Total Portfolio Allocation and Withdrawal (TPAW)

[Ben Mathew]

Just curious what does implementing duration matching mean? Is the tool going to recommend an average duration for your bond holdings? Or something else?

The target duration of bonds will be shown in Tasks. But the goal is to also capture the impact of duration matching in the simulations. That will happen as part of a broader goal of modeling changing expected returns for both stocks and bonds. Asset allocation and withdrawals in the simulation would reflect these changing expected returns. There would be the option of duration matching bonds for the simulation, so the impact of duration matching can be seen in the spending graph.

[jic]

This is on the to-do list. This post talks about adding the ability to start an income stream at your or your partner's max age.

Thanks for the suggestion and for sharing your use case. It's always helpful to know how a feature will be used.

Thanks - I hadn’t seen the linked post but that’s exactly how I ended up working around. Just a little clunky at the moment as we wanted to look at scenarios, e.g. both survivor scenarios with 5, 10, 15 years difference.

FWIW the process of selecting risk preference by adjusting the slider while looking at monthly withdrawals really clicked with them. I’d never seen a tool with that workflow and think it was a clever design decision.

[Ben Mathew]

FWIW the process of selecting risk preference by adjusting the slider while looking at monthly withdrawals really clicked with them.

Thanks for sharing this. That's great to hear!

[NatureBoy]

I'm starting to put TPAW planner into action this month, the first month of my retirement, and have a really basic question. I've had the tool configured as though I was retired for the last few months, and the balance history shows that the withdrawal occurs on the 2nd to last day of the month, and the rebalance occurs on the first of the month. Is there a way to change what day of the month the withdrawal occurs? I was planning on taking my monthly withdrawal (as shown in tasks) the first day of the month, but that means the current portfolio balance won't match for most of the month. Is there something I'm missing?

[Ben Mathew]

the balance history shows that the withdrawal occurs on the 2nd to last day of the month, and the rebalance occurs on the first of the month. Is there a way to change what day of the month the withdrawal occurs? I was planning on taking my monthly withdrawal (as shown in tasks) the first day of the month, but that means the current portfolio balance won't match for most of the month. Is there something I'm missing?

There are two aspects to this: one is the assumptions for the simulation, and the other is the portfolio balance estimate.

The simulation is happening at the monthly level. So anytime during March 2025 is seen as the "current month" and the planning horizon is the same—say n months remaining. When we move into April, the current month becomes April 2025 and the planning horizon becomes n-1 and stays that way for all of April. Usually the planning horizon is large enough and monthly flows small enough that it doesn't matter very much whether whether we're amortizing across n months or n-1 months, and whether contributions/withdrawals are complete for the month.

The assumptions used for the portfolio balance affect the simulation only through the current portfolio balance, and can be over-ridden by manually entering the current portfolio balance. There is no option for specifying the day of the month when contributions, withdrawals and rebalancing are assumed to happen. But even if the portfolio balance is modeled accurately at the daily level, the simulation is still happening at the less granular monthly level. i.e. Even if you've taken out the withdrawal for March, the simulation will assume that you haven't and will continue to include March in the amortization. Given this, trying to make the portfolio balance estimate accurate is of limited benefit.

Eventually, we may try to get more accurate with the simulation and create a way for you to say when you're done with the contributions/withdrawals for the current month so we know to exclude it from the calculation. But the benefit of doing this is small, so it is pretty far down the list.

[idc]

I was playing with the risk tolerance, and I observed something I did not expect. I set no ceiling/floor and no legacy for my portfolio. I set out nominal amounts for social security and then I started to play with the risk tolerance slider. I assumed I can retire today.

The thing that surprised me the most was that the suggested monthly spending amount for the first withdrawal decreased as I increased the risk. This was rather unexpected, as I was expecting to be able to spend more earlier with an increased risk.

The way the model is structured, it suggests much higher monthly spending for very conservative portfolios than for very aggressive. Ben, can you comment on why that is?

I am trying to address that with the spending tilt, but unfortunately it is very hard to conceptualize how to correlate a certain spending tilt value to actual spending. Also, even when moving the tilt quite negative, making the risk more aggressive still makes the median spending increase in time, rather than decrease. Not saying there is something bad with the model, but rather trying to understand the logic behind it so I can use it better.

