The Four Percent Rule

[pb4uski]

I guess one way to reset would be to simply run a new Firecalc analysis as if I was retiring anew with my then current portfolio and time horizon and use that to calculate a new sustainable withdrawal. I think it is the same thing.

I'm willing to accept a slightly higher chance of failure in exchange for a significantly lower chance of underspending.... especially since I'm targeting a 98% success rate which implicitly include some built-in conservatism.

 

[growing_older]

Here is the problem with ratcheting, which I apologize if my explanation is not clear enough. When the FireCalc examples show 95% success rate, that is taking a random starting point anywhere in the historic sequence of returns. You can think of this as 95 of the starting points are green and 5 are red. We don't actually know which years are green or red going forward, because we don't yet know the future sequence of returns, but for calculation purposes FireCalc just uses all of them, combines the results as a percentage success.

Now in a usual 30 year retirement, as you go though the red years in the sequence (years someone who started their retirement then would have a problem) as an existing retiree, you probably are fine because previous years have been good to you or your remaining retirement is short enough that the bad sequence starting on that red year won't be long enough before your retirement ends and you heirs get whats left. FireCalc still counts that as success as long as you end with more than zero.

By ratcheting, you are eliminating the case above where previous years have been good to you and your pile is more than big enough to power through the red year starting point, because you effectively re-started on the new ratchet higher amount. Also if you are an early retiree (depending on how early) you may not escape zero because you are not the case where remaining retirement is short enough that the bad sequence won't be long enough to hurt you (your retirement is still 30+ years to go). In effect you start on a green year, possibly a very green year, and by ratcheting you move to a new starting point each year in the sequence that is higher, until finally you ratchet up to start on a red year. At that point if your remaining retirement is short enough, you make it (maybe barely). But if your retirement is long enough, the ratchet has pushed you to select a red (bad) year by constantly switching your starting point to any year that was more red-ish than your previous starting year. If you eventually hit a red one given your length of retirement remaining, you have adversely selected yourself into a failing sequence of returns.

This is how ratcheting will push you into a failing sequence for any withdrawal rate that is historically less than 100% success. Of course a future sequence of returns could be worse than the worst recorded in FireCalc, so even a 100% success rate starting withdrawal rate could still push you to find that failure case, but the hope is that the actual chance of that is rare.

Note that all of this is NOT a problem if you have a sufficiently effective means of REDUCING your spending when a bad sequence of returns is occurring. All the failures shown are because the spending rate cannot be supported by future returns - in large part because the spending is assumed to be inflexible.

 

[ERD50]

I guess one way to reset would be to simply run a new Firecalc analysis as if I was retiring anew with my then current portfolio and time horizon and use that to calculate a new sustainable withdrawal. I think it is the same thing.

I'm willing to accept a slightly higher chance of failure in exchange for a significantly lower chance of underspending.... especially since I'm targeting a 98% success rate which implicitly include some built-in conservatism.

Yes, I think you can find some postings on a 'retire again and again' strategy. Essentially reset/ratchet on each new high. Once you get your head around it, you see it has to work (with the data provided). It also resolves the apparent paradox of the two retirees who retire a few years apart after a large market shift.

-ERD50

 

[ERD50]

...
Now in a usual 30 year retirement, as you go though the red years in the sequence (years someone who started their retirement then would have a problem) as an existing retiree, you probably are fine because previous years have been good to you or your remaining retirement is short enough that the bad sequence starting on that red year won't be long enough before your retirement ends and you heirs get whats left. FireCalc still counts that as success as long as you end with more than zero. ...

I didn't read the rest, because this is just incorrect. It's not going to help clarify anything.

The red years are the failures. period.

All the years assume the same starting portfolio, there is no adjustment for 'previous years that have been good to you' - those are included as successes in other runs, they make up the 95%.

Please try again.

-ERD50

 

[pb4uski]

+1 If a new retiree has a 95% success rate then a ratcheting retiree with the same WR also has a 95% success rate if their portfolio balance, AA, time horizon, etc. are the same.... to claim differently is poppycock.

