Interesting article on 4% rule

[Andre1969]

I think one reason the 1999 cycle would have been so bad for me if I had retired then, is because I had three bad years in a row right off the bat. I was down about 5.4% in 2000, down another 30.4% in 2001, and finished 2002 down 23.1%. I don't know what the overall market return was, but that's what mine was. And pulling 4% adjusted for inflation each year would have devastated the portfolio.

While 2003-2007 were great years, I wouldn't have recovered enough, in time to ride out the Great Recession. As it is, I lost about 42% in 2008. So whatever 4% adjusting for inflation would have jacked up to by then would have really smacked it hard. Hard enough that the withdrawals from there on out would offset the gains I made in the following years.

But, I was also invested pretty risky during those times, if that cycle would ultimately fail, a good deal of the blame sits right on me!

 

[Bir48die]

I think one reason the 1999 cycle would have been so bad for me if I had retired then, is because I had three bad years in a row right off the bat. I was down about 5.4% in 2000, down another 30.4% in 2001, and finished 2002 down 23.1%. I don't know what the overall market return was, but that's what mine was. And pulling 4% adjusted for inflation each year would have devastated the portfolio.

While 2003-2007 were great years, I wouldn't have recovered enough, in time to ride out the Great Recession. As it is, I lost about 42% in 2008. So whatever 4% adjusting for inflation would have jacked up to by then would have really smacked it hard. Hard enough that the withdrawals from there on out would offset the gains I made in the following years.

But, I was also invested pretty risky during those times, if that cycle would ultimately fail, a good deal of the blame sits right on me!

I agree and this is not beyond the realm of possibilities. I don't know if I'm a realist or pessimist. Could be just an axietiest. But hitting at least part of an up cycle in the first few years makes so much difference in how the next decade or two can be managed.

If I had those losses early on 4% would not even have been close to getting it done. Timing is definitely everything along with some common AA sense

 

[Dtail]

I think one reason the 1999 cycle would have been so bad for me if I had retired then, is because I had three bad years in a row right off the bat. I was down about 5.4% in 2000, down another 30.4% in 2001, and finished 2002 down 23.1%. I don't know what the overall market return was, but that's what mine was. And pulling 4% adjusted for inflation each year would have devastated the portfolio.

While 2003-2007 were great years, I wouldn't have recovered enough, in time to ride out the Great Recession. As it is, I lost about 42% in 2008. So whatever 4% adjusting for inflation would have jacked up to by then would have really smacked it hard. Hard enough that the withdrawals from there on out would offset the gains I made in the following years.

But, I was also invested pretty risky during those times, if that cycle would ultimately fail, a good deal of the blame sits right on me!

So that is probably the difference in that it sounds like you were invested 80%+ in equities during this time.
If one is depending mostly on their portfolio for their withdrawals and invested 80% or so in equities, they are potentially subjecting themselves to a very risky scenario.

 

[corn18]

The SORR thing is my biggest worry bead heading into retirement. While I am happy when firecalc says 99%, I cannot imagine withdrawing the same amount in a year when the market dropped 40%. Yikes!

So we are establishing a lower risk portfolio for our base expenses. And a higher risk portfolio for our blow that dough expenses. In reality, though, it's all the same 60/40 portfolio. I just play mind games in that the 40% part covers 10 years of base expenses. I think we will use a form of VPW and if we hit a bad year at the beginning, I am certain that will be a lean year for spending.

 

[Navigator]

The one aspect of the 4% rule that has always confused me is this. Lets say I retired last year with $1 M and took out the 40,000. Assume inflation at 3% so next year my WD is $41,200. It's been a good year for the market so the balance has increased to $1.2 M. My Pal who also had $1 M last year (Exactly identical portfolios) Decides to retire this year so he takes his 4% of 1.2 M i.e. $48,000. Why can he take $48K plus inflation and live happily for 30 years while I can only take $41.2 K plus inflation and live happily for only 29 years (since I already used one of the 30). Doesn't make sense to me.

You raise an interesting question. I think the answer is that the 4% rule is based on the worst 30 year period in the study. In most cases, you can withdraw much more than 4% and be fine. Your question also points out another interesting point. The 30 year periods the 4% rule is based on are most definitely NOT independent. There is a 29 year overlap between your 30 year retirement and your Pal’s.

