Ratchet Up

[Independent]

A recent thread brought up the old question "The 4% SWR is far too conservative in most scenarios. If my early results are good, when can I adjust up?"

The first thought is to ratchet up. After all, if 4% of my initial assets is "safe", why isn't 4% of my new, higher, assets also safe?

We shared opinions, but I thought I'd do some number research. FireCalc does not provide a ratchet alternative spending plan. I thought I could download a couple key scenario factors from FireCalc and build my own model that had this possibility.

Here's the result. First, the standard 4% SWR. I started with a $1 million portfolio and $40,000 of spending. This table has the results. All these numbers could be found in the regular FireCalc Excel output, but I reproduced them to check my model.

The four rows represent four quartiles of scenarios. They are sorted by the assets at the end of year 10. The first four columns are average withdrawals for the indicated years. The next two are average assets at the end of years 10 and 30. The last two are the number of scenarios that end below $0, and the number of scenarios where the assets get below $200k, but stay above $0.

Fixed…	Yr 1-5 ….	Yr 6-10 …..	Yr 11-20 ..	Yr 21-30 ..	EOY 10 ………	EOY 30 ………	< $0	< $200k
Q1 …	40,000	40,000	40,000	40,000	2,010,508	3,289,304	-	-
Q2	40,000	40,000	40,000	40,000	1,393,635	2,121,818	-	-
Q3	40,000	40,000	40,000	40,000	1,031,607	1,382,206	-	1
Q4	40,000	40,000	40,000	40,000	576,981	490,319	6	2

As we all know, there are 6 failures. They are all in the quatile of scenarios that start out poorly.

Next, I have a simple ratchet. The payout goes up to 4% of the last year end balance, whenever that results in an increase, and never goes down.

Ratchet…	Yr 1-5 ….	Yr 6-10 …..	Yr 11-20 ..	Yr 21-30 ..	EOY 10 ………	EOY 30 ………	< $0	< $200k
Q1	46,915	65,607	85,946	95,608	1,787,993	1,272,450	-	-
Q2	44,189	52,850	62,780	71,903	1,291,791	963,000	-	2
Q3	43,093	46,653	49,973	56,400	976,101	692,599	2	2
Q4	40,615	40,840	41,594	43,584	570,161	374,296	7	1

This introduces new failures. As you'd expect, they tend to come in years that are just prior to failure years in the simple 4% SWR. One or two good years cause a ratchet up, then you're in a failure year with no cushion.

But, of course, I get to spend more money in the good scenarios. I'm somewhat disappointed that much of the extra money comes in the later years. And, ending assets are lower, as expected. I leave less money on the table.


I tried to get rid of those extra failures by doing a bump up, without the ratchet. So this is 4% of the higher of the initial assets or the assets at the last year end. Unlike the ratchet, a bad year following a good year results in withdrawals that go back toward the initial level.

4/4…	Yr 1-5 ….	Yr 6-10 …..	Yr 11-20 ..	Yr 21-30 ..	EOY 10 ………	EOY 30 ………	< $0	< $200k
Q1	46,557	62,854	78,621	71,276	1,812,332	1,806,705	-	-
Q2	43,773	49,670	56,784	59,830	1,314,378	1,360,384	-	-
Q3	42,107	42,424	44,401	51,502	1,008,030	1,080,837	-	1
Q4	40,329	40,049	40,453	42,308	575,330	447,735	6	2

This was successful in getting rid of the extra failures. That's kind of surprising, I'll guess there are other sets of scenarios where it wouldn't do quite as well.

But, the payout is less than the ratchet. Net, I think the success/payout combination is better here than in the simple ratchet.



This brings up the common point that people with good downside flexibility can spend more in general. To illustrate that, I did a percent of last balance plan. Unlike FireCalc, my floor is a percent of the initial portfolio.

In this case I picked 6% of last year's assets, with a floor of 3.4% of the initial portfolio. This illustrates the not surprising fact that retirees with a lot of cushion in their beginning assets can be quite aggressive with early payouts.

