3-PEAT Retirement Withdrawal Strategy

[audreyh1]

I thought VPW meant variable percentage withdrawal whereas %remaining portfolio is a fixed percentage withdrawal method, so I'm still a bit confused.

Oh, OK - I see it's simply increasing the % withdrawal rate each year based on age, and determined by your AA. Otherwise it's calculated off the remaining portfolio. https://www.bogleheads.org/wiki/Variable_percentage_withdrawal

I would mention to folks that this does not reduce income volatility in a noticeable way. Yes, during a multi-year down streak your percent withdrawn may sneak up a bit, but it is in such small percent increments, that it will easily be overwhelmed by bear market drops and/or rapid inflation rises.

Clyatt's rule has a similar "feature" as it is not inflation adjusted, so even though you are never withdrawing income less than 5% of the prior year, that's in nominal terms, and when a run like 1966 is compared against the %remaining portfolio method the nominal income withdrawn each year is very close. Meanwhile inflation is eating your lunch for a while before you catch back up. Short bear markets without significant inflation (like in the 2000s) may be relatively less painful using the Clyatt rule rather than straight % of remaining portfolio.

 

[Lsbcal]

I thought VPW meant variable percentage withdrawal whereas %remaining portfolio is a fixed percentage withdrawal method, so I'm still a bit confused.

I guess I should stick to just talking about VPW methods since that's what I'm using. What I do with the excess not spent in any year is to move it to something safe as suggested by you and Longinvest. Short term IG bonds for the unspent monies.

Sorry for the muddle.

 

[donheff]

Having looked at the paper I understand what is being proposed. It is similar to a fixed percentage approach but rather than an arbitrary percent I-ORP is consulted to choose an optional percent. Since this is recalculated each year that optional percentage will gradually increase as you get older and have a shorter life expectancy.

Like several others I dump excess into a fund to enhance down year spending. Unlike others mine is a pseudo fund within my overal portfolio thus the fund goes up and down with market performance.

 

[sengsational]

i-orp and FIRECalc walk into a bar...

For the i-orp nay-sayers, it's always been that i-orp calculates the entire plan based on a fixed return percentages; you set equities and 6% and bonds at 3% and it's smooth sailin'! Well, I agree with the nay-sayers that approach is incomplete since it ignores the sequence of return aspect.

That's not to say that poio (plain old i-orp) doesn't serve a purpose. You can fiddle around with Roth conversions and get an idea about the magnitude of the savings, for instance. It also has a monte-carlo option, but some people are only going to be happy with an historical aproach, a la FIRECalc.

Ok, so using poio suffers from not using historical variability. Now we see 3-PEAT, which, as the author mentioned, is modeling what i-orp users, myself included, have been doing for years: coming back every year and running the optimization. It probably helps them decide on various actionable choices (ACA, Roth conversions, etc), as well as giving them an optimized spending number.

The purpose of 3-PEAT seems to be to show how well this method will work compared to other withdrawal strategies. Run a complete i-orp 30 year model with an expected average return of 6% on equities, and the only thing you take from that is the recommended spending for the very first year of the model. You apply a historical return to your starting-year balance (a la FIRECalc), and run a 29 year model, again, taking only the recommended spending from the very first year of that model. Repeat ad nausium.

First, let me commend the author on keeping i-orp running and being prolific in the field. I hope everyone that uses i-orp donates at least a little bit.

I need to challenge the claim that a 3-PEAT plan "doesn't fail". If the scenario is such that I can keep my house as long as my budget doesn't get cut by more than 30%, and the 3-PEAT model has me cut by 40%, well, that's a plan failure IMHO. Maybe not a catastrophic failure, but a failure. The paper says the retiree is not ready to retire. Going back to the purpose of the paper, I realize this would be out of scope, but an improvement would to prevent the model from offering a spend amount below a non-discretionary amount. Of course that would mean that some models would fail (as in, run completely out of money). Also, this goes against the grain of the idea behind the optimization process that's the heart of i-orp.

Another improvement I can envision is some kind of axe/scalpel metric. If one buys into the i-orp idea that we might as well spend as much as we can within the safety of not running out of money, one must also ask "how bumpy was the ride"? It would be nice to have some kind of a bumpy ride metric that goes beyond how little was left at plan end. In a perfect world, we've always joked, the account would be zero, but how variable was the ride? Those two, together (not just the lowest ending balance) would need to be in the metric. Of course, the construction of such a metric would be the source of endless discourse, but I think would be useful.