Some of my expectations when using this:
1. An increased risk would allow for more spending initially, with a wider spread for the outcome. I suspect that a higher risk making the spread more severe in later years forces the early withdrawals to be smaller. It might make sense, but it is unexpected.
2. When I favor more spending early on (negative tilt), I would have expected the median allowed monthly spending to decrease in time, regardless how I play with the risk. But for my numbers monthly spending decreases only for the more conservative risks and can increase quite significantly with more aggressive risk. Again, I was not expecting this.

I have more to add, especially as I add legacy into the discussion and about the volatility of the asset allocation but can you please comment on the logic? I am not saying it is wrong, I suspect it makes sense and could be a very interesting insight, just trying to understand it.

[Ben Mathew]

The thing that surprised me the most was that the suggested monthly spending amount for the first withdrawal decreased as I increased the risk. This was rather unexpected, as I was expecting to be able to spend more earlier with an increased risk.

When you increase risk tolerance, the stock allocation and so the expected return of the portfolio goes up. But the base spending tilt given by Merton's formula increases as well. This increase in the spending tilt is more than offsetting the increased expected return, reducing starting withdrawals.

The base spending tilt is increasing because a higher risk tolerance means that you are are getting a higher risk adjusted return by investing in the risky portfolio (because the certainty equivalent of a risky bet is higher.) The higher risk adjusted return makes future spending look more attractive, so you tilt your spending more to the future. A higher risk tolerance also implies a higher elasticity of intertemporal substitution (EIS) in the model, which means you don't need to smooth spending as much across time. This makes you more willing to shift your spending to the later years where you can have higher spending (due to compounding). In short, a high risk tolerance makes you closer to a risk neutral person who just wants to maximize their expected spending, which means taking on more risk but also scheduling more of your spending in later years.

If instead you want to take more risk but don't want to delay spending, you can reduce the spending tilt using the spending tilt slider.

Also, even when moving the tilt quite negative, making the risk more aggressive still makes the median spending increase in time, rather than decrease. [...]

When I favor more spending early on (negative tilt), I would have expected the median allowed monthly spending to decrease in time, regardless how I play with the risk. But for my numbers monthly spending decreases only for the more conservative risks and can increase quite significantly with more aggressive risk.

The spending tilt has two components: the base spending tilt and the extra spending tilt. The base spending tilt is the one given by Merton's formula. The extra spending tilt is what you enter on the slider. You can see the breakdown between the two by clicking on "Spending Tilt Breakdown" under the slider.

When you increase your risk tolerance, the base spending tilt increases and if it exceeds the negative extra spending tilt you entered, the total spending tilt will become positive. Just adjust the risk slider and the spending tilt slider till you arrive at the spending graph that you most prefer.

[idc]

Thank you for the detailed response Ben, I think I understand what you are saying.

I guess it is inevitable that with so many variables the interdependencies are so subtle. Not a criticism, I expect this complexity to limit the attractiveness of the model to broader audiences, at least until the complexity can be exposed in the UI in a way that is more intuitive.

But this is a learning process, I am assuming part of what you're doing is using this type of feedback to evolve the model and I hope I can help. I am still in awe of this tool and more than happy to be able to provide this type of input.

Thank you.

[Ben Mathew]

I am assuming part of what you're doing is using this type of feedback to evolve the model and I hope I can help. I am still in awe of this tool and more than happy to be able to provide this type of input.

Thanks. Feedback is always appreciated. It's useful for us to know what types of issues people are encountering when they use the planner. We can't always fix it, but sometimes we can improve how we handle something. The planner has benefited a lot from the feedback we have received so far!

[jmk]

I'm curious about TPAW's monte carlo. In the old excel sheet the population samples got adjusted proportionally to users' specification to match their expected returns, which were then drawn randomly. In the online rendition, the user is now able to directly specify the standard deviation as well as the expected average, which both seems to affect the monte carlo. What kind of bootstrapping is this?

[Ben Mathew]

I'm curious about TPAW's monte carlo. In the old excel sheet the population samples got adjusted proportionally to users' specification to match their expected returns, which were then drawn randomly. In the online rendition, the user is now able to directly specify the standard deviation as well as the expected average, which both seems to affect the monte carlo. What kind of bootstrapping is this?

You can do Monte Carlo simulations using any return distribution—historical distribution, normal distribution, log-normal distribution, logistic distribution, etc. The online planner takes the historical distribution and adjusts it to match the expected return and volatility entered by the user. I'm not aware of a special term for Monte Carlo simulations done with this (or any other) return distribution assumption.

[Raspberry-503]

I think this random sampling of historical data is a form of "bootstrapping"

[jmk]

I think this random sampling of historical data is a form of "bootstrapping"

Yah, i used the term "monte carlo" cause that’s the word tpaw uses. (To state the obvious, that tpaw samples blocks of historical data shows it’s bootstrapping.) Granted I’m nerding out in the weeds, but with my naive finance knowledge, outside of academic research I was not aware of any bootstrapping that scaled not only the historical returns but also the volatility of historical returns. Linear scaling of both measures with z scores would do the trick. Something like: Target return i=Target Return + Target SD *(Original Return i − Original Mean)/ Original SD) for each return year gives you a distribution with the same shape (fat tails, etc) as the historical record but with user ER and expected SD. So TPAW must do something like that.

Frankly, other than TPAW, I just haven't seen anything that sophisticated with other finance sites/apps. TPAW should get more credit for its simulation IMO! Programs like Bolin and Pralana or VPW use simple monte carlo (based on normal curves that minimize tail risk), or simple historical runs. TPAW bootstraps scaled historical block samples, which really has the "best of all worlds" and is "best in class": it maintains the autocorrelation and and covariance of the historical record, while still bending to expected future returns and variance. Bravo!

[Ben Mathew]

Linear scaling of both measures with z scores would do the trick. Something like: Return-scaled i=Expected Return + Expected SD *(Original Return i − Original Mean)/ Original SD) for each return year gives you a distribution with the same shape (fat tails, etc) as the historical record but with user ER and expected SD. So TPAW must do something like that.

Yes, this is what is being done.

Frankly, other than TPAW, I just haven't seen anything that sophisticated with other finance sites/apps. TPAW should get more credit for its simulation IMO! Programs like Bolin and Pralana or VPW use simple monte carlo (based on normal curves that minimize tail risk), or simple historical runs. TPAW bootstraps scaled historical block samples, which really has the "best of all worlds" and is "best in class": it maintains the autocorrelation and and covariance of the historical record, while still bending to expected future returns and variance. Bravo!

Thanks. Happy to hear that you appreciate these aspects of the simulation. I do think that this is a better way to model returns because, as you said, it combines the fat tails and correlations of the historical distribution with user inputs for expected return and volatility.

[idc]

Hey Ben, I played some more with the tool and studied intently the Merton's formula. I can't get past the problem of tying the risk aversion to something I can easily relate to. We all know that is a general problem.

Here is what I did instead, maybe this info will be helpful to you in thinking through the design of the tool. I am in the situation where I know relatively well how much I need to live a very comfortable life, taking into account all expenses, including very generous allowances for travel, health care, etc. I am very lucky that all tools I have used so far indicate that what I can afford to spend is more than that number by a good margin.

So instead of using TPAW to determine the max I can spend, I instead used it to model asset allocation and legacy based on the ceiling spend I am already comfortable with.

I kept all setting default, except for spending ceiling, which is a known quantity for me. I set that to a number that exceeds my max projected spend by 20% and added the social security income.

Then I proceeded to play with the risk tolerance slider until I arrived at a situation where my monthly spend could be kept constant at that ceiling, while the net worth would stay relatively constant. I then determined an asset allocation that will help me achieve those objectives. The top of the monthly spending line is the constant ceiling for me and the P5 is a curve that would still keep me very comfortable.

I am super happy with this result. For my situation the risk tolerance implies a ~70/30 allocation now with some slow decrease in time. This simply validates my plan.

I find this approach much more comfortable for my personal circumstances and I felt it eliminates the fuzzy feeling I would have on playing with spending tilts, trying to model my risk aversion in different ways, etc. Instead of trying to understand and quantify my risk aversion, I am now seeing what different asset allocations would do for my ceiling, P5 curve and legacy, which to me is a very powerfool tool to validate that I am ready for retirement and have a more quantifiable assessment on how things can play out.

It does require some degree of trust in the statistics and Merton's equation, but that is a lot easier for me to be comfortable than assessing my risk tolerance.

Obviously this is an exercise I would have to continue to do on a regular basis.

Let me know what you think of this approach and I hope this is something it can help you.

[jic]

So instead of using TPAW to determine the max I can spend, I instead used it to model asset allocation and legacy based on the ceiling spend I am already comfortable with.

I kept all setting default, except for spending ceiling, which is a known quantity for me. I set that to a number that exceeds my max projected spend by 20% and added the social security income.

Then I proceeded to play with the risk tolerance slider until I arrived at a situation where my monthly spend could be kept constant at that ceiling, while the net worth would stay relatively constant. I then determined an asset allocation that will help me achieve those objectives. The top of the monthly spending line is the constant ceiling for me and the P5 is a curve that would still keep me very comfortable.

Very interesting to read as I arrived at almost exactly the same procedure for my own planning - a generous ceiling, then adjust risk tolerance to consider median / p5 spending curve. In my case lower risk tolerance means hitting the flat ceiling almost immediately with a very low equity percentage, and what feels “right” to me is a lower monthly spend initially to keep a higher equity allocation, leading to improved p5 curve and higher legacy.

I found this quite intuitive and haven’t really played with spending tilt as I don’t have specific reasons to favour earlier or later spend. I guess this is situational and someone whose portfolio is a lower multiple of expenses would find the tilt more useful for exploring tradeoffs.

I also watched with interest as markets have dropped recently and TPAW has increased suggested equity allocation while the curves and median legacy remain roughly constant. I believe this is consistent with lifestyle investing theory on removing sequence of return risk, so I found it cool to see in realtime! Good encouragement to not get jitters and stay the course

[Ben Mathew]

Instead of trying to understand and quantify my risk aversion, I am now seeing what different asset allocations would do for my ceiling, P5 curve and legacy, which to me is a very powerfool tool to validate that I am ready for retirement and have a more quantifiable assessment on how things can play out.

This is using the planner in the way that it is intended to be used. I think you may just not be realizing that it is the same approach.

The planner does not require that you know your risk aversion or spending tilt. You just need to adjust the sliders till you arrive at the spending graph and legacy outcomes that you most prefer. From "How to Plan" under "Help Me Understand This" (located below the spending graph):

• How to Plan
  
  Spending During Retirement and Legacy are your primary views into your retirement. They will be at the higher end of the range shown on the cards if the markets do better than expected and at the lower end of the range if the markets do worse than expected.
  
  Spending During Retirement and Legacy together capture what ultimately matters financially for your retirement. Your decisions about retirement—how much to save, when to retire, how much risk to take, etc.—should be driven by their impact on these two outcomes. Other things like portfolio balance, asset allocation and withdrawal rates are just the tools used to control retirement spending and legacy outcomes. They are means to an end, not an end in itself. So don't focus on those numbers.
  
  To create your retirement plan, simply adjust your inputs till you arrive at the Spending During Retirement and Legacy ranges that you most prefer.
  

So what you are doing is consistent with that. Applying a ceiling is just a way to control the spending and legacy distributions. It makes sense to apply a ceiling if you don't care to consume more than the ceiling and would prefer to leave a larger legacy instead. As long as you are choosing asset allocation, spending tilt, ceiling, floor and other inputs based on their impact on the spending and legacy distributions, you are using the planner as intended.

Adjusting risk tolerance, spending tilt etc. till you arrive at the graph you like best might seem almost too simple and easy. But that is all there is to it. There is no deeper meaning to what risk tolerance is other than what kind of spending distribution you most prefer. So the best way to figure it out is to focus on the spending graph and look for the spending distribution that you like the best. More about that in this post:

• The best way to gauge your risk tolerance/RRA is by simply adjusting the risk tolerance slider till you arrive at the spending graph (and legacy outcomes) that you most prefer. There are tools like surveys and thought experiments to help you evaluate your RRA, but those are less reliable because you are not making the evaluation in the context of real world financial circumstances. They are based on artificial scenarios that may not translate well enough to your actual circumstance and don't provide the full context of the choices that you are facing. This is especially a problem if you have created a spending floor (by adding essential extra expenses). Your risk aversion should be evaluated in the context of that floor. If you have a floor that covers your essential expenses, you can take more risk on the risk portfolio. That would mean entering a lower RRA (higher risk tolerance) on the slider. A survey can miss this if you're not careful to account for the floor when answering questions. But the planner will catch it. Evaluating your risk tolerance inside the planner by adjusting the risk tolerance slider will avoid these types of errors.
  
  RRA simply measures how willing you are to take on risk. That risk is displayed in the spending graph. You don't need anything more. If you get an RRA from a survey, that is simply a guess about how you would adjust the risk slider when faced with these choices. You don't need that guess—you can just adjust the slider and find it directly!
  
  If you still want to get an RRA estimate from outside the planner, the Missing Billionaires survey (pages 175-180) and Professor Campbell's demon thought experiment that ConstantChrysalis referred to above are good sources. But make sure that you don't blindly enter the results in the planner. Instead, look at the spending graph in the planner and adjust the risk aversion slider till you find the spending distribution that you most prefer. If the RRA you get from the planner is different from the survey and thought experiment results, go with the RRA you found through the planner. It's effectively a richer survey administered in the perfect context—the financial choices that you actually face in real life.
  
  Note also that this approach of adjusting the inputs based on the outcomes shown in the spending graph is how the planner is meant to be used for all inputs, not just for risk tolerance. For example, the choices of how much to save, when to retire, etc, should be made based on how it’s affecting the spending graph. The spending graph is what captures the material impact of all of those decisions on retirement. This is why it's the default view of the planner and why we have taken care to make the graph visible when you enter any of the inputs. It always boils down to what's happening in the spending graph.

[Ben Mathew]

I also watched with interest as markets have dropped recently and TPAW has increased suggested equity allocation while the curves and median legacy remain roughly constant. I believe this is consistent with lifestyle investing theory on removing sequence of return risk, so I found it cool to see in realtime! Good encouragement to not get jitters and stay the course

Great to hear that this feature of the planner is working out well for you. Focusing on projected spending instead of the portfolio balance and recent returns helps put us in the right frame of mind and leads to better decisions.

[idc]

This is using the planner in the way that it is intended to be used. I think you may just not be realizing that it is the same approach.

Thank you for the detailed response, I guess I should have read the fine manual more closely. In fairness, it is all too tempting to try to exploit all those nice knobs you put in there to the max I also assume that some of those might be intimidating to some, which can result in preventing more people from using the tool.

Separate question now. Have you given any thought to an optimal withdrawal frequency/cadence? Intuition says there might be one that can make a material difference for the long term.

The tool implies a monthly cadence, but I am curious if there is there any statistical evidence for an optimal cadence. I guess there are practical reasons that restrict withdrawing and rebalancing too often. It could be transaction costs, custodial restrictions, or even wash sales rules, since withdrawal can often result in a need to rebalance.

I am assuming the sweet spot to be between one year and something less frequent than monthly, but wonder if that makes a material difference.

Thank you.

[Ben Mathew]

Have you given any thought to an optimal withdrawal frequency/cadence? Intuition says there might be one that can make a material difference for the long term.

The tool implies a monthly cadence, but I am curious if there is there any statistical evidence for an optimal cadence. I guess there are practical reasons that restrict withdrawing and rebalancing too often. It could be transaction costs, custodial restrictions, or even wash sales rules, since withdrawal can often result in a need to rebalance.

I am assuming the sweet spot to be between one year and something less frequent than monthly, but wonder if that makes a material difference.

Because there tends to be momentum at short horizons, rebalancing frequently can be a problem. I caution against doing that here. Eventually we'll try to model transaction costs and rebalancing frequency in more detail, but that will be pretty far down the list of things to do.

For withdrawals, small frequent withdrawals should be better than large infrequent withdrawals because it keeps the funds invested longer. To get some sense for the magnitudes involved, suppose we need to withdraw $5,000 per month * 12 = $60,000 per year. If we start with a balance of $60,000 and withdraw $5,000 per month, we'll have $552 left after a year if the portfolio grows 2% per year, and $1,108 left if the portfolio grows 4% per year. This is what we would lose if we take out all $60,000 at the start of the year. This is not a large loss relative to the $60,000 in annual spending. But if you can tolerate monthly or quarterly withdrawals, that should be better than an annual withdrawal.

[idc]

But if you can tolerate monthly or quarterly withdrawals, that should be better than an annual withdrawal.

Perfect, thank you! So anything between monthly and quarterly seems to be the sweet spot. Matches what I was thinking intuitively

[GaryA505]

In TPAW is there a way to simulate having 50% of assets in a SPIA and 50% in stock?

[Raspberry-503]

In TPAW is there a way to simulate having 50% of assets in a SPIA and 50% in stock?

you can enter the SPIA as an income stream of XX/per month. Check "not adjusted for inflation" unless your SPIA has a rider to keep up with inflation. I'm not super familiar with SPIAS, so if the rider is a fixed number rather guaranteed CPI-adjusted COLA, it might be a little inexact.
You then enter your portfolio value on top of the SPIA, since it's the only variable component. If you are 100% stock you may want to go in "advanced" and select SPAW and specify the AA now and in the future. the default TPAW assumes you let TAPW dictate your AA. (it's possible that with 50/50 it will tell you to be 100% stock in your portfolio either way)

[GaryA505]

In TPAW is there a way to simulate having 50% of assets in a SPIA and 50% in stock?

you can enter the SPIA as an income stream of XX/per month. Check "not adjusted for inflation" unless your SPIA has a rider to keep up with inflation. I'm not super familiar with SPIAS, so if the rider is a fixed number rather guaranteed CPI-adjusted COLA, it might be a little inexact.
You then enter your portfolio value on top of the SPIA, since it's the only variable component. If you are 100% stock you may want to go in "advanced" and select SPAW and specify the AA now and in the future. the default TPAW assumes you let TAPW dictate your AA. (it's possible that with 50/50 it will tell you to be 100% stock in your portfolio either way)

Thanks. While I specified 100% stock that was just for simplicity. I'll put in the numbers and see what it recommends for an AA. It should be interesting.

[jmk]

I imagine I'm not the only person who uses a basic TPAW model supplemented by a program like Boldin, Pralana or ProjectionLab. I'm curious how others integrate the two types of programs.

In broad terms I use TPAW for most of the strategic decisions (e.g. Asset Allocation glidepath, determining my amortized withdrawal amounts) and Pralana for tactical projections of "core" (floor) expenses like medical (ACA), taxation, charity, medical, irmaa, pension impact; and some strategic decisions like Roth conversions.

In practice I iterate results between the two, playing to the strengths of each. Below I outline what's worked for me, in the hopes it might help others. But I'm really curious how others approach things. (I hope this topic doesn't stray too far out away from the main thread; I'll move it out if it does.)

1. Entering income streams is almost identical in both programs. I know my projected employment, pension, social security, house sale, and total portfolio. I struggle in both programs on entering lumpy liability-matched individual TIPs; but in practice it's merely the tedious task of entering a bunch of winfalls in both. (I suppose i could simply exclude the bonds and liabilities on both sides, but that's also tedious.)

2. Pralana is useful to project yearly essential expenses--especially medical, housing costs, education costs; and total taxation. A graph of the Pralana expense summary enables me to visually notice essential spending regimes that I then enter into TPAW under "Extra Expenses" to be covered by bonds. (For instance, my essential expenses are pretty constant age 62 until my son finishes college at my age 68, with another regime age 68-80.) I separate out some of my Go-Go early retirement expenses in TPAW's (underappreciated) "Discretionary Expenses" subsection, used for things that you will want to splurge on if the market does well, but are willing to cut back on if the market does poorly—e.g. vacation travel.

3. With accurate expenses/income, TPAW then derives (a) my General (discretionary) Spending amount, and (b) Asset allocation glidepath (with merton share in the background). This is pretty much by-the-book TPAW. Since I prefer a constant (flat) general spending "income" on top of my quite lumpy essential expenses, I end up using the General Spending card more than I use the main spending card. After setting risk to generate a General Spending curve I like (and cross-checking the main overall spending curve), I flatten the general spending curve by tinkering with spending tilt. (For people like me with lumpy expense floors, the visual assessment of proper risk will be easier once Ben builds in the ability to see essential vs discretionary expenses on the main spending home card.)

4. I then enter TPAW's derived "General Spending" $ amount back into Pralana as a constant discretionary expense under Pralana's SWR option. (See below for issues with doing so.) I also enter the Asset Allocation glidepath into Pralana that roughly matches what I derived in TPAW (from Merton share).

Having a complete income and expense picture in Pralana then enables me to leverage Pralana's strengths in (a) optimizing account withdrawal order, (b) optimizing taxation, e.g. roth conversions versus ACA subsidies etc. These results temper the basic TPAW monthly guidance.

5. #4 will often result in spending changes due to roth conversions and changing ACA subsidies, crossing taxation thