFP to review my plan.  I'm not quite happy with her work product, even though she did give me a thumbs up, because some of her advice seems sketchy to me (all bonds in my ROTH?), so I kind of discount her opinion on the larger picture. 

[Telecaster]

What are you using for your planning?   How do you think about Monte Carlo vs historical models?   Are you seeing meaningful differences depending on which models you use?

Historical.  I don't think Monte Carlo is valid for SWR calculations.   But I also try to be mindful of what the historical data are saying.   We all know that the future will be different than the past, and then promptly forget it.   For example, Jaske calculates the SWR out to two decimal places (and in some cases even three!).  I do not believe for a solitary second that past results have predictive power out to two decimal places.   There is no way.   Markets are different, the economy is different, the world is different, the way P/E is calculated is different, the way businesses return money to shareholders is different.   I flat reject the level of precision he's implying.   For his part, at times he seems to acknowledge this but I think huge amounts of his work is not actionable because of the false implied precision.   

But all the models seem to converge around 4%.  And although we all know that's a guess to a certain extent, that really does seem to be the magic number we can start with for planning purposes.   One thing the data do tell us that I feel confident about is that small input changes in the early years can greatly improve the final portfolio balances.   So if things start going south in the early years, some combination of cutting spend and part time work should get you through.    If things don't go south in the early years, a different danger arises:  The danger of falling off of your wallet.   

So for me, 4% fine as I long as I keep an eye on things.  And basically everyone plans to keep an eye on things regardless of WR they use, so I don't see much utility in trying to optimize beyond that.   

[TempusFugit]

What are you using for your planning?   How do you think about Monte Carlo vs historical models?   Are you seeing meaningful differences depending on which models you use?

...  I do not believe for a solitary second that past results have predictive power out to two decimal places.   There is no way.   Markets are different, the economy is different, the world is different, the way P/E is calculated is different, the way businesses return money to shareholders is different.   I flat reject the level of precision he's implying.  ...

In fairness, I don't think he is saying that it is a perfect prediction.  His numbers can be precise to 2 decimal places because it's an analysis of something that has already happened, so it can be known.  Granted, it is for the purpose of informing our plans for the future, and in that regard I totally agree that it would be false precision. 

Just like the 4% rule itself, it does not claim to predict the future. People like to interpret it that way, which is fine inasmuch as we have to use something to help make plans.  I cringe every time I hear some financial guru in a podcast say, "the four percent rule shows that you will never run out of money..."   NO! It shows you the past, not the future.  And even then, people gloss over that there were failures in the 60s retirement cohorts.  The 4% WR failed in the past. It could fail more often in the future.  We don't know. 

What are you using for your planning?   How do you think about Monte Carlo vs historical models?   Are you seeing meaningful differences depending on which models you use?

...models seem to converge around 4%...

Models that have the same constraints as those used to generate the rule, yes, but most people don't have the same real world inputs.  Almost no one retiring today has a 50/50 portfolio with exactly 30 years retirement horizon, no social security, and spends exactly 4% inflation adjusted every year of your retirement.  If we start pushing those numbers around it can make meaningful differences in the SWR using historical market conditions.  Some for the better, some not. 

Just to say that the 4% is not to blame if people misuse it. 

[Telecaster]

Just to say that the 4% is not to blame if people misuse it.

Totally agree, and that was my real point.  Big ERN wrote a post about planning for assisted living.   I read it a while back and don't fully remember it, but the jist of it was that it was so far in the future he wasn't going to worry about it.   Which makes total sense.   At his age, it is too far in the future to plan for.   But for someone in their 60's probably need to start thinking about it.    But the 4% rule doesn't know about those kinds of things.   So we have to personalize it and make decisions as best we can based on the knowledge we have. 

[TempusFugit]

Just to say that the 4% is not to blame if people misuse it.

Totally agree, and that was my real point.  Big ERN wrote a post about planning for assisted living.   I read it a while back and don't fully remember it, but the jist of it was that it was so far in the future he wasn't going to worry about it.   Which makes total sense.   At his age, it is too far in the future to plan for.   But for someone in their 60's probably need to start thinking about it.    But the 4% rule doesn't know about those kinds of things.   So we have to personalize it and make decisions as best we can based on the knowledge we have.

I haven't read his post on that topic.  It isn't in my nature to just forget about it though!

I have modeled in my plan selling my home when I'm in my late 70's and I've increased my monthly expenses for a few years to account for independent living type arrangements, then a few more years of even higher cost assisted living, all based on today's rates in my city for those places.  Boldin also adds long-term / nursing home expenses of 100K/yr for the last two years.

I realize that it is far in the future but that doesn't mean there is no value in considering what it might look like.  The plan may be imperfect, but the act of planning is valuable! 

In reality, the thing that is most likely to have a big impact on my plan is that I probably won't live as long as I'm planning.  That makes the plan finances work out a lot better.. but I'm hesitant to call that a great scenario  : )   

[Telecaster]

There's a great quote from Eisenhower that I like  "“Plans are useless, but planning is indispensable."

[TempusFugit]

That’s what I was thinking of, just couldn’t recall who said it. 

The other quote that comes to mind is:  “Everyone has a plan until they get punched in the mouth” - Mike Tyson

[cwilkinson1ski]

@TempusFugit @cwilkinson1ski 
So true about feeding Boldin "good" information - I've way over budgeted on some items in my like to spend category to be safe and my need to spend budget is actually more current actual spend - so could be cut further if needed.

I’m also using return values that average about 6 - 6.25% - I was using a low inflation rate of between 1 and 2 % but after I read your post I bumped it up to 3.25% to stress test my plan.  Even with the higher inflation rate I still had an 89% success rate on my pessimistic scenario and 99% on average and optimistic.   

Good to know about being able to change my rate of return for future years.  I'll try that. 

Have you worked with any of the Boldin financial advisors?  I was considering making an appointment but wondered if it's worth the cost...

[reeshau]

Both good quotes on planning.  Another appropriate one:

"All models are wrong, but some are useful."
- George Box

[TempusFugit]

To project my desired retirement budget, I've taken my current annual spend (which I've tracked in detail for several years) and added on health insurance at $8K/yr and also added about 20% for additional travel and hobby spending for the 'go-go' years of retirement (the first 18 years in my plan). 

I think it's important to know how much of one's budget is discretionary, to be leveraged in the case of that needed 'flexibility' everyone talks about. In my case I identified discretionary items- dining out, club/hobby costs, subscriptions, 'miscellaneous' spend like Amazon shopping, and I cut them in half. That half I consider discretionary, meaning in really bad times I could fairly easily cut these things back by half and not really feel like my lifestyle was downgraded materially.  I'd eat out 6-7 times a month instead of 12-15, I'd buy the $20 bottle of wine instead of the $40. Stuff like that. Hard core.  : )

I saw recently a podcast with Mad FIentist discussing a new analysis he took part in that considered historic SWR if you followed a rule about discretionary spending: 

- Markets 10% down, only 50% of planned discretionary
- Markets 20% down, no discretionary

In that case, the back tested SWR (for a 30% discretionary budget, as mine is under the definition above) was 4.75% (other factors being unchanged from the traditional 4% rule analysis)

[TempusFugit]

@TempusFugit @cwilkinson1ski 

Have you worked with any of the Boldin financial advisors?  I was considering making an appointment but wondered if it's worth the cost...

I have not.  I think they have a free 'office hours' Q&A kind of thing every so often.   They also have a youtube channel if you haven't seen that.  One of those videos is where they explain how they seed the MC model standard deviation based on your rate of return. 

[TempusFugit]

@TempusFugit @cwilkinson1ski 

Good to know about being able to change my rate of return for future years.  I'll try that. 

This would allow you to assume the paltry ~1.5% REAL returns from the US market for the next decade (as Vanguard projects) just to see what the implications are on your plan.  At least maybe showing you a sense of the magnitude of the impact.

[scottish]

@scottish  Yes, I have read about some of the shortcomings of a Monte Carlo model, primarily by Jeske.   I’m not a mathematician or anything but I intuitively ‘get’ that the Monte Carlo model will have sequences that are not representative of how real world markets work.  I think perhaps the bigger issue with future-looking models is that you have to make assumptions about return ranges, and if you base those on historical returns then you’re kind of just duplicating a historical model with these weird MC sequences thrown in. 

My reasoning is that historical data provides more information than random data.   Think Bayesian!     And history does have an odd habit of repeating...

[Ron Scott]

@TempusFugit @cwilkinson1ski 

Good to know about being able to change my rate of return for future years.  I'll try that. 

This would allow you to assume the paltry ~1.5% REAL returns from the US market for the next decade (as Vanguard projects) just to see what the implications are on your plan.  At least maybe showing you a sense of the magnitude of the impact.

The assumption of achieving a real return over a retirement period is sound. But keep in mind that just breaking even with inflation isn’t a horror show. Better to be frugal in your assumptions and surprised in your retirement than bet the farm and suffer.

[TempusFugit]

The assumption of achieving a real return over a retirement period is sound. But keep in mind that just breaking even with inflation isn’t a horror show. Better to be frugal in your assumptions and surprised in your retirement than bet the farm and suffer.

Perhaps, but I don’t think that’s a particularly useful plan.  If I start retirement based on a spending plan that assumes zero real return over decades, I’m (almost certainly) denying myself the fruits of my labors.  By the time I decided that enough years had passed to prove the hypothesis wrong, I may have lost the opportunities to enjoy that money during the best years of my healthy retirement. 

That’s not to say it isn’t worthwhile to compare one’s income sources with one’s basic needs to see what life might look like in that scenario, but I wouldn’t want to use a spending plan that assumed it from the start. That’s too pessimistic for me.  On the flip side, projecting returns that match historical is too optimistic in my opinion.  That is actually one of my main concerns with using historical models for planning purposes. 

As someone suggested above, I’ve been experimenting with FICalc and adding in fees to artificially reduce historical returns.  I haven’t finished playing with that model yet, but I like the idea just as another data point.