6/3.4…	Yr 1-5 ….	Yr 6-10 …..	Yr 11-20 ..	Yr 21-30 ..	EOY 10 ………	EOY 30 ………	< $0	< $200k
Q1	66,334	81,321	86,978	64,132	1,471,048	961,188	-	-
Q2	61,942	64,367	62,624	53,594	1,070,334	716,025	-	1
Q3	57,571	52,166	46,799	47,823	845,534	619,930	-	5
Q4	50,506	38,753	38,610	40,476	538,151	452,880	3	4

I think the increase in payouts is pretty remarkable, given the good success rate. I'd say that percent of current portfolio strategies definitely dominate for people who can stomach the downside.

(I'll admit to data mining here. I backed into the 3.4% to get the success rate I wanted. Other scenarios might have other results.)

 

[pb4uski]

So are you saying that a fixed 4% WR resulted in a 94.9% success rate and a 4% ratchet (renewing in any year where 4% of beginning of year portfolio exceeds last years withdrawals plus inflation) resulted in a 92.3% success rate?

And with the ratchet strategy that, on average, your available spending is ~40% higher in years 11-20?

If so, for me 40% more spending in exchange for a 2.6% increase in failure seems like a reasonable tradeoff.

 

[ERD50]

I didn't go through the numbers in detail (yet), but I don't think the results are surprising.

If you start with some failures in the data-set (95% success, 5% failure), ratcheting up (or more simply, increasing any amount of spending at any time), could push a success scenario to a failure. And while it can be seen as data mining, starting with 100% success scenarios will still provide 100% success with a ratchet up approach.

The concept of ratchet up, but back down in a bad year is an interesting one. I meant to comment on one of the other threads that a ratchet up isn't carved in stone - I'd expect a retiree to bring it back down (maybe just not take the inflation adjustment for X years) if they saw storm clouds in the distance. I just never tried to apply numbers/algorithm to it.

Thanks, will probably check it out in more detail later.

-ERD50

 

[Perryinva]

Kitces already published a Ratching plan. It was not as aggressive as the OPs, because the aim was to maintain principle at the same non inflation adjusted value over a 30 year period. The objective included stomaching drops in income over protracted downturns. The upside was similar, typically much higher income for the same success with the usual sequence of ROR influencing it. If the first 15 years of retirement were all bull market, then even ending in a 10 year bear never had the income lower than a continuous 4% Trinity withdrawal would. Allowing for prince reduction increased the SWR. Starting with a 10 year bear, then a 10 year bull, would mean never passing rhe 4% amount until 20 years in and no more bear., ut never going undner the 4% either. So really no downside to the ratchet approach.

 

[Midpack]

Kitces already published a Ratching plan. It was not as aggressive as the OPs, because the aim was to maintain principle at the same non inflation adjusted value over a 30 year period...

https://www.kitces.com/blog/the-rat...l-rate-a-more-dominant-version-of-the-4-rule/

 

[Perryinva]

Yup. That’s the one.

 

[Independent]

Kitces already published a Ratching plan. It was not as aggressive as the OPs, because the aim was to maintain principle at the same non inflation adjusted value over a 30 year period. The objective included stomaching drops in income over protracted downturns. The upside was similar, typically much higher income for the same success with the usual sequence of ROR influencing it. If the first 15 years of retirement were all bull market, then even ending in a 10 year bear never had the income lower than a continuous 4% Trinity withdrawal would. Allowing for prince reduction increased the SWR. Starting with a 10 year bear, then a 10 year bull, would mean never passing rhe 4% amount until 20 years in and no more bear., ut never going undner the 4% either. So really no downside to the ratchet approach.

Yep

the “rule” might be that any time the account balance is up 50% over the original value, spending is increased by 10% (over and above any ongoing inflation adjustments), but such spending bumps can only occur once every 3 years at most

My first interest here was to put some FireCalc-consistent numbers on the very simple idea that always seems to puzzle people: "If Joe, retiring one year later than me can safely take 4% of his portfolio, why can't I safely take 4% of any portfolio I might have one year after I retire?"

Kitche's approach is much more complex, and doesn't answer the question above. It seems more conservative than necessary, but I haven't checked.
Certainly, we could generate an infinite number of rules that allow increases and will look good with back testing.

I lean toward "Percent of current, with floor based on original" as a simple approach that allows early increases, that's why I showed a couple examples in that category.

