Proposed rule of thumb for when your FIRE is in trouble

[secondcor521]

Hi all,

I wanted to start a new thread for this.  In another thread people are discussing FIRE failures and asking if there is a rule of thumb to determine if one is in trouble.

I have been wondering about this as well and had an idea that ended up producing a simple rule of thumb for my specific circumstances that might be able to be generalized.  I'll describe how I came up with the rule of thumb and then give the rule of thumb that I came up with.

1.  I started with the FIREcalc graph that plots the portfolio balances over time - year of retirement (1 through n) on the x axis, and portfolio value on the y axis, with one line for each starting retirement year.  (I like cfiresim better, but I couldn't find a way for it to give me all of the raw data I wanted, so I used FIREcalc instead.)

2.  For the particular set of numbers I entered, there were 90 data series, of which 36 ended up failing (falling below 0).

3.  I then imagined the graph from step 1 with only these 36 failing sequences on it.

4.  I then imagined a new data series which was equal to the highest portfolio value of all of the failing series for each year of retirement (1 through n).  In other words, for the particular FIREcalc run that I did, it was possible for an unlucky retiree with this much money to still ultimately fail.  Call this the "unlucky line".

5.  For each year of retirement (1 through n), I then counted how many of the 54 successful data series were below the "unlucky line".  These are situations where the retiree looks like they could be in trouble but their lucky future returns ended up rescuing them.  Call this the "lucky count".

6.  As you might expect, in the first few years of FIRE, the "lucky count" is pretty high because all of the series are jumbled together, but towards the end it is quite low because the series start to separate themselves.  What was surprising to me was that for pretty much the last half of this FIREcalc data, the "lucky count" was basically zero.

What this means is that -- for the particular dataset I looked at -- for a 40 year FIRE period, if at the 20 year point I had less than my starting amount then I had over a 90% historical probability of failure AND if at the 20 year point I had more than my starting amount then I had a 100% historical probability of success.

I like the symmetry that the picture becomes clear halfway through the FIRE period, and also that the unlucky line pretty much happens to pass through the original portfolio amount at that halfway point.

The rule is not particularly sensitive to the year or the amount.  So for example, at 16 years, the percentages are 90% and about 95% respectively.  At 23 years, if you're less than 75% of your initial amount, the percentages are 100% and 100% respectively.

It does look like it is still quite possible to be lucky up through 15 years into the FIRE period.  In that case there were 17 sequences that were below the "unlucky line" that eventually recovered.  (I.e., the "lucky count" was 17.)

All of the above analysis is on my particular numbers for my particular FIREcalc inputs (which were for a 40 year retirement, 85% equities, .05% expense ratio, and a few years of higher than 4% spending followed by ~37 years of ~4% spending).  I am not sure to what degree it can be generalized, but I thought I would throw it out there for discussion.

2Cor521

People like Pfau have attempted to mathematically analyze it, so reading his stuff would be a good place to start, if you want to get beyond a gut-level "hmm, maybe I should pare back spending" feeling.

I'm not aware of any guidelines that have come out of the research to help early retirees gauge whether or not their portfolio performance is on a successful trajectory.  People like Pfau are generally focused on determining the withdrawal rate that can be expected to be sustainable in anything but the worst-case and near-worst-case scenarios, or developing more dynamic withdrawal strategies that can protect against downside risk and/or allow retirees to take advantage of upside outcomes.  Has anyone put forward a history-based rule of thumb (akin to the 4% rule) that tries to quantify the "gut feeling"--something like:  if you're using a 4% WR and your portfolio's average performance over the first W years is less than X% then you need to drop to a Y% WR until the portfolio recovers by Z%?

[Monkey Uncle]

I'll preface this by saying that it is always much easier to throw rocks at a product that someone else creates than it is to create the product yourself.  So, I applaud your efforts to think this through and come up with a guideline to address the thorny problem of how to know when your FIRE is headed toward failure.