 

[Fermion]

You raise an interesting question. I think the answer is that the 4% rule is based on the worst 30 year period in the study. In most cases, you can withdraw much more than 4% and be fine. Your question also points out another interesting point. The 30 year periods the 4% rule is based on are most definitely NOT independent. There is a 29 year overlap between your 30 year retirement and your Pal’s.

Yes there is something there that is contaminating the data.

I think it does not make sense to say you can reset the 4% withdrawal rate at any arbitrary time and still maintain the 100% success rate because the data set has been compromised. It is sort of a Heisenberg Uncertainty Principle at play. If you adjust your withdrawal rate up during your retirement based on current market performance, then you are limiting the 30 year periods to only those which contain examples of that early market overperformance; thus the study covers many fewer examples and cannot be as reliable.

I may have just confused myself though.

Let me think it through more. Okay I see it like this:

Say you enjoy 5 really good market years and your porfolio grows from $1 million to $1.5 million. If you say that you are going to use the $1.5 million as your "new" starting amount, then you must consider that a calculator should use only those scenarios which contained 5 previous good years in addition to the remaining 30, or a 35 year study. This would limit the pool of studies and reduce the reliability of the overall prediction.

 

[Navigator]

The SORR thing is my biggest worry bead heading into retirement. While I am happy when firecalc says 99%, I cannot imagine withdrawing the same amount in a year when the market dropped 40%. Yikes!

So we are establishing a lower risk portfolio for our base expenses. And a higher risk portfolio for our blow that dough expenses. In reality, though, it's all the same 60/40 portfolio. I just play mind games in that the 40% part covers 10 years of base expenses. I think we will use a form of VPW and if we hit a bad year at the beginning, I am certain that will be a lean year for spending.

It’s natural to want to have a conservative portfolio before an anticipated downturn. The irony of this is that over a 30 year period, conservative portfolios are actually more risky than a 60/40 one. At least they are in the sense that they are more likely to run out of money than a 60/40 is for a given withdrawal rate.

Also, if you are willing to cut back on withdrawals when the market drops, you can actually begin with a much higher withdrawal rate than 4% and still be safe.

 

[Dtail]

Yes there is something there that is contaminating the data.

I think it does not make sense to say you can reset the 4% withdrawal rate at any arbitrary time and still maintain the 100% success rate because the data set has been compromised. It is sort of a Heisenberg Uncertainty Principle at play. If you adjust your withdrawal rate up during your retirement based on current market performance, then you are limiting the 30 year periods to only those which contain examples of that early market overperformance; thus the study covers many fewer examples and cannot be as reliable.

I may have just confused myself though.

Let me think it through more. Okay I see it like this:

Say you enjoy 5 really good market years and your porfolio grows from $1 million to $1.5 million. If you say that you are going to use the $1.5 million as your "new" starting amount, then you must consider that a calculator should use only those scenarios which contained 5 previous good years in addition to the remaining 30, or a 35 year study. This would limit the pool of studies and reduce the reliability of the overall prediction.

But the person who retires 5 years later and uses the 4% guidance. They will use it based on the 1.5m starting balance and Firecalc will not calculate his withdrawal scheme any different just because he had a most recent good 5 years.
Thus a variable type withdrawal strategy is more common such as VPW or % of current portfolio, etc.
The downside of these strategies is one has to be willing to make the necessary cuts in a down market similar to upping the withdrawals in good markets.

 

[Fermion]

But the person who retires 5 years later and uses the 4% guidance. They will use it based on the 1.5m starting balance and Firecalc will not calculate his withdrawal scheme any different just because he had a most recent good 5 years.
Thus a variable type withdrawal strategy is more common such as VPW or % of current portfolio, etc.
The downside of these strategies is one has to be willing to make the necessary cuts in a down market similar to upping the withdrawals in good markets.

But see, if you follow that line of thought, then you would also have to agree that a person whose portfolio falls in year one from $1,000,000 to $500,000 should need to adjust their $40,000 a year withdrawal down to $20,000 a year in order to be the same as a person who runs firecalc that year with a $500,000 starting portfolio. It is either that or claim that a 8% withdrawal rate has a 100% success rate.

 

[Dtail]

But see, if you follow that line of thought, then you would also have to agree that a person whose portfolio falls in year one from $1,000,000 to $500,000 should need to adjust their $40,000 a year withdrawal down to $20,000 a year in order to be the same as a person who runs firecalc that year with a $500,000 starting portfolio. It is either that or claim that a 8% withdrawal rate has a 100% success rate.