 

[samclem]

I guess one way to reset would be to simply run a new Firecalc analysis as if I was retiring anew with my then current portfolio and time horizon and use that to calculate a new sustainable withdrawal. I think it is the same thing.

At what point do we just call it what it is: "I'm now using a 'percent of portfolio year end value rather than a starting amount adjusted annually for inflation" ?

 

[ERD50]

At what point do we just call it what it is: "I'm now using a 'percent of portfolio year end value rather than a starting amount adjusted annually for inflation" ?

Because it's different?

A 'percent of portfolio year end value' would have you adjusting down as well. This is ratcheting, a 'peak-hold' function if you will. And it works, within the historical data.

If you use this approach on the bad sequences, you never increase (or decrease) your initial withdraw amount, other than for inflation (same as default). It only applies when the inflation adjusted portfolio rises.

-ERD50

 

[pb4uski]

At what point do we just call it what it is: "I'm now using a 'percent of portfolio year end value rather than a starting amount adjusted annually for inflation" ?

Absolutely not... because it is a blend of the two.

In good years there will be a reset, in sideways or bad years the withdrawal will increase with inflation based the 4% rule and the last reset. And the measurement is based on performance since the last reset date (be it a year ago or many years ago) and today.

I like the recalculate version slightly better than just 4% of new balance version because it takes into account both growth in the portfolio value and the reducing time horizon but I'm not sure there is a huge difference between them.

To me the big problem with use portfolio year end value is that you will likely die rich since the average annual return exceeds the WR. I guess that you could mitigate that by shifting to a RMD approach at a target age.

 

[samclem]

A 'percent of portfolio year end value' would have you adjusting down as well. This is ratcheting, a 'peak-hold' function if you will. And it works, within the historical data.

I hear the data being tortured. "Help"

 

[audreyh1]

I guess one way to reset would be to simply run a new Firecalc analysis as if I was retiring anew with my then current portfolio and time horizon and use that to calculate a new sustainable withdrawal. I think it is the same thing.

I'm willing to accept a slightly higher chance of failure in exchange for a significantly lower chance of underspending.... especially since I'm targeting a 98% success rate which implicitly include some built-in conservatism.

Personally, I thought the easiest way was to just do away with inflation adjustments altogether and just stick with a fixed % of remaining portfolio. My spending doesn't immediately jump up to match an increase in income after a good year, so unspent funds are accumulated to help with future lower income after lean years.

To me the big problem with use portfolio year end value is that you will likely die rich since the average annual return exceeds the WR. I guess that you could mitigate that by shifting to a RMD approach at a target age.

I plan to address this in the future by reviewing my % withdrawal and increasing it if I decide it is warranted.

True - it perhaps means I spend less today than I could, but since my withdrawals have been way outpacing my actual spending anyway, the point is moot for me. (knock on wood!!!)

All I need is for a big permanent market selloff to make my withdrawals match my actual spending.

I hope this is not a jinx!!

 

[ERD50]

I hear the data being tortured. "Help"

Yes, you can call this 'data mining' if you want. I've considered that.

I'm not sure it is 'data mining' any more than a regular run though. It's all the data we got, and we can only get so much info out of it. You can think of the ratcheting like this:

If you retired at a historical market peak, that will likely be one of the bad sequences. Yet, FIRECalc will show that a WR of ~ 3.2% will be 100% safe (historically).

But, if you retired at a historical low point, right before a big lift in the market, FIRECalc gives you the same ~ 3.2% as 100% safe, and you will end up with a very large portfolio unspent. It doesn't discriminate. The 3.2% is (historically) 100% safe. So going forward along those historical sequences, it is always safe, so you can raise your withdraw amount at any peak (they are safe at 3.2% as well).

And I guess, what concerns me overall regarding this data, is when you do the long runs, the success is generally because we had a good run after a bad run. A mediocre period after a bad run would probably sink a lot of ships. While we may have 100+ sequences in the data, we probably really only have about 4 economic cycles. The future could look very different.