But here's the rock:  20 years in, it is probably too late to fix the situation without a drastic, sustained drop in spending, and/or going back to work.  What you need is some way to separate the lucky series from the unlucky series in the first 10 years or so.  I know, easy for me to say.  But that's why this is such a thorny problem.

[brooklynguy]

I like dragoncar's suggestion in the other thread of simply rerunning cfiresim periodically with your updated numbers to see if you're still comfortable with the then-current historical odds.  This will give you updated odds based not only on the actual performance of your portfolio but also on the addition of the newer periods that get included in the historical dataset.

[Retire-Canada]

Here is a simple plan to avoid FIRE failure:

- keep the door open to part time work
- give yourself some spending flexibility ~4% WR = 3% - 6% range
- don't pull more out of your investments in a given year than they've grown [adjust the difference by part time work and decreased spending and emergency fund]
- in good years pull out some extra and establish a liquid reserve fund not tied to the markets

This will buffer a lot of the possible market fluctuations.

Even if you have to break the rule about not pulling $$ out greater than your investments have grown if that's only a worst case scenario you've kept your principal secure to allow for this.

As you get older and there is less time left to fund plus you have gov't benefits coming in you can be less concerned.

-- Vik

[stuckinmn]

Building on Vikb's comment, what i am planning, at least the first few years, is to take out the lesser of 1) 4% of starting portfolio, inflation adjusted, or 2) 4% of portfolio value at that time.  Seems like a nice little self correcting rule to tell you when it is time to cut expenses or get a part time job.

I've run this through cfiresim and it has a 100% success rate (obviously, as any plan to withdraw a set percentage of current value per year will always have something left) and the lowest the annual spending gets down to is about 50% of the original for a bit. 

And these dips in what you can take out always happen in the first 3rd of retirement- by the time you get to the middle part you are typically just taking 4% of the original.  I guess that's yet another way of saying that if you are going to get screwed, it is going to happen fairly early, so you can quickly correct it.   

[DoubleDown]

Some other things to consider for warm and fuzzy feelings against portfolio failure:

- Get an annuity to cover most basic living expenses
- Paid off SFH will always provide a place to live regardless of market fluctuations or inflationary pressures

[retired?]

Some other things to consider for warm and fuzzy feelings against portfolio failure:

- Get an annuity to cover most basic living expenses
- Paid off SFH will always provide a place to live regardless of market fluctuations or inflationary pressures

It may give a warm and fuzzy feeling, but inflation will eat at the annuity.  Also, with low rates, your annuity will be minimal.  Watch out for fees, too.

It will be taxes or inflation (increasing the money supply) that will cover all the deficits over the past decade and spending to come.  Taxes are harder to pass.....inflation is sneaky.  Can anyone stop the Fed from buying?

[Eric]

Some other things to consider for warm and fuzzy feelings against portfolio failure:

- Get an annuity to cover most basic living expenses
- Paid off SFH will always provide a place to live regardless of market fluctuations or inflationary pressures

It may give a warm and fuzzy feeling, but inflation will eat at the annuity.  Also, with low rates, your annuity will be minimal.  Watch out for fees, too.

I don't really get this.  I'm pretty sure you can set up an annuity to be both inflation adjusted and non-minimal.

[stuckinmn]

Some other things to consider for warm and fuzzy feelings against portfolio failure:

- Get an annuity to cover most basic living expenses
- Paid off SFH will always provide a place to live regardless of market fluctuations or inflationary pressures

It may give a warm and fuzzy feeling, but inflation will eat at the annuity.  Also, with low rates, your annuity will be minimal.  Watch out for fees, too.

I don't really get this.  I'm pretty sure you can set up an annuity to be both inflation adjusted and non-minimal.

An inflation adjusted non-minimal annuity will cost you a lot of principal.  I have no idea of the exact amount but it's probably high enough you can do the same at some really low swr of 1.5 or 2%.

