USA Today article on SWR

[Independent]

T

(Mr Independent, you'll also find the answers to your questions, btw)

I looked at the paper, but didn't find the answers to my questions.

The typical 4% SWR assumes that I will take $40k in the first year and increase it by the CPI every year, even if investment returns are very poor. Typical models and assumptions might show that I have a 5% chance of running out of money if I live 30 years.

This one says that if I have some downside flexibility, I can start at 5% or $50k. But, how much flexibility do I need? Suppose I think my basic needs are $20k, can I "safely" take $50k in the first year and apply these rules, knowing that they will never require that I live on less than $20k? What if my basic needs are $30k, can I still start at $50k? What if I can't live on less than $40k, ... oops, I know the answer to that one, I can't safely start at $50k.

These Monte Carlo models are generating the answers in their calculations, I don't see the author reporting them.

 

[siamond]

I looked at the paper, but didn't find the answers to my questions.

Ah yes, you're right. I had noticed that the comments associated with Figure 3 do describe the lowest withdrawals that might happen. But didn't pay attention that the (more interesting) following figures do not have similar comments.

Personally, I am not overly fazed, as tough times would demand tough measures (work again, gasp! possibly part-time), but I'm with you, I'd love to see more details.

Well, either one of us buys the simulator from Mr Klinger, or we can convince the Firecalc fellows to beef up the spending model tab (and underlying math)! Which would be very cool, but of course, it's easy to say.

PS. check Mr Klinger's Web site, there is more data in there. Thanks for the pointer, bmcgonig.

 

[pb4uski]

I read the paper and it just seemed to confirm with analysis what many of us have been saying - if one has a high withdrawal rate that you need to have some flexibility to reduce withdrawals if returns are poor and you may be able to increase withdrawals if returns are good. This is similar to the authors' capital preservation and prosperity rules.

Actually, I think the authors would say that if you can live on $40k you could probably start at $50k since you could absorb up to two 10% cuts under the capital preservation rule and still have enough to live on.

 

[Independent]

I read the paper and it just seemed to confirm with analysis what many of us have been saying - if one has a high withdrawal rate that you need to have some flexibility to reduce withdrawals if returns are poor and you may be able to increase withdrawals if returns are good. This is similar to the authors' capital preservation and prosperity rules.

Actually, I think the authors would say that if you can live on $40k you could probably start at $50k since you could absorb up to two 10% cuts under the capital preservation rule and still have enough to live on.

Suppose I have a model and a set of assumptions that say a constant real $40k withdrawal will fail in 5% of the 30 year scenarios.

I'm pretty sure that if I start with $50k, use the authors rules, but put a floor of $40k on my withdrawals, than I will fail more than 5% of the time.

I don't see the free lunch.

 

[Independent]

Ah yes, you're right. I had noticed that the comments associated with Figure 3 do describe the lowest withdrawals that might happen. But didn't pay attention that the (more interesting) following figures do not have similar comments.

Personally, I am not overly fazed, as tough times would demand tough measures (work again, gasp! possibly part-time), but I'm with you, I'd love to see more details.

Well, either one of us buys the simulator from Mr Klinger, or we can convince the Firecalc fellows to beef up the spending model tab (and underlying math)! Which would be very cool, but of course, it's easy to say.

PS. check Mr Klinger's Web site, there is more data in there. Thanks for the pointer, bmcgonig.

If you're good with Excel, I think you can duplicate FireCalc for a little testing with an Excel spreadsheet. The key is that FireCalc gives you the option of downloading a detailed sample calculation, and you can back out the formulas, extract the assumptions, and change the spending rule.
(I tried this some years ago and I think I was fairly successful.) Of course, then we get into the "easy" issue

 

[Rambler]

I'm going to try to hang on to my 2.75-3% rate for now. If Mr Market performs really well over the next 10 years, I might buy a toy or two here and there...like a diesel pusher RV, which would throw the percentage way out of whack for a year or so. I like the idea of not having to worry about having money when I need it, and I am pretty confident that we will be quite happy with our ability to spend at that WR. That said, this year is a bit of a nightmare, as we are doing a lot of repairs and fixits on our home that we left vacant for the past 8years while on assignment overseas. The compensated non-compete is taking care of that. Next year, we won't have those expenses, for the most part, and spending will be way down.