So using the strict 4% guidance with no inflation let's say, that person would still draw 40k the 2nd year even though his portfolio is now 500k which would be 8%.
But who in their right mind would actually draw 40k after his portfolio dropped 50% at the beginning of retirement.
Thus your example is just one of many reasons why strictly following the 4% guidance is really more of a guidance for being able to retire and not a WR strategy.
OTOH, the variable rate WR person would need to cut their spending down to 20k from 40k which unless most of it is discretionary would also be difficult.

 

[pb4uski]

...Let me think it through more. Okay I see it like this:

Say you enjoy 5 really good market years and your porfolio grows from $1 million to $1.5 million. If you say that you are going to use the $1.5 million as your "new" starting amount, then you must consider that a calculator should use only those scenarios which contained 5 previous good years in addition to the remaining 30, or a 35 year study. This would limit the pool of studies and reduce the reliability of the overall prediction.

But, I think that we would all agree that some stranger retiring at the point that you have $1.5 million can withdraw $60k a year adjusted for inflation safely... right? Also, you have 5 less years in the time horizon, so in your case the $1.5 million only needs to last 25 years and not 30 years.

We frequently ask... what does FIRECalc say?

According to FIRECalc someone retiring with $1 million and 4% WR ($40k first year withdrawal) and a 30 year time horizon and a 50/50 AA would have a 95% success rate. Someone retiring with $1.5 million and a 4% WR ($60k first year withdrawal) and a 30 year time horizon and a 50/50 AA would have a 95% success rate. That makes sense right?

Someone retiring with $1.5 million and a 4% WR ($60k first year withdrawal) and a 25 year time horizon and a 50/50 AA would have a 100% success rate... so that 17% reduction in the time horizon is a big deal... so it is safe to go ahead and ratchet.

 

[tmm99]

A great article, thank you for posting this, OP.

 

[Dtail]

But, I think that we would all agree that some stranger retiring at the point that you have $1.5 million can withdraw $60k a year adjusted for inflation safely... right? Also, you have 5 less years in the time horizon, so in your case the $1.5 million only needs to last 25 years and not 30 years.

We frequently ask... what does FIRECalc say?

According to FIRECalc someone retiring with $1 million and 4% WR ($40k first year withdrawal) and a 30 year time horizon and a 50/50 AA would have a 95% success rate. Someone retiring with $1.5 million and a 4% WR ($60k first year withdrawal) and a 30 year time horizon and a 50/50 AA would have a 95% success rate. That makes sense right?

Someone retiring with $1.5 million and a 4% WR ($60k first year withdrawal) and a 25 year time horizon and a 50/50 AA would have a 100% success rate... so that 17% reduction in the time horizon is a big deal... so it is safe to go ahead and ratchet.

Agree on that.
But if one only has a 15 year retirement and since the worst 15 year time period is worse than the worst 30 year time period, I wonder how much of this concept would negatively affect the success rates logically growing as one's remaining retirement period gets shorter.

 

[foxfirev5]

Once you've stopped your paycheck, you're going to need some money to live on right away. You don't have to take the full 4% at the beginning of that year, but you need something. Or you have to leave your first year of expenses out of your investments balance. I would like to see a study that uses everything else Bengen had but modified it for the reality of needing money right away.

Personally I am firmly committed to taking my IRA withdrawals near the end of the year. Prior to retirement we stashed enough cash to get us through the first year. Now at the end of the year I can manage my income and pay whatever I want in taxes. I wish I had that option while working. Huge swings in income that were beyond my control.

 

[Fermion]

But, I think that we would all agree that some stranger retiring at the point that you have $1.5 million can withdraw $60k a year adjusted for inflation safely... right? Also, you have 5 less years in the time horizon, so in your case the $1.5 million only needs to last 25 years and not 30 years.

We frequently ask... what does FIRECalc say?

According to FIRECalc someone retiring with $1 million and 4% WR ($40k first year withdrawal) and a 30 year time horizon and a 50/50 AA would have a 95% success rate. Someone retiring with $1.5 million and a 4% WR ($60k first year withdrawal) and a 30 year time horizon and a 50/50 AA would have a 95% success rate. That makes sense right?

Someone retiring with $1.5 million and a 4% WR ($60k first year withdrawal) and a 25 year time horizon and a 50/50 AA would have a 100% success rate... so that 17% reduction in the time horizon is a big deal... so it is safe to go ahead and ratchet.

But it *isn't* a 25 year time period. It is a 25 year time period with a 5 year lookback. You should only be able to use scenarios that contain a similar 5 good years, which reduces the available number of scenarios to run and reduces the firecalc accuracy.

Right? I dunno, maybe not.

I consider it similar to someone running Firecalc on a 70 year retirement. There just are not enough 70 year time periods available to reliably predict the future.

 

[pb4uski]

Agree on that.
But if one only has a 15 year retirement and since the worst 15 year time period is worse than the worst 30 year time period, I wonder how much of this concept would negatively affect the success rates logically growing as one's remaining retirement period gets shorter.

Doesn't matter... the shorter payout period overwhelms the aspect that you are referring to... using the same data and changing the time horizon to 20, 15, 10 or 5 years are all 100%.

Using your 15 year example, 15 years at $60k a years would be $900k of withdrawals but you're starting with $1.5 million... 167% of your withdrawals.

 

[pb4uski]

But it *isn't* a 25 year time period. It is a 25 year time period with a 5 year lookback. You should only be able to use scenarios that contain a similar 5 good years, which reduces the available number of scenarios to run and reduces the firecalc accuracy.

Right? I dunno, maybe not.

I consider it similar to someone running Firecalc on a 70 year retirement. There just are not enough 70 year time periods available to reliably predict the future.

In high theory you might have a minor point if and only if reversion to the mean is real... but if what you say is correct the the 4% rule should be 4% if the last 5 years have been normal, less than 4% if the last 5 years have been good and more than 4% if the last 5 years have been bad... but I haven't heard anyone advocate that and it seems a bit silly.... especially when the 4% is effectively based on the worst case scenario of returns and the vast majority of scenarios end up with the retiree having a lot of money at the end of the 30 years and only a handful fail.

 

[Fermion]

Well, let us use this year as an example. We have had a large gain in the market, roughly 20% for quite a few people.

So, if last year I ran Firecalc with $1,000,000 and a $40,000 withdrawal rate for a 30 year retirement, I got a 95% success rate (6 failures)

This year with $1,160,000 (20% return - $40,000) if I run Firecalc with a $40,800 withdrawal rate and a 29 year retirement, I get a 100% success rate (0 failures)

So one good year puts me at 100% success rate, or is it still the original 95% success rate?

Can you start with a 100% success rate and find a bad first year that drops you below 100% if you run Firecalc again with the 29 year horizon? I bet you can, which leads much doubt to the 100% to begin with.

 

[Fermion]

I just ran a 2009 scenario.

Retire in 2008 with $1,000,000 and a $35,000 withdrawal rate for 30 years, for a 100% success in Firecalc

2009 returns drop your portfolio to $700,000. Running Firecalc again with 2% inflation for a $35,700 withdrawal rate and a 29 year time horizon gives you a success rate of only 71.7%

So was Firecalc wrong in 2008 on the 100% prediction?

 

[corn18]

I just ran a 2009 scenario.

Retire in 2008 with $1,000,000 and a $35,000 withdrawal rate for 30 years, for a 100% success in Firecalc

2009 returns drop your portfolio to $700,000. Running Firecalc again with 2% inflation for a $35,700 withdrawal rate and a 29 year time horizon gives you a success rate of only 71.7%

So was Firecalc wrong in 2008 on the 100% prediction?

Firecalc uses data to drive results, so I think there are many scenarios where a 40% drop in equities is followed by above average returns for the next X years.

I can tell you that a 40% drop in equities followed by average returns is a retirement killer, at least in my model. A modest increase in average returns following a 40% drop really helps out a lot. But if the market drops a lot and then stays average, we will need to cut spending to survive.

 

[RunningBum]

Personally I am firmly committed to taking my IRA withdrawals near the end of the year. Prior to retirement we stashed enough cash to get us through the first year. Now at the end of the year I can manage my income and pay whatever I want in taxes. I wish I had that option while working. Huge swings in income that were beyond my control.