[MrMoogle]

Building on Vikb's comment, what i am planning, at least the first few years, is to take out the lesser of 1) 4% of starting portfolio, inflation adjusted, or 2) 4% of portfolio value at that time.  Seems like a nice little self correcting rule to tell you when it is time to cut expenses or get a part time job.

I was thinking about doing something like this.  But change 1) to 3.3%, and keep 2) as 4%.  This way, I wouldn't be affected if the first year has a ~15% drop.  And I think 3.3% has a 100% success rate, so this is only helping if the next 70 years are worse than all previous 70.

[stuckinmn]

If you can get to a 3.3 swr, you probably don't need the contingency plan in 2), considering the 100% success rate in cfire.  Of course, this means 30x expenses rather than 25x which calls for a couple more years of work. 

[Turkey Leg]

We are shooting for 80/20 asset allocation. With 80% stocks, I plan to refer to Living AFI's drawdown series, especially Part 4, to keep emotion out of 'stache decisions. Stay the course, and cut back spending where needed, if portfolio failure looms.

[TheFrugalFox]

Maybe a bit naive of me, but have never done used FIREcalc or any other system/ formula. I am not touching my capital and only using dividends - and that I do one year in arrears. Basically, every August withdraw all my dividends, divided by 12 and that is my monthly budget for the next year - so no surprises and I like that I am not touching the capital. Kind of simple, but I like it.

[Eric]

Maybe a bit naive of me, but have never done used FIREcalc or any other system/ formula. I am not touching my capital and only using dividends - and that I do one year in arrears. Basically, every August withdraw all my dividends, divided by 12 and that is my monthly budget for the next year - so no surprises and I like that I am not touching the capital. Kind of simple, but I like it.

That's a perfectly valid way to do it.  Just make sure that you consider that dividend yields change all the time, so your budget would need to be more flexible.  But I'd guess that the main reason most of us reject that path is that you'd have to work many extra years to build up enough principle to produce dividend payments that covered all of your spending.  It hardly seems worth it, considering that dividends are mathematically equivalent to selling shares.  But if you're already there, then you're in a good spot.

[Retire-Canada]

I like dragoncar's suggestion in the other thread of simply rerunning cfiresim periodically with your updated numbers to see if you're still comfortable with the then-current historical odds.  This will give you updated odds based not only on the actual performance of your portfolio but also on the addition of the newer periods that get included in the historical dataset.

Most definitely. When I am FIREd keeping my 'stach healthy will be my job.

Reviewing invest performance.

Reviewing new investment options/theories.

Managing my portfolio.

Same with my spending and cash flow management.

I'll spend a bunch of time optimizing my taxes so I benefit as much from my 'stash as possible.

Of course since I am FIREd I may well be doing  all of this from my RV camped on a beach in Baja.

-- Vik

[secondcor521]

I like dragoncar's suggestion in the other thread of simply rerunning cfiresim periodically with your updated numbers to see if you're still comfortable with the then-current historical odds.  This will give you updated odds based not only on the actual performance of your portfolio but also on the addition of the newer periods that get included in the historical dataset.

I'm kinda wondering why I even went through all the analysis, because dragoncar's idea is (a) simple, (b) easy to understand, (c) takes all of the information into account, and (d) is what i was already planning to do anyway.

@Monkey Uncle, you are right that after 20 years, you might already be in trouble.  In the particular data set I ran, out of the 36 data sequences that go on to fail at 40 years, 15 had already failed by the 20 year point.

For a 5 year rule of thumb, if you have 125% of your original stash in nominal terms after five years, then you had a 100% chance of success.  If you had less than 125% of your original stash in nominal terms after five years, then you had about a 53% chance of failure.  So that may be a useful rule of thumb also.  The "risk" in that case is going back to work even though you historically may not have needed to.  The surprising thing about that result is that after being retired for five years and seeing my stash 25% higher, I don't think that would feel very uncomfortable even though that's what the data says.  Strange.