But this gets back to the age-old debate here - how safe do you need to be? Work till you die, no worries of needing to rely on your portfolio. At some stage, we decide (or have it decided for us) and make the best of it.

-ERD50

 

[ERD50]

Personally, I thought the easiest way was to just do away with inflation adjustments altogether and just stick with a fixed % of remaining portfolio. My spending doesn't immediately jump up to match an increase in income after a good year, so unspent funds are accumulated to help with future lower income after lean years.

But if you 'accumulate' them, you aren't really doing what you are saying. Effectively, you didn't withdraw them, so it isn't a 'fixed % of remaining portfolio'.

Not that there's anything wrong with your approach, but it's confusing to call it something it isn't.

I think what you are saying is that you will cut spending in bad times, and bank the excess in good times. The normal inflation adjusted withdraw approach does the banking as well, it just starts with a point that meets your safety level (historically), and has demonstrated that will survive through the bad times, without cuts.

So do you start with a higher rate, figuring the cuts and banking will support it? Or just being conservative overall?

Again, nothing wrong with either approach, it's a personal comfort level decision.

-ERD50

 

[audreyh1]

But if you 'accumulate' them, you aren't really doing what you are saying. Effectively, you didn't withdraw them, so it isn't a 'fixed % of remaining portfolio'.

Not that there's anything wrong with your approach, but it's confusing to call it something it isn't.

I think what you are saying is that you will cut spending in bad times, and bank the excess in good times. The normal inflation adjusted withdraw approach does the banking as well, it just starts with a point that meets your safety level (historically), and has demonstrated that will survive through the bad times, without cuts.

So do you start with a higher rate, figuring the cuts and banking will support it? Or just being conservative overall?

Again, nothing wrong with either approach, it's a personal comfort level decision.

-ERD50

No - I am withdrawing all the funds. They are no longer part of my retirement portfolio. They are no longer counted when I compute my withdrawal as % of my retirement portfolio the next year.

Unlike some folks, I do not return unspent funds to my portfolio. Once withdrawn, I choose not to expose unspent funds to the market risks of my retirement portfolio which is invested for the long term. I prefer to have them available for near term spending.

No - I don't expect to cut spending in bad times (unless really prolonged), so I'm not saying that either. My income will be cut, but perhaps not my spending.

I'm just conservative for my personal comfort. It's my version of belts and suspenders. Even more conservative in that not all of my investable assets are counted in my retirement portfolio. I have other assets that I many spend however I please whenever I please. But I have a set that are cordoned off, labeled "retirement portfolio" and I withdraw a fixed % of that every January to live on.

BTW - there is no rule that says all investable assets must be counted when one computes their annual withdrawal. I know many folk here do it that way. But I never have. I always had a subset of my assets labeled "retirement portfolio" and that is what I apply the withdrawal rules on.

 

[growing_older]

I didn't read the rest, because this is just incorrect. It's not going to help clarify anything.

Sorry, if this was unclear, but it is absolutely correct. FireCalc uses historic sequences of returns, and some of those years are years in which if that is the year you start and follow the 4% plus inflation, your portfolio will not last 30 years. If you start in an earlier year, you succeed because portfolio growth (plus shortened retirement) was sufficient to support the 4% plus inflation through the bad sequence. If you restart and have a long enough retirement still ahead of you, you cannot get this insulating effect and your 4% plus inflation is too much for the sequence that starts on that "bad" year.

I apologize (again) if this is unclear, but if so it is the fault of the explanation, not the mathematics. Adverse selection is a real phenomena.

 

[ERD50]

Sorry, if this was unclear, but it is absolutely correct. FireCalc uses historic sequences of returns, and some of those years are years in which if that is the year you start and follow the 4% plus inflation, your portfolio will not last 30 years. If you start in an earlier year, you succeed because portfolio growth (plus shortened retirement) was sufficient to support the 4% plus inflation through the bad sequence. If you restart and have a long enough retirement still ahead of you, you cannot get this insulating effect and your 4% plus inflation is too much for the sequence that starts on that "bad" year.