R

 

[ERD50]

If you're good with Excel, I think you can duplicate FireCalc for a little testing with an Excel spreadsheet. The key is that FireCalc gives you the option of downloading a detailed sample calculation, and you can back out the formulas, extract the assumptions, and change the spending rule.
(I tried this some years ago and I think I was fairly successful.) Of course, then we get into the "easy" issue

While it's not automatic, I finally got around to creating some runs to take a stab at this. I started with that 1973 output, and you see (and it can be seen on the graphs, kinda, if you squint), that the portfolio takes a major hit in the first few years, and that's is hard to recover from.

So here's what I got (maybe this should be a separate thread?):

I first found that you can hit 100% success with a straight $35,947/$1,000,000 spend/portfolio (so 3.595% WR).

Since that drop hits early, I modeled starting at $40K/$1M (4%), cutting to $20K (2%) in 2015. Through trial/error found I needed to do this for so 6 years just to get back to 100%. I 'undid' the spending cut by adding it back as pension in year 2021.

FIRECalc: 4% - 2% - 4%

40K/1M, rest are defaults, then;

$20,0000 income Infl Adjusted start 2015 (effectively cuts spending in half)

$20,0000 spending Infl Adjusted start 2021 (effectively returns spend to original)

______________________________


try similar, start at 5%, cut in half, return to 4%

FIRECalc: 5% - 2.5% - 4%

Start at 5%, cut to 2.5% in 2015, hold thru 2024 (ten years inclusive), then return to 4% of initial. Took 10 years to hit 100% success.

50K/1M defaults;

$25,0000 income Infl Adjusted start 2015 (effectively cuts spending in half to $25,000)

$15,0000 spending Infl Adjusted start 2024 (effectively returns spend to $40K - 4%)

______________________________



Using Investigate - for 100% straight, 3.59% spending through-out.

So let's SS this and see what the total spend (in buying power) would be summing all 30 years:

Straight 3.5947%                 : $1,078,410	

Adjust 4% - 2% (6 years) - 4%    : $1,080,000	100.15%	

Adjust 5% - 2.5% (10 years) - 4% : $1,130,000	104.78%

​

To each their own, but I would not care to make a long term cut like that in my early years. And the long-term benefit is pretty small. Others might want to play with cuts at different points, and see what they get. I also did not attempt to find the low point in the portfolio in each case. I suspect that even with these drastic cuts, one would see a drastically reduced portfolio.

OK, here's one - FIRECalc: 5% init, adj - 5 years - that takes the 5% initial and shortens to 5 years to give that ending portfolio value - it's down to less than half - $461,564/$1,000,000 - a 54% dip!

Oh, what the heck, for the others 4% - 2% -4% - FIRECalc: 4% init, adjust - 5 years

$491,291/$1,000,000 - a 51% dip.

and straight 3.5947% - FIRECalc: 3.5947% NO adjust - 5 years

$448,935 - a little worse actually, but all in the same scary neighborhood.

So when people say they would obviously cut their spending if their portfolio tanked, would they really cut it in half, for 6 years, in their prime retirement years?

Or something else? Please share any scenario that gets you to 100% - maybe it can be done with better times or more shallow cuts to spending. But then we may be getting close to data-mining - what would a generalized rule set be?

FireCalc already models this pretty closely with the "work Less, Live More" rule in the spending tab. You can adjust the % of subsequent year's spending with this tab. It's not exactly Guyton/Klinger but, it's probably close.

The problem I have with that is when the portfolio grows, you spend that much more, regardless of inflation. I think most people spend based on a combo of needs and resources, and don't just spend because they have it. They probably never would have attained FIRE status if they did that.