 

[Independent]

So are you saying that a fixed 4% WR resulted in a 94.9% success rate and a 4% ratchet (renewing in any year where 4% of beginning of year portfolio exceeds last years withdrawals plus inflation) resulted in a 92.3% success rate?

And with the ratchet strategy that, on average, your available spending is ~40% higher in years 11-20?

If so, for me 40% more spending in exchange for a 2.6% increase in failure seems like a reasonable tradeoff.

If you trust back testing, yes.

Though, as you can guess, I wouldn't ratchet, I'd be quick to give up that extra (in the first few years) if the market went back down. Note that the 4/4 example looks pretty good in the up paths, and survives a little better in the down paths.

Based on one of your earlier posts, I thought you might gravitate toward the 6/3.4 example.

 

[ERD50]

I skimmed some of it, but the Kitces plan just seems overly complex to me.

If you were OK starting at say 4%, why not something simple like if 3.5% of your current portfolio is greater than your normal inflation adjusted amount, take the new, higher 3.5% and inflation adjust that going forward (that means your inflation adjusted portfolio grew to 1.14x, why wait for 1.5x?). If the portfolio keeps growing, repeat at 3.0% (1.33x), 2.5% (1.6x) etc.

That's essentially the "retire again & again" idea, but a little more conservative each step. Or (what I'd likely do), just split the difference between the new % of portfolio and the current inflation adjusted amount?

If that amount scares you in a downturn, set an upper limit for your WR%, which might be 6-8% if you started with 4%.

-ERD50

 

[pb4uski]

Yes, the 50% hurdle in the article seemed senseless to me.

 

[Independent]

I skimmed some of it, but the Kitces plan just seems overly complex to me.

If you were OK starting at say 4%, why not something simple like if 3.5% of your current portfolio is greater than your normal inflation adjusted amount, take the new, higher 3.5% and inflation adjust that going forward (that means your inflation adjusted portfolio grew to 1.14x, why wait for 1.5x?). If the portfolio keeps growing, repeat at 3.0% (1.33x), 2.5% (1.6x) etc.

That's essentially the "retire again & again" idea, but a little more conservative each step. Or (what I'd likely do), just split the difference between the new % of portfolio and the current inflation adjusted amount?

If that amount scares you in a downturn, set an upper limit for your WR%, which might be 6-8% if you started with 4%.

-ERD50

Yes, or the simpler version using 3.5% indefinitely. Sure seems like it should work. I'd like to run some numbers to compare withdrawals to the other option, but probably won't have time today.

 

[Independent]

I finally got back to this. This is a plan that starts with the regular 4.0% SWR, but ratchets up to 3.5% of the prior year's balance if that is a higher payout. The "ratchet" means it doesn't go down if returns weaken.

R.4.0/3.5…	Yr 1-5 ….	Yr 6-10 …..	Yr 11-20 ..	Yr 21-30 ..	EOY 10 ………	EOY 30 ………	< $0	< $200k
Q1	43,254	58,846	79,589	90,500	1,868,042	1,642,711	-	-
Q2	41,390	47,426	57,616	67,856	1,345,223	1,268,433	-	-
Q3	41,058	42,597	45,385	52,051	1,012,044	1,006,041	-	1
Q4	40,088	40,147	40,488	42,223	576,040	448,041	6	2

It does everything we'd expect. It ..
.. does not decrease the number of failures because it has no impact on the sharply down scenarios.
.. does not increase the number of failures because the ratchet level of 3.5% is a 100% safe WR for FireCalc's database.
.. increases the payouts over the pure 4.0% SWR
.. but does not increase them as fast as the other variable options in the OP

In particular, compared to the "greater of 4% or 4%" strategy, it doesn't go up as fast in the early years but doesn't go back down in any of the later years. Again, as expected.

 

[big-papa]

Yep



My first interest here was to put some FireCalc-consistent numbers on the very simple idea that always seems to puzzle people: "If Joe, retiring one year later than me can safely take 4% of his portfolio, why can't I safely take 4% of any portfolio I might have one year after I retire?"

Kitche's approach is much more complex, and doesn't answer the question above. It seems more conservative than necessary, but I haven't checked.
Certainly, we could generate an infinite number of rules that allow increases and will look good with back testing.