I apologize (again) if this is unclear, but if so it is the fault of the explanation, not the mathematics. Adverse selection is a real phenomena.

What it seems you are missing is, those scenarios are already in the runs that make up the successes and failures. If you 'start in an earlier year' - that is just another one of the squiggly lines, you don't re-run anything, it was already run.

No - I am withdrawing all the funds. They are no longer part of my retirement portfolio. ...

BTW - there is no rule that says all investable assets must be counted when one computes their annual withdrawal. I know many folk here do it that way. But I never have. I always had a subset of my assets labeled "retirement portfolio" and that is what I apply the withdrawal rules on.

So you take money out of your portfolio, it is available for you to spend, but you no longer consider it 'part of your portfolio'? That really redefines what a success or failures is. Your 'portfolio' could fail, but you will still 'succeed', because you have a stash accumulated - from your 'portfolio' (that maybe would not have failed if you didn't take money out to accumulate it somewhere else)?

Yes, there are 'rules' that FIRECalc and the Trinity study uses. Without 'rules' we can't know how the study is done or what is being demonstrated. Of course there is no 'rule' on what you decide to do, but if you are trying to communicate with others on this forum, I think people expect 'portfolio' to mean 'portfolio'. It's almost getting into an Alice in Wonderland situation on the meaning of words.

-ERD50

 

[liberty53]

Ratcheting up sounds like a great approach - after all if you re-run firecalc with current numbers and your desired increase in WR is still within acceptable probability of portfolio failure (say <= 5%) then why not enjoy the extra spend?

Ratcheting handles those firecalc curves that show you dying with way more money than your wife or kids deserve!

As a rule of thumb - it might be good to run firecalc every year to make sure your initial plan at retirement is still valid. After all, things can change. Assumptions about pensions and social security are just that - assumptions. Occasionally Re-running firecalc on your current stash to make sure your withdrawal rate is still in that <= 5% failure probability range would make a lot of sense.

Imagine some unexpected event that depletes your portfolio more than expected. Say somebody breaks their neck on your property and you have an uncovered judgment of $500k against you that you have to pay. It would be wise to re-run firecalc to find out what a reasonable withdrawal rate would be based on your newly reduced stash.

In other words - it seems prudent to keep running firecalc throughout retirement to do a sanity check on your specific circumstances. If you are doing fantastic and want to increase WR - so be it. If things haven't been going well and you are on the path to failure - adjust fire and consider contingency plans.

I'm not sure this is the same as the "retire again and again" approach - it might be.

 

[samclem]

Yes, you can call this 'data mining' if you want. I've considered that.
I'm not sure it is 'data mining' any more than a regular run though. It's all the data we got, and we can only get so much info out of it. You can think of the ratcheting like this:

If you retired at a historical market peak, that will likely be one of the bad sequences. Yet, FIRECalc will show that a WR of ~ 3.2% will be 100% safe (historically).

Assuming our retiree has a fair slug of equities in his AA, we know that retiring at time when they have been bid up to very high valuations increases the chance of a nasty dip, maybe in the early years, that increases the chance of portfolio failure. Now, our very general rule of thumb (95% survival rate) is based on a broad data set that includes all types of starting valuations. If we instead used just those cases that started in periods of high stock valuations, then our success rate would be lower. Assuming a normal AA, what this "ratcheting" we're talking about does is to deliberately reset the the withdrawal amount after during periods of higher valuation (because that's the main thing that causes the portfolio value to spike, and which will trigger a "re-ratchet")--never down, only up. Is it realistic to think that the "general rule" 95% survival will apply?

"Old man, I'm told you track the success of all boats which sail from this shore to cross the ocean. I want to do that. What are my chances for success?"
"Well, Young Snapper, 95% of the boats I've tracked over many years have made it across. Big and small, fancy and plain--overall it is 95%"
"That sounds great!. Are there some boats that do less well?
"Yes. The boats made by the "Very High Valuation" shipyard fail to make it across at a much higher rate. They are loaded with nice things inside--cappuccino bars, heavy brass ornamentation, etc. They sit low in the water, and when a storm comes up, sometimes they have insufficient freeboard. That's how you can tell the most luxurious boats--they sit low in the water."
"Noted. They sound like very comfortable boats. Here's my plan: I'll set sail, and every time I see a boat that sits lower than the one I'm on, I'll trade and get aboard that one. I'll keep trading for more luxurious boats every chance I get. So, old man, my chances are still 95%, like you said, right?"
"Hmmm. Would you like to play a little poker before you set sail?"

I realize I sound uncomfortably like H***s!

 

[W2R]

I'm headed out and in a rush, but just wanted to quickly add that I like to compute my withdrawal as a percentage of my 12/31 portfolio value. Normally I keep that % low enough that it also is below the CPi adjusted 4%SWR, as well as being below 4% of my lowest balance in 2008-2009 and below my dividend yield.

2015 was an exception for me, since I bought my house that year. But otherwise I've been pretty much a belt and suspenders type of person too.

 

[NW-Bound]

+1 If a new retiree has a 95% success rate then a ratcheting retiree with the same WR also has a 95% success rate if their portfolio balance, AA, time horizon, etc. are the same.... to claim differently is poppycock.

It is the same, but at the same time it is not. Both cases may survive a 30-year retirement (the same part), but one likely with a lot of surplus, the other barely scraping by at the end (the different part).

Somewhere in between is best. So, one can ratchet up but not all the way, because that increases the chance he will be sweating bullets in his old age, or if he happens to live longer than originally thought.

 

[ERD50]

...
"Hmmm. Would you like to play a little poker before you set sail?"

...

Where your analogy falls apart is that ALL those sailing runs are already in the data set that was tested.If you apply those rules to the 5% historical failures, you'll find that they never get to ratchet up anyhow, they never showed an excess - it's irrelevant to those scenarios.

What it does historically, is to pull the excess from some successful runs, and the final distribution becomes tighter.

Unquestionably, in real life with an uncertain future, increasing your withdrawals at any time (by any method) will decrease your chances of success if everything goes very bad in the future (worse than the worst of the past). No different than any other approach. No free lunch.

But I also consider that a ratchet approach starting with a relatively safer initial WR% will likely be safer than a fixed, but higher initial WR%. The ratchet method allows for increases if things go well, the fixed amount might hurt you beyond repair in the early years.

-ERD50

 

[photoguy]

This is how ratcheting will push you into a failing sequence for any withdrawal rate that is historically less than 100% success.

I think this is a good explanation and how I see the impact of ratcheting. Basically we're all going to experience a few more big nasty bear markets in our lifetime, and ratcheting guarantees we'll reset to a 4% WR right before that.

I don't think that it's necessarily a bad approach though, because as noted, you can reduce spending. And of course, every new year is one less year that you need to pay for.

 

[NW-Bound]

Where your analogy falls apart is that ALL those sailing runs are already in the data set that was tested.If you apply those rules to the 5% historical failures, you'll find that they never get to ratchet up anyhow, they never showed an excess - it's irrelevant to those scenarios.

What it does historically, is to pull the excess from some successful runs, and the final distribution becomes tighter...

Yes. I have said the same. With full ratcheting, all runs will converge to the worst run. The one where one is exactly broke at 30 years. That leaves no margin left for extra longevity, or market getting worse than the worst of the past years.

Again, ratchet is OK, but don't make it so tight that it leaves you no wriggling room later on.


PS. I assume that one does "ratcheting" he accounts for the shorter remaining period. The increase in WR due to fewer years left is quite legitimate, eg. if you have one year left, WR should be 100% and no FIRECalc is needed. I was just talking about the increase of WR in dollar amounts due to portfolio outperformance in bull years.

 

[ERD50]

Ratcheting up sounds like a great approach - after all if you re-run firecalc with current numbers and your desired increase in WR is still within acceptable probability of portfolio failure (say <= 5%) then why not enjoy the extra spend?