That's fine. But what did you do when you first retired? If you retired in the middle of the year, you couldn't just wait until the end of the year to take a withdrawal to live on, since you need money to eat, etc. That's what the Bengen study seems to do though. You can retired on $1M at the start of the year, and withdraw $25K at the end of the year. What did you live on during that year?

More specifically I'm not talking about a calendar year. You retire on Day 1, Year 1, which can be any day of the year. Bengen says you can take out 4% at the end of the year, Day 365. In reality, you have to take out or set aside money for year 1 on day 1, not day 365.

Bengen is basically skipping a year of withdrawals. That's cheating. You simply can't live this way. As far as I know Bengen doesn't give a caveat that you set aside your first year of expenses and then withdraw 4% at the end of every year.

 

[pb4uski]

Well, let us use this year as an example. We have had a large gain in the market, roughly 20% for quite a few people.

So, if last year I ran Firecalc with $1,000,000 and a $40,000 withdrawal rate for a 30 year retirement, I got a 95% success rate (6 failures)

This year with $1,160,000 (20% return - $40,000) if I run Firecalc with a $40,800 withdrawal rate and a 29 year retirement, I get a 100% success rate (0 failures)

So one good year puts me at 100% success rate, or is it still the original 95% success rate?

Can you start with a 100% success rate and find a bad first year that drops you below 100% if you run Firecalc again with the 29 year horizon? I bet you can, which leads much doubt to the 100% to begin with.

Of course it is 100% for the second scenario... because the WR in the second scenario is only 3.5% ($40.8k/$1,160k) and 3.5% is bulletproof for a 29 year time horizon.

If you look at the scattergraph for the $1 million base scenario and follow the lines, you'll see that the failures are at the end of year 25, ~27.5 (looks like 3) and a couple at the very end.

The same 4% WR with a 29 year time horizon has a success rate of 96.7% and 4 failures, because the couple at the very end drop off and are successes at 29 years.

 

[teejayevans]

People, it’s a study, the parameters have to be fixed and consistent so that the model is easy to program, you are reading too much into it. As I said before, it’s a good rule of thumb.I’ve taken out much more than 4% and less depending on the market returns, but I still use 4% as a target at the beginning of the year.

 

[pb4uski]

I just ran a 2009 scenario.

Retire in 2008 with $1,000,000 and a $35,000 withdrawal rate for 30 years, for a 100% success in Firecalc

2009 returns drop your portfolio to $700,000. Running Firecalc again with 2% inflation for a $35,700 withdrawal rate and a 29 year time horizon gives you a success rate of only 71.7%

So was Firecalc wrong in 2008 on the 100% prediction?

No, because you don't know what happened subsequent to the first year.

However a real life example using Portfolio Visualizer and a 50/50 AA proves that 4% would be ok even if one retired in 2008.

The $1,000k would have dipped down to $724k at the end of Feb 2009 but would be $1,385k at the end of Oct 2019.

https://www.portfoliovisualizer.com...location1_1=50&symbol2=VBMFX&allocation2_1=50

 

[Andre1969]

I just ran a 2009 scenario.

Retire in 2008 with $1,000,000 and a $35,000 withdrawal rate for 30 years, for a 100% success in Firecalc

2009 returns drop your portfolio to $700,000. Running Firecalc again with 2% inflation for a $35,700 withdrawal rate and a 29 year time horizon gives you a success rate of only 71.7%

So was Firecalc wrong in 2008 on the 100% prediction?

I know I've asked this before, but at what point in the year does FireCalc presume you retire? That can make a huge difference.


For instance, on the first trading day of January, 2008, my balance was around $415K. On the last trading day of December 2008, it was around $260K.

On the first trading day of January 2009, I had that ~$260k, but at the end of the year I was around $460K.


BTW, in your example above, if you re-run FireCalc again for the 29 year scenario, I think you need to reset the initial withdrawal back to 3.5%. So it would be $700K for 29 years, which would be $24,500. $35.7K would be 5.1%.

I think the reason the success rate drops off so much is partly the higher initial withdrawal rate, but also, in the 30 year cycle, you just retired going into one of the worst one-year market drops in history. Chances are, you aren't going to do that again in the near future, and the market will most likely recover fairly quickly. But the 29 year cycle starts all over again, and throws in the possibility of you again, retiring into one of the worst downturns in history, but at a higher withdrawal rate, so you're less likely to recover.


At least, that's my rationale. I'm sure there's holes that can be poked in it