I lean toward "Percent of current, with floor based on original" as a simple approach that allows early increases, that's why I showed a couple examples in that category.

reminds me of retire again (and again) posted by siamond on Mr. Money Moustache - Siamond is a frequent poster on Bogleheads, by the way:

https://forum.mrmoneymustache.com/investor-alley/new-withdrawal-method-retire-again-again/

 

[Independent]

reminds me of retire again (and again) posted by siamond on Mr. Money Moustache - Siamond is a frequent poster on Bogleheads, by the way:

https://forum.mrmoneymustache.com/investor-alley/new-withdrawal-method-retire-again-again/

Yep. This thread follows up that type of discussion on another thread. I just wanted to put some FireCalc numbers on the idea.

And, as expected, the 4% ratchet increases the probability of failure, but the 3.5% ratchet doesn't. This agrees with MM.

But, I'd say that people who are okay with fluctuations should consider "greater of x% of original balance or y% of current balance", without the ratchet (ie, allowing reversions to x). Those numbers seem better for this data.

 

[big-papa]

Yep. This thread follows up that type of discussion on another thread. I just wanted to put some FireCalc numbers on the idea.

And, as expected, the 4% ratchet increases the probability of failure, but the 3.5% ratchet doesn't. This agrees with MM.

But, I'd say that people who are okay with fluctuations should consider "greater of x% of original balance or y% of current balance", without the ratchet (ie, allowing reversions to x). Those numbers seem better for this data.

I am OK with fluctuations, but I prefer something closer to VPW (aka PMT method). The disagreements that arise on bogleheads stem from what to use as the expected returns in the PMT equation. The VPW author likes to simply use the weighted average of historical real stocks and bond returns (based on your AA). In Siegel's "the only spending rule article you'll ever need", they choose to update it yearly based on remaining life expectancy and current TIPs yield as an (in my opinion) overly conservative choice. The VPW author's position has been that if you want smoother withdrawals with VPW, use a more conservative AA, but there are those of us who want the potential upside of a higher stock allocation without roadbumps that are too severe.

Several of us instead use future looking expected real returns and update it each year. For future returns, you can try CAPE based, Tobin's Q, DDM, Philosophical Economics "Single Greatest Predictor of future stock market returns", or an average of annually reported future expected returns from several investment houses. A large number of choices available. Some are empirically derived curve fits and some are based more on models that use price, PE, etc. Unlike using a fixed expected return for your entire retirement period, which tends to have withdrawals that shoot to the moon in the last several years in a large number of backtested cases, using an updated expected return tends to smooth withdrawals a little bit more evenly all the way to the end. Not perfectly smooth like a 4% rule would be, but smoother than using a fixed expected return.

Cheers,
Big-Papa

 

[Independent]

I am OK with fluctuations, but I prefer something closer to VPW (aka PMT method). The disagreements that arise on bogleheads stem from what to use as the expected returns in the PMT equation. The VPW author likes to simply use the weighted average of historical real stocks and bond returns (based on your AA). In Siegel's "the only spending rule article you'll ever need", they choose to update it yearly based on remaining life expectancy and current TIPs yield as an (in my opinion) overly conservative choice. The VPW author's position has been that if you want smoother withdrawals with VPW, use a more conservative AA, but there are those of us who want the potential upside of a higher stock allocation without roadbumps that are too severe.

Several of us instead use future looking expected real returns and update it each year. For future returns, you can try CAPE based, Tobin's Q, DDM, Philosophical Economics "Single Greatest Predictor of future stock market returns", or an average of annually reported future expected returns from several investment houses. A large number of choices available. Some are empirically derived curve fits and some are based more on models that use price, PE, etc. Unlike using a fixed expected return for your entire retirement period, which tends to have withdrawals that shoot to the moon in the last several years in a large number of backtested cases, using an updated expected return tends to smooth withdrawals a little bit more evenly all the way to the end. Not perfectly smooth like a 4% rule would be, but smoother than using a fixed expected return.

Cheers,
Big-Papa

Yep, people who can handle quick downside adjustments should think about "percent of current portfolio" methods. We've had VPW threads here -- with some overlap on the individuals posting at Bogleheads.

The Trinity study SWR was aimed at retirees didn't want to make downside adjustments except in extreme cases.