-ERD50

 

[Independent]

Thanks, lots of good work there. I hadn't thought about that approach.

I'm going to paraphrase and see if I have this right. If I have a $1 million portfolio, a traditional level $$ SWR lets me spend $36k every year.

If I want to move the early years up to $40k, I need to be willing to cut to $20k for 6 years in some scenarios just to make up my extra early withdrawals.

If I want to move the early years up to $50k, I need to be willing to cut to $25k for 10 years in some scenarios to make up for the early withdrawals.

Of course, $20k and $25k are significantly below the level $36k.

 

[urn2bfree]

Thanks, lots of good work there. I hadn't thought about that approach.

I'm going to paraphrase and see if I have this right. If I have a $1 million portfolio, a traditional level $$ SWR lets me spend $36k every year.

If I want to move the early years up to $40k, I need to be willing to cut to $20k for 6 years in some scenarios just to make up my extra early withdrawals.

If I want to move the early years up to $50k, I need to be willing to cut to $25k for 10 years in some scenarios to make up for the early withdrawals.

Of course, $20k and $25k are significantly below the level $36k.

This is not how I understand these papers. If you start taking $50000 and adjust for inflation, if the value of the portfolio drops by some you either don't take the inflation adjustment that year and if it drops by 20% or more you would cut by 10% or in the case of $50,000 down to $45,000. That $45,000 becomes your new base to adjust from Say the following year you lost another 20%- then you would cut $45000 down to $40,500 or by 10%. This is now the base. If the portfolio does not lose from there, you increase $40,500 by an inflation amount. And if you have a big gain in the portfolio, you could increase by 10% of that $40,500 base. I do not know where or why one would need to drop down to 2%WR according to these dynamic WR strategies.

 

[Huston55]

This is not how I understand these papers. If you start taking $50000 and adjust for inflation, if the value of the portfolio drops by some you either don't take the inflation adjustment that year and if it drops by 20% or more you would cut by 10% or in the case of $50,000 down to $45,000. That $45,000 becomes your new base to adjust from Say the following year you lost another 20%- then you would cut $45000 down to $40,500 or by 10%. This is now the base. If the portfolio does not lose from there, you increase $40,500 by an inflation amount. And if you have a big gain in the portfolio, you could increase by 10% of that $40,500 base. I do not know where or why one would need to drop down to 2%WR according to these dynamic WR strategies.

This is my understanding as well. I also seem to remember that, based on the research, only two such 10% spending reductions were required during a 30 (40?) year period.

I could live with that.

 

[ERD50]

This is not how I understand these papers. If you start taking $50000 and adjust for inflation, if the value of the portfolio drops by some you either don't take the inflation adjustment that year and if it drops by 20% or more you would cut by 10% or in the case of $50,000 down to $45,000. That $45,000 becomes your new base to adjust from Say the following year you lost another 20%- then you would cut $45000 down to $40,500 or by 10%. This is now the base. If the portfolio does not lose from there, you increase $40,500 by an inflation amount. And if you have a big gain in the portfolio, you could increase by 10% of that $40,500 base. I do not know where or why one would need to drop down to 2%WR according to these dynamic WR strategies.

Those dynamic level of adjustments aren't easy (or possible?) to do in FIRECALC, so that was a rough cut stab I took at it, with the available tools. So while other approaches might work better, cuts of those magnitude and length did get back to 100%. Reducing the length or depth of the cuts resulted in failures.

I have not gone through the paper yet - do they indicate just how low your spending would go, and how far a portfolio would dip when applying this algorithm to some of the worst 30 year periods since 1871? Also, my runs were for 100% historical success - that's my preferred approach, since I figure the future could certainly be worst than the worst of the past, I ought to at least plan for what has been observed. And it makes this test easier - cut spending until you get 100%. If they were basing on 95%, then they can get by with lesser cuts - but what about those failures?

There are some bad cycles there, I would think that you would see 6-7 down years that could have you making that 10% drop, and you would be ~ half your original spending (if I read that right). I'll try to dig into the papers later.