Beyond a bumpy ride metric, and yet another possible improvement to the 3-PEAT process would be to include a 'smart' spending reduction rule. The idea I've come-up with would be to use the historical PE10 (CAPE) to clamp-down on spending. For instance, say we are iterating in the 3-PEAT process and the CAPE for the historical year we're emulating is above the historical average. That fact could be used to calculate some spending limitation and entered into i-orp's "Maximum Spending" field. Given the predictive value of CAPE, this might smooth spending by conserving money when the market is high, making it available for market lows. This seems to be an improvement on limiting spending as a fraction of the previous year's spending. There is probably no end to ideas for experiments with regards to rules that would limit spending in certain circumstances, but this CAPE idea might be worth exploring.

I have no problem with the chunks of history that were selected in the paper (Modern Era: 1985-2015, Bengen’s Big Bang: 1973-2003, Stagflation: 1966-1996, Interbellum: 1915-1945, Worst Case: 1903-1933, Gilded Age: 1880-1910). If you think about it, the wild majority of starting dates that are used in FIRECalc are completely uninteresting; it would be a waste of time to explore every one. I'm personally more concerned about needing more than a 30 year horizon, but the paper is about exploring a concept, and I'm confident the conclusions will hold up in alternate plan durations.

One might envision a process that automates the 3-PEAT process, possibly with customizable spend-limiting rules and plan comparison metrics. That would be a whole lotta calculations, but "computers is cheap". I wonder how much it costs to keep those i-orp servers running.

 

[haha]

All this emphasis on technical aspects of money withdrawal in retirement is just a cult pastime. The "insights"are nothing more than curve fitting and data mining. A % of current portfolio taken each year trades yearly spending fluctuations of allowed spending for a much smaller chance of bombing out. I happen to think that this is the only rational approach. If one wants to also smooth annual spend. he or she must give up some of the failure-proof features of an annual % of current portfolio. Depending on assumptions and the data marshaled to support this or that plan, some paln will look better or worse than others.

I personally cut back hard under stress, and for me that has been possible. It may not always be, or it may never be for some retirees

To take an adequate draw to maintain a jobless life depends on a) do you have a pension? With a good (govt) pension, history has shown that one really does not need much else. 2) if no pension do you have enough capital? How much capital is dependent on lots of things, but almost never on which sensible withdrawal scheme you choose. There really cannot be any magic here. Spend your effort on collecting pensions, on making more money while working, on socking away more money while working, and on investing this money profitably and sensibly. When retired, give thought to controlling your expenses in ways that have fewer negative effects on your happiness 3) Do you feel lucky? IMO, we have all been quite lucky since 2009. Equities have done way better than the underlying businesses, and bonds have not yet fallen back to levels that normal interest rates would require. This is likely more luck than we can count on always having.

Last point- a withdrawal rate higher than the cash generation of the underlying securities is clearly going to be time limited. This is the old "don't eat the seed corn" argument. This cannot be wrong over time.

Ha

 

[sengsational]

One might envision a process that automates the 3-PEAT process, possibly with customizable spend-limiting rules and plan comparison metrics.

Have you ever wondered what your annually adjusted retirement spending would look like if we had another market that looked like 1999/2008/2009? I did, and now I've got a pretty good idea what the answer to that question would be.

There is a 3-PEAT automation app now in beta testing to help answer questions like that. The app is not available searching through Google Play (yet?), but if you go to the opt-in link for the app, you can get the app. I just pushed "publish", and the play store takes a while to update, so if it doesn't work, try again in an hour or two.

Although the project started-out as just something I wanted to do for myself, I put a few finishing touches on it to allow fellow retirement plan tinkerers that have an Android device to try it. Not every single feature of i-orp is implemented, but I think most things will work. There are a bunch of input fields that map one to one with i-orp.

If you want to tinker there's a field at the bottom where you can put values of fields that I didn't include in my UI. So, for instance, if you wanted to include your house, you'd put "illiqudh=250, uprinblh=100" if your house was worth $250K and you had a $100K loan. You can also get those field names if you go to the bottom of a regular i-orp run.

When you first start out, rather than use a long model duration (taking you to age 95, or whatever), you should probably start with a short duration of a few years, to save yourself some time, since each model run takes a second or two. If your tired of waiting or see something wrong, you can hit the "red X" to stop the process.

As the 3-PEAT process progresses, each first-year model run result will be added to the results listing on the device. I tried to put something up that described how the model results were progressing, but really, the true output is not what you see on the phone, it's what you email to yourself when the process is done. You press the email button and mail yourself a spreadsheet (csv file, really, so it's just text that can go in any spreadsheet program).