Ratcheting handles those firecalc curves that show you dying with way more money than your wife or kids deserve! ....

You got it!

Easiest to explain starting with a 100% historically successful WR, I think ~ 3.2% with other things at default on FIRECalc. All those runs succeed. Jumping in 'a few years before, or after', is just the same as starting the run that began that year - and it succeeded, just like all the other ones!

Yep, it just pulls excess from the very succesful runs, and never pulls from the ones on the edge of failure. If those edge-of-failure runs could succeed with more pulled from them, then FIRECalc would have provided a higher SWR under the 'investigate spending for 100% success' scenario! But they didn't, so you can't, and the ratchet method won't either, cause it's working with the same data set.

As a rule of thumb - it might be good to run firecalc every year to make sure your initial plan at retirement is still valid. After all, things can change. Assumptions about pensions and social security are just that - assumptions. Occasionally Re-running firecalc on your current stash to make sure your withdrawal rate is still in that <= 5% failure probability range would make a lot of sense.

Imagine some unexpected event that depletes your portfolio more than expected. Say somebody breaks their neck on your property and you have an uncovered judgment of $500k against you that you have to pay. It would be wise to re-run firecalc to find out what a reasonable withdrawal rate would be based on your newly reduced stash.

In other words - it seems prudent to keep running firecalc throughout retirement to do a sanity check on your specific circumstances. If you are doing fantastic and want to increase WR - so be it. If things haven't been going well and you are on the path to failure - adjust fire and consider contingency plans.

I'm not sure this is the same as the "retire again and again" approach - it might be.

Yes, that wasn't really the intention of 'retire again & again', that was meant as a static apples-apples comparison with the same data set, but it still makes sense to me to utilize it in that was as well.

Things change, stuff happens - adapt or die! (OK, that's a little too serious!)

It is the same, but at the same time it is not. Both cases may survive a 30-year retirement (the same part), but one likely with a lot of surplus, the other barely scraping by at the end (the different part).

Somewhere in between is best. So, one can ratchet up but not all the way, because that increases the chance he will be sweating bullets in his old age, or if he happens to live longer than originally thought.

Exactly. I don't think anyone will blindly follow any plan. This is just another way to re-evaluate as you go. Things are looking pretty good for me, unless some unexpected expenses come my way, or the market goes really bad. So I think this will likely be more along the lines of taking a conservative WR and feeling even more secure in it, and maybe feeling OK to raise it, but still stay in 'conservative territory'.

-ERD50

 

[W2R]

I think this is a good explanation and how I see the impact of ratcheting. Basically we're all going to experience a few more big nasty bear markets in our lifetime, and ratcheting guarantees we'll reset to a 4% WR right before that.

I don't think that it's necessarily a bad approach though, because as noted, you can reduce spending. And of course, every new year is one less year that you need to pay for.

This is why it is so smart to live a modest enough lifestyle in retirement, that you have plenty of money for discretionary stuff after the bills are paid. Makes it easier to cut back.

 

[NW-Bound]

... just wanted to quickly add that I like to compute my withdrawal as a percentage of my 12/31 portfolio value...

If one withdraws a fixed percentage of the current portfolio value, she will never run out of money. The only problem is that in lean years, one can be very hungry.

From real-life data I mentioned in an earlier post,

Here are some actual numbers. On 10/9/2007, the S&P was at 1565. On 03/09/2009, the S&P was at 677! That's a huge drop of 57%.

Ignoring the bitty dividend, a $1M at the top of the market would become $430K later. So, if you retired at the bottom of the market, you could withdraw only 4% x $430K = $17K instead of $40K/yr.

So, a retiree drawing 4% of current portfolio started out with $40K in 2007, she would have to make do with $17K in 2009.

I have fluff in my expenses, but I do not know if I can cut down to less than 1/2 of what I normally spend. It would really require me to let go of both houses and go live in the motorhome. Even then, I would really need subsidy for my $38K/year healthcare cost ($24K premium + $14K deductible).