-ERD50

 

[urn2bfree]

Those dynamic level of adjustments aren't easy (or possible?) to do in FIRECALC, so that was a rough cut stab I took at it, with the available tools. So while other approaches might work better, cuts of those magnitude and length did get back to 100%. Reducing the length or depth of the cuts resulted in failures.

I have not gone through the paper yet - do they indicate just how low your spending would go, and how far a portfolio would dip when applying this algorithm to some of the worst 30 year periods since 1871? Also, my runs were for 100% historical success - that's my preferred approach, since I figure the future could certainly be worst than the worst of the past, I ought to at least plan for what has been observed. And it makes this test easier - cut spending until you get 100%. If they were basing on 95%, then they can get by with lesser cuts - but what about those failures?

There are some bad cycles there, I would think that you would see 6-7 down years that could have you making that 10% drop, and you would be ~ half your original spending (if I read that right). I'll try to dig into the papers later.

-ERD50

I do not think there has yet been a sustained drop of 20% per year 6 or 7 years in a row. Flat markets can go for a while, sure. A few bad years in a row, sure, but sustained hemorrhage on the scale needed to cut every year?? After a few bad years the market rebounds even if only briefly even after 1929 Crash.

 

[ERD50]

This is my understanding as well. I also seem to remember that, based on the research, only two such 10% spending reductions were required during a 30 (40?) year period.

I could live with that.

Does that include 'cuts' from not keeping up with CPI? I think we saw consecutive years of double-digit inflation in the 80's. A few years of that is way more than a 10% cut in buying power.

-ERD50

 

[ERD50]

I do not think there has yet been a sustained drop of 20% per year 6 or 7 years in a row. Flat markets can go for a while, sure. A few bad years in a row, sure, but sustained hemorrhage on the scale needed to cut every year?? After a few bad years the market rebounds even if only briefly even after 1929 Crash.

OK, I didn't read that as 20% per year, I was thinking you cut until it recovered to above 80% of original portfolio. But if you skip the CPI adjustment on a down year, seems those sure could add up (subtract down?) quickly.

I'll try to get through the paper tomorrow. Thanks.


-ERD50

 

[explanade]

Is there a spread sheet somewhere which will generate yearly balances you should have given a beginning balance, a withdraw rate, inflation and market return?

So that you can benchmark how actual balances are 5 years out, 7 years out, etc. compared to the modeled balances based on your original assumptions?

 

[obgyn65]

Like you I am also quite conservative. My planned SWR is about 3%, maybe a bit less if include the use of deferred annuities.

Curious to see what others think of the article. I tend to agree with it, even though I'm more on the conservative side myself.

 

[rayvt]

But if you skip the CPI adjustment on a down year, seems those sure could add up (subtract down?) quickly.

I'll try to get through the paper tomorrow. Thanks.
-ERD50

I think you'll find that under historical circumstances where the Guyton-Klinger rules reduce your withdrawal drastically, that a constant SWR portfolio would be absolutely crushed in that same period.

Realise that G-K depends on the market having strings of good periods, where the cuts in the extended bad periods would be made up. Not only is this an accurate view of the market, but **any** withdrawal strategy depends on having good periods on net.

If the market goes down for an extended period of time, then any withdrawal strategy of more than $0 will fail.

The G-K rules automatically & mechanically adjust your withdrawals to preserve your portfolio. Fixed SWR doesn't do that. It maintains constant withdrawals all the way down to zero.

============
I imagine it would be possible to create an excel spreadsheet containing the historical S&P500 (or portfolio of your choice or whatever) returns and then another few columns implementing the G-K rules and see what would happen under various scenarios. I've done a spreadsheet for my own portfolio this way, to guide my own withdrawals.

It's not hard to do, but I'm not sure it would tell you what you want. To do Monte Carlo or something is a lot of work. You really need to do it in a program, like Firecalc is. It would probably be relatively easy to add to Firecalc, for somebody who had access to the source code.

 

[Major Tom]

Oh man, I've read about your low rent and living expenses. Forget about RV'ing!