Practically speaking, though, my personal FIRE plan calls for me to evaluate my situation using the dragoncar approach twice a year, and invoke my contingencies if I fall below my original safety rate (which is currently 95% success in cfiresim with my particular inputs).

[brooklynguy]

For a 5 year rule of thumb, if you have 125% of your original stash in nominal terms after five years, then you had a 100% chance of success.  If you had less than 125% of your original stash in nominal terms after five years, then you had about a 53% chance of failure.  So that may be a useful rule of thumb also.  The "risk" in that case is going back to work even though you historically may not have needed to.  The surprising thing about that result is that after being retired for five years and seeing my stash 25% higher, I don't think that would feel very uncomfortable even though that's what the data says.  Strange.

In the bolded text, is what you are saying that out of the universe of cases that were below 125% of the original stash in nominal terms at the five year point, 53% of them turned out to be failures?  If so, that in itself is not a very useful metric (because it includes all cases where the stash size was below 125% at the five year mark, whether they were 124% or 50%), and it doesn't tell you that you should be worried merely because your stash is under 125% five years out (take another look at the data you're using; maybe 99% of cases where the stash size was 120% or higher five years out were successes, and so on).

FYI, there's a parallel discussion going on in the other thread that inspired this one.

[secondcor521]

For a 5 year rule of thumb, if you have 125% of your original stash in nominal terms after five years, then you had a 100% chance of success.  If you had less than 125% of your original stash in nominal terms after five years, then you had about a 53% chance of failure.  So that may be a useful rule of thumb also.  The "risk" in that case is going back to work even though you historically may not have needed to.  The surprising thing about that result is that after being retired for five years and seeing my stash 25% higher, I don't think that would feel very uncomfortable even though that's what the data says.  Strange.

In the bolded text, is what you are saying that out of the universe of cases that were below 125% of the original stash in nominal terms at the five year point, 53% of them turned out to be failures?  If so, that in itself is not a very useful metric (because it includes all cases where the stash size was below 125% at the five year mark, whether they were 124% or 50%), and it doesn't tell you that you should be worried merely because your stash is under 125% five years out (take another look at the data you're using; maybe 99% of cases where the stash size was 120% or higher five years out were successes, and so on).

FYI, there's a parallel discussion going on in the other thread that inspired this one.

Yes, that is what I am saying.  There were 90 data series.  Of these 90 data series, all of the 36 data series that failed had a 5 year nominal balance of 125% or less.  There were also 31 data series that had a 5 year nominal balance of 125% or less but eventually turned out to be successful.  There were also 23 data series that had a 5 year nominal balance of 125% or more that also were all successful.

If you don't find it useful, OK.  I was just looking at the idea and data and responding to the earlier poster who pointed out that they wanted a 5 year rule of thumb rather than a 20 year rule of thumb.

I agree with you that it doesn't address the magnitude of the stash (124% vs. 50%).  However, if one starts down that train of thought, I think one gets into a, "well, if you have between 75% and 125%, then your chance of failure is X%, and if you have between 25% and 75%, your chance of failure is Y%", which starts moving more away from a yes/no rule of thumb to a set of guidelines.  Which is OK also.  It seems to me that some people are more willing to take a "I'll know it when I see it" or gut feel approach, and some want more of a data-driven "I'll go back to work as soon as X happens".

Oh, I also agree with you (I think this is where you're going) that people should look at the individual data runs and also do sensitivity analyses on their inputs to get a feel for what the historical data really tells us.  One data series I looked at was the earliest failure series.  IIRC (I can't see my data at the moment) it was 1968, and it was an obvious disaster from the start that failed after 12 years or so.  A 1968 retiree following a 4% withdrawal would have gone back to work after just a few years.  The other surprising one, though, was that there was a series (I don't know which one, sorry), that did manage to be at 125% of original stash at 5 years that ended up failing.

(And yeah, I knew about the other thread; I quoted from it in my original post :-)  Good discussion going on there also.)