Your second paragraph talks about various ways of trying to soften the downside. They may make VPW more appealing to more people. I don't know if you've discussed them in a thread here.

 

[big-papa]

By the way, you might want to check out Simba's spreadsheet over on bogleheads. Many more asset classes available (including intermediate treasuries, notably missing in Firecalc). It also has an SWR calculation built it. No ratcheting - understandably since there are just way too many ideas for withdrawal methods.

 

[Finance Dave]

A recent thread brought up the old question "The 4% SWR is far too conservative in most scenarios. If my early results are good, when can I adjust up?"

The first thought is to ratchet up. After all, if 4% of my initial assets is "safe", why isn't 4% of my new, higher, assets also safe?

Using this logic, you are assuming there will never be a downturn. The 4% rule assumes variability in returns.

 

[audreyh1]

I am OK with fluctuations, but I prefer something closer to VPW (aka PMT method). The disagreements that arise on bogleheads stem from what to use as the expected returns in the PMT equation. The VPW author likes to simply use the weighted average of historical real stocks and bond returns (based on your AA). In Siegel's "the only spending rule article you'll ever need", they choose to update it yearly based on remaining life expectancy and current TIPs yield as an (in my opinion) overly conservative choice. The VPW author's position has been that if you want smoother withdrawals with VPW, use a more conservative AA, but there are those of us who want the potential upside of a higher stock allocation without roadbumps that are too severe.

Several of us instead use future looking expected real returns and update it each year. For future returns, you can try CAPE based, Tobin's Q, DDM, Philosophical Economics "Single Greatest Predictor of future stock market returns", or an average of annually reported future expected returns from several investment houses. A large number of choices available. Some are empirically derived curve fits and some are based more on models that use price, PE, etc. Unlike using a fixed expected return for your entire retirement period, which tends to have withdrawals that shoot to the moon in the last several years in a large number of backtested cases, using an updated expected return tends to smooth withdrawals a little bit more evenly all the way to the end. Not perfectly smooth like a 4% rule would be, but smoother than using a fixed expected return.

Cheers,
Big-Papa

Interesting. Do you have links for VPW spreadsheets using future looking expected returns?

I’m currently using % remaining portfolio method and am OK with income varying.

 

[big-papa]

Interesting. Do you have links for VPW spreadsheets using future looking expected returns?

I’m currently using % remaining portfolio method and am OK with income varying.

No. Sorry. It's pretty straightforward: The Excel function for each year's % withdrawal is

PMT(expected returns, # years remaining, -1, 0, 1)

The regular VPW spreadsheet available on BH basically reinvented the PMT calculation and doesn't use the built in function in Excel. It is also based on a BH 2 or 3 fund portfolio so if you deviate from that, you'll need to create your own spreadsheet anyway. Returns of other assets are available on the Simba spreadsheet over at Bogleheads.

For US large cap stock, an OK estimate of expected real returns is 1/PE10
For bonds, just use the current interest rate and take off expected inflation.
Then weight the two based on your AA. If you're backtesting, you can try the previous year's inflation as expected inflation or maybe an average of the last 2 or 3 years. For year's they existed, the delta between nominal bonds and TIPs is another choice.

That should get you going. If people want to discuss this further or have other ideas, then a new thread should be opened.

 

[garyt]

I haven't read through your post or any follow up posts, but the obvious thing to me would be that you would eventually be setting yourself up for a bad sequence of returns . Eventually you're going to start your new year at the beginning of a big correction/bear market. The reason the 4% rule works is because of those first few good years youhad.

 

[Perryinva]

Actually, the success rate for the 4% SWR is based on the assumption of 30 years. There IS nothing wrong with a 4% assumption of the newer higher amount, as long as the 30 year assumption is still in place. One must also remember that FIRECALC uses returns based on returns meeting the market average plus assumptions. This is somewhat unrealistic when applied to the guy that is more Risk tolerant ie (bought heavy in to Netflix, etc, then got out with a 1000% increase) compared to the guy that only invests in Artistocratic Dividend stocks. The volatility of one is greater than the volatility of the other, so crossing the threshold of failure is more likely for one that the other. In reality, as Kitces does point out, astute investers never blindly follow a fixed plan. They are active money managers, and once a portfolio gets below a personal threshold, income cutbacks ensue. Since astute investors typically bave much more money and income, the cutbacks are not as sinister to their budget. There is no real way, to prevent too much safe i come in later years, if one does not reassess their minimum balance as they age. You don’t know when you will die so you keep adequate funds to generate income. You know you need a higher balance to begin retirement and a lower balance when finishing but no one knows when they are finishing until you hit say, 90, when you KNOW the likelihood of 5-10 more years is very low and can act accordingly. But who among us is certain they will have the same mental capacity at 90 that they do today and will make the changes?