RV'ing is supposed to be cheap, but if you want to move about, gas cost is high. And when your vehicle breaks down, it hurts like the Dickens in the pocketbook.

You should stay put, my friend. It's better to live vicariously through the blogs of the other guys. Or wait till you get SS, and indulge yourself then.

I hesitated to reply, not wanting to get too off-topic, but your assessment of my financial situation is accurate NW-Bound. I have achieved such low living expenses that a move to even a modest RV would involve more bills, and less easily predictable expenses.

I plan to stay put for a few more years and take advantage of this simple and very affordable life before considering the move to an RV. Unfortunately, my current lifestyle couldn't get too much simpler or cheaper, so any changes I make in the future will most likely involve greater expense and complexity

 

[Independent]

This is not how I understand these papers. If you start taking $50000 and adjust for inflation, if the value of the portfolio drops by some you either don't take the inflation adjustment that year and if it drops by 20% or more you would cut by 10% or in the case of $50,000 down to $45,000. That $45,000 becomes your new base to adjust from Say the following year you lost another 20%- then you would cut $45000 down to $40,500 or by 10%. This is now the base. If the portfolio does not lose from there, you increase $40,500 by an inflation amount. And if you have a big gain in the portfolio, you could increase by 10% of that $40,500 base. I do not know where or why one would need to drop down to 2%WR according to these dynamic WR strategies.

The big reduction is because ERD50 wanted to recover all the excess payments in a short period, just 6 years. He didn't know exactly what the sequence of returns would be, so he didn't know how spread out the adjustment would be.

Yes, it seems like it would be a smaller adjustment for more years. Let's at least agree that there has to be an adjustment:

Suppose we are in a scenario where the traditional SWR of $40,000 just barely survives. There is a positive balance at the end of 30 years, but it's trivial.

Also, lets assume that the portfolio returns in the first two years were -9% and -21%, and that inflation in the first couple years was 5%.

Then the traditional case has me withdrawing $40,000, $42,000, and $44,100 in the first three years.

This alternate has me withdrawing $50,000, $50,000, and $45,000.

If I withdraw may entire amount at the beginning of each year, my fund balance just after my third SWR withdrawal is $612,864. Similarly, my fund balance at the same point in the alternate scenario is $598,455.

Since the 4% SWR approach just barely survived 30 years, the alternate approach has to pay less than the flat 4%, on average, over the next 27 years. It has to because we know that in this scenario the $612,864 is just barely adequate to provide the traditional SWR, so $598,455 has to pay less.

My question earlier in this thread, the one that ERD50 is trying to answer, is "How far below the traditional SWR will my withdrawals fall in the poor scenarios?" That seems to be an obvious question for someone who might embark on this route. The model that supports the papers contains the answer, but I don't see the author reporting it.

 

[siamond]

My question earlier in this thread, the one that ERD50 is trying to answer, is "How far below the traditional SWR will my withdrawals fall in the poor scenarios?" That seems to be an obvious question for someone who might embark on this route. The model that supports the papers contains the answer, but I don't see the author reporting it.

I agree with Rayvt, intuitively, the G-K model seems more likely to help you go through tough times than the classic fixed WR approach.

But I also agree with you, how low can this go is a key question. And for sure, the G-K model probably should be combined with some kind of floor, which corresponds to the very strict minimum of expenses you need to do, even when under duress. And maybe also an upper cap too.

That being said, I am quite convinced that under such extreme scenario, one would start playing with other parameters (e.g. downsize your house, move, try to find a part-time job, etc).

Now it would indeed be good to model when and how often such situation might arise for a given set of input parameters.

PS. all right, I opened a corresponding thread on the Firecalc section of the forum...

 

[NW-Bound]

I hesitated to reply, not wanting to get too off-topic...

All this talk about SWR is for people to find a way to maximize WR. Why? Because additional money may buy that little extra luxury or comfort. However, that additional money comes with a worry any time the market takes a downturn.