Now you can look at how, with your asset allocation, your available "DI" spend number jumps around. As with i-orp, each subsequent year's values go up with inflation. But the spreadsheet includes a cumulative inflation value, so you can always divide by (1+cumInflation) to get constant dollars.

There is a message when all model year runs are complete that attempts to tell you how bumpy the road was (number of years where budget cuts were more than 5%). Not the greatest metric in the world, but something to compare plans with each other. Like I said, the real results will be when you tinker around with the spreadsheet output.

There is one "setting" and that was kind of an experiment with using PE10/CAPE ratio to limit spending when the market was "overheated". It uses an arbitrary formula that reduces spending. The inputs include the PE10 of the year of the run and the number of years left in the model. Earlier years in the model, the spend-limiting is "stronger". In the middle it gets weaker, and at the end, there is no spend limiting. If the PE10 is below it's historical average, there is no spend limiting. This was supposed to address the concern of the guy deciding to retire when the market was at all-time highs, only to be living on a heck of a lot less when the market "reverts to mean". Anyway, if you do a model run starting in, say, 1999 or something, it's going to give you a very small budget compared to what you "could" spend in an unlimited spending scenario. But the default is that spend-limit is "off", and you needn't deal with it unless curious.

James Welch and I have corresponded, and he's ok with my app banging on his web UI, but if you have a question about this app, it's going to be me to ask. If one optimization run (one row in the spreadsheet) has you wondering, then just use the parameters provided in the spreadsheet to do a single, regular web run. That will get one of those GUIDs and that will enable you to talk specifics with James. I looked for a donate button on i-orp, but it appears to be gone now. The app, of course, is free.

I suppose this thread is as good as any to talk through questions, problems, etc.

 

[Chuckanut]

I'd be interested in how the 3-PEAT method improves on VPW if anyone knows. I confess to not having read the working paper above. But I have used I-orp in my early retirement days. It helped me to see that tapping into my Roth was a good strategy until RMD's kick in.

Why do you say tapping into your Roth is a good strategy until RMD's kick in? Would not it be better to tap into the Reg IRA and use the Roths when one reaches the age of RMD? Would that not minimize taxes?

 

[growing_older]

It would take more modelling to see if this is so, but I would be inclined to use the non-Roth assets first, which when the expected market recovery eventually happens would have the effect of boosting my relative ratio of Roth to Trad IRA. This in turn puts me in a more favorable position in the future to reduce RMD and taxes that would trigger. The money is the same, but the location (account type) could really matter.

 

[Whisper66]

........I looked for a donate button on i-orp, but it appears to be gone now. The app, of course, is free.....

If interested in donating, I found a button at the bottom of the "About" page at https://i-orp.com/about.html.

 

[respond2u]

Why do you say tapping into your Roth is a good strategy until RMD's kick in? Would not it be better to tap into the Reg IRA and use the Roths when one reaches the age of RMD? Would that not minimize taxes?

That's a good question. It seems that a right thing to do would be to drain the tIRA so that the RMD (starting at 3.65%) doesn't cause a tax bracket change.

With differing social security, pensions, etc., I'm not sure if there's anything really general that can be said, is there?

Of course, all these projections assume that future tax code changes won't be malignant to retirees. GenX + Millenials will outvote Boomers by the mid-2020s and they may get tired of the favorable tax deals our parents worked out for us.

 

[Cut-Throat]

Last point- a withdrawal rate higher than the cash generation of the underlying securities is clearly going to be time limited. This is the old "don't eat the seed corn" argument. This cannot be wrong over time.

Ha

A lot of people don't realize that they are also "time limited".

 

[Whisper66]

Why do you say tapping into your Roth is a good strategy until RMD's kick in? Would not it be better to tap into the Reg IRA and use the Roths when one reaches the age of RMD? Would that not minimize taxes?

ORP provides a plan that optimizes taxes over your lifetime. The impact is to maximize money available for personal use. In some high $$ asset cases, ORP suggests it makes sense for withdrawals before 70.5 to include some Roth assets on top of tIRA funds.

Can see this by running some ORP cases where one has a very large tIRA fund. If the yearly withdrawal for living expenses is all from the tIRA, the taxes in early years can be so high that the goal of minimizing life time taxes is not met. It's cases like this that ORP suggests plans that may include Roth conversions and/or early Roth withdrawals.

 

[sengsational]

If interested in donating, I found a button at the bottom of the "About" page at https://i-orp.com/about.html.

Thanks for finding that. There used to be donate buttons on the form input pages. And my usual trick, to search a site with google "donate site:i-orp.com" didn't turn-up anything except for a comment.

Anyway, I just pressed 'publish' on the next version of the app that has a new "donate" action which takes you to PayPal for orplanner@gmail.com. It also removes the permission to write to local storage (that was just for debugging and I had forgotten to remove it).

 

[haha]

A lot of people don't realize that they are also "time limited".

Oh Cut-Throat, now you tell me!

Ha

 

[Lsbcal]

Why do you say tapping into your Roth is a good strategy until RMD's kick in? Would not it be better to tap into the Reg IRA and use the Roths when one reaches the age of RMD? Would that not minimize taxes?

Probably depends on several factors. We do draw from my largish IRA first and then blend with some Roth money to manage taxes effeciently. No way to impact the IRA size enough to change the RMD taxes.

 

[sengsational]

Probably depends on several factors. We do draw from my largish IRA first and then blend with some Roth money to manage taxes effeciently. No way to impact the IRA size enough to change the RMD taxes.

That's what i-orp recommends for me (IRA money plus a little Roth). I plan to take the IRA (actually 401k) money out in December (due to mandatory withholding), and take Roth money "whenever".

 

[Montecfo]

All this emphasis on technical aspects of money withdrawal in retirement is just a cult pastime. The "insights"are nothing more than curve fitting and data mining. A % of current portfolio taken each year trades yearly spending fluctuations of allowed spending for a much smaller chance of bombing out. I happen to think that this is the only rational approach. If one wants to also smooth annual spend. he or she must give up some of the failure-proof features of an annual % of current portfolio. Depending on assumptions and the data marshaled to support this or that plan, some paln will look better or worse than others.

I personally cut back hard under stress, and for me that has been possible. It may not always be, or it may never be for some retirees

To take an adequate draw to maintain a jobless life depends on a) do you have a pension? With a good (govt) pension, history has shown that one really does not need much else. 2) if no pension do you have enough capital? How much capital is dependent on lots of things, but almost never on which sensible withdrawal scheme you choose. There really cannot be any magic here. Spend your effort on collecting pensions, on making more money while working, on socking away more money while working, and on investing this money profitably and sensibly. When retired, give thought to controlling your expenses in ways that have fewer negative effects on your happiness 3) Do you feel lucky? IMO, we have all been quite lucky since 2009. Equities have done way better than the underlying businesses, and bonds have not yet fallen back to levels that normal interest rates would require. This is likely more luck than we can count on always having.

Last point- a withdrawal rate higher than the cash generation of the underlying securities is clearly going to be time limited. This is the old "don't eat the seed corn" argument. This cannot be wrong over time.

Ha

Lots of excellent points.

I would tweak your seed corn statement to say

"Last point- a withdrawal rate higher than the total return of the underlying securities is clearly going to be time limited."

You can actually exceed cash generation, if you do not spend all of the increase.

 

[Whisper66]

....This is the old "don't eat the seed corn" argument. This cannot be wrong over time.

"Don't eat the seed corn" is a smart philosophy for those that
(a) are happy to work long enough to
(b) develop high enough assets to
(c) support low enough yearly spending and are
(d) happy to have money left over at their passing for beneficiaries or other causes.

I gave significant thought to that philosophy before retiring partly because i have kids that would benefit from the large amount that would be left over after my passing (item d above). I thought of working those extra years as basically just working for them. When I honestly looked at our numbers and the market expected returns, I decided that this philosphy would require too many of my remaining years to address a very small probability that our money would not outlast us. Thus began the exploring what HaHa calls the "cult past time" methods.

In our case, there is still a very high probability that the kids will still have much passed on to them even using "3-Peat" or similar methods. Chances for running out of funds before we pass is basically nil unless dramatic changes occur in the economy and we blindly (and stupidly) keep spending at maximum levels. Eventually that could lead to our maximum yearly available spending funds being less than non-discretionary spending and the plan would "fail". Not likely for us but something anyone considering "eating the corn seed" should consider.

Good discussion.

 

[haha]

Lots of excellent points.

I would tweak your seed corn statement to say

"Last point- a withdrawal rate higher than the total return of the underlying securities is clearly going to be time limited."

You can actually exceed cash generation, if you do not spend all of the increase.

I agree completely with this. I was referring to the internal free cash generated by the securities, not the payouts to holders. Total return is certainly a way to approach this, but total return can deviate markedly based on current emotions in the market.

Ha

 

[sengsational]

Concerning the 3-PEAT Android app, only 3 downloads to the beta release, and no feedback as of yet.

I released the app to the Google Play Store just now, so should be discoverable in the PlayStore in a few hours. That means no need to go to the beta link any more...just search for it and install it. I can't imagine it will generate a big following, but I'm ready to move on to writing another app now. You can email me from the app if you are wondering about something or can think of more features.