Plus, like so many say here, what would you spend it on after a lifetime of LBYM.

Kitces Ratcheting plan is a stab at seeing if a formula similar to his has potential to work (seems it does) and he says it is not an attempt to define that his is correct. His just knocks the top too large ending balances down, while minimizing the reduced incomes for very long bear runs, hence the 50 %. The 50% value depends on how much you have. Certainly someone that goes from $400k to $600k is not in the same boat as someone that went from $2M to $3M.

 

[big-papa]

The 4% rule (should really be a guideline) came about by both the Trinity Study and Bengen's study at a time when people were throwing around numbers like 8% inflation adjusted withdrawals being sustainable.

4% comes about because the studies show that one could withdraw 4%, adjusted by inflation annually, and sustain a 30 year retirement, based on historical sequences of returns. The number 4% generally comes about due to the worst ever starting year for retirement - for most Asset Allocations that is 1966. Some allocations show a higher number than 4%, some less than 4% by the way. Any other starting year would have been able to sustain a higher than 4% withdrawal rate - problem is you wouldn't have known what market you were retiring into on year 1. And if you knew ahead of time that 4% was what you were going to start with, in the majority of cases you would have found yourself 6 feet under with a huge pile left over that you could have spent along the way. There have been some studies that indicate that the first few years returns during retirement will set the stage for whether you're going to be on the edge of running out or likely to have a pile left over.

Enter various variable withdrawal methods which attempt to trade off variability of withdrawals for a better chance of your portfolio lasting, with potential for upside if things are going well. They can be simple such as just taking a fixed % of your remaining portfolio - guaranteed to last forever. They can be more complex such as Guyton-Klinger decision rules. They can be PMT based such as ARVA and VPW. And there are others if you're so inclined to google. The possibilities are endless and if there was one perfect method, we'd all be using it.

Ratcheting, such as what siamond shows on Mr. Money Moustache (retire again and again) or the Kitces article are methods to increase your withdrawals if it looks like you're on a good trajectory while keeping things at the floor of 4% if it looks like a 1966 scenario. For some of these, remember, if you're assuming a 30 year retirement, the so called "safe" withdrawal is around 4%. If you think you'll live longer, then common sense (and firecalc) will tell you that it's less than 4%. If you think you'll have a shorter lifespan, likewise, the safe withdrawal would be more than 4%.

Then there is Taylor Larimore's method (he's the grand poo-bah over at Bogleheads)
Here's a quote he likes to give when discussions of withdrawal methods start:

"I retired in June of 1982 at the age of 57. We had about a $1 million dollar portfolio to last us the rest of our lives. I didn't know about safe withdrawal rates (the Trinity Study wasn't published until 1998). We had no computers, Internet, Monte Carlo, or sophisticated calculators. We only knew that we had to be careful to make our money last ($1M at 4% = $40,000/year before tax).

So what happened? We simply withdrew what we needed and kept an eye on our portfolio balance. Most years our balance went up and we spent the money on vacations, luxuries and charity. When our balance went down we tightened our belt and economized.

This is what most people do and it works."

Cheers,
Big-Papa

 

[Perryinva]

^^^ yup. Perfectly good reasoning IMHO. It also makes a big difference if the investments are needed to actually live on vs just supply a higher level of luxury and emergency support and inheritance.

 

[Independent]

Using this logic, you are assuming there will never be a downturn. The 4% rule assumes variability in returns.

I didn't say that was my opinion, but one that I had seen on other threads.

I wanted to find an example where a simple ratchet like that converts a surviving scenario into a failing one. FireCalc doesn't provide a simple ratchet as a withdrawal option, so I did my own calculation.