In your case, once you have settled in a stable living condition, it may be better to maintain it. The peace of mind is important too.

The last time I earned an income was just a bit more than a year ago so it may be too early to tell, but my Quicken expense tracking tells me that my expenses would be at 3.5%WR to maintain status quo. So, that's where I am trying to stay.

 

[ERD50]

OK, I started to read the paper, I don't have much time today but...

it appears everything is based on Monte Carlo!? Dang it. How the heck do I know how their choices compare with historical scenarios? While there isn't exactly any magic to history, the future may look way different, it is at least something with some grounding. And using tools with published data, we can compare. Does he provide all his Monte Carlo inputs? Did he have a scenario as bad as 1973? Worse? We know that the sequence of returns is crucial - how do these algorithms work with real, historical cycles?

The algorithm may have value, but I'd sure like to see how it handled the 1973 retiree.


-ERD50

 

[jkern]

I never understood this.

Person A retires with $1M and uses a 4% SWR. The initial withdrawal of $40K/yr provides Person A with a comfortable but far from extravagant retirement lifestyle. Person A prudently spends down the portfolio and has little left at death. Person A is complemented for living a well-planned financially responsible life.

Person B retires with $10M and uses a 0.4% withdrawal rate. At an initial withdrawal of $40K/yr, Person B has an identical retirement lifestyle as Person A. Due to investment growth, Person B has a $50M portfolio at death. These funds are given to family and/or charity. Person B is chastised for being the "richest person in the graveyard," accused of being "stingy and cheap," and is said to have led a life of "deprivation." People constantly told Person B, "you would be happier if you spent more money because you can't take it with you."

Both individuals led identical retirement lifestyles. However, Person A is complimented for living a productive happy life and Person B is criticized for being an unhappy miser. Personally, I'd much rather be Person B than Person A. I do not need to spend all or even most of my money to be happy. More so, the financial security provided by "excess money" would provide significant emotional value.

I agree completely!

 

[rayvt]

The algorithm may have value, but I'd sure like to see how it handled the 1973 retiree.

One of the plusses with financial debates about strategies, such as Covered Calls and Indexed Universal Life, is that once you create a spreadsheet that models a strategy using actual historical data, it's pretty easy to revise it to create a spreadsheet for a different strategy.

Since I already have developed several, and since I already have a spreadsheet that I use to guide my own portfolio withdrawals, it only took a couple of hours to create a spreadsheet to backtest Guyton-Klinger rules using the S&P500 historical returns.

It also shows a side-by-side comparison of the standard SWR withdrawal rules.

We don't care about the good times -- what we all worry about is what happens in the bad times. So 1973 is a good year to start the backtest from.

It takes a few parameters. The SWR for the G-K rules, the SWR for standard rules, and the cap for the CPI adjustment.

Download link:
https://www.dropbox.com/s/cwprtn6y8ouyajj/SPY_Withdraw_by_Guyton_rules.xls

I'll eschew editorial comment right now. Other than to note that it was interesting to see the action of the Capital Preservation Rule and the Prosperity Rule.

Dowload it. Play with the parameters as you desire.

I welcome any comments about errors that may be in the sheet.

 

[Fred123]

OK, I started to read the paper, I don't have much time today but...

it appears everything is based on Monte Carlo!? Dang it. How the heck do I know how their choices compare with historical scenarios? While there isn't exactly any magic to history, the future may look way different, it is at least something with some grounding. And using tools with published data, we can compare. Does he provide all his Monte Carlo inputs? Did he have a scenario as bad as 1973? Worse? We know that the sequence of returns is crucial - how do these algorithms work with real, historical cycles?

The algorithm may have value, but I'd sure like to see how it handled the 1973 retiree.


-ERD50

Just wondering -- what's the issue with Monte Carlo sampling? My understanding (not a statistics expert) is that as long as the probability distribution used for sampling correctly fits the historical data, then it's probably a better way of simulating returns than just using strictly historical data.

A bigger concern for me is that neither paper has been updated to include the 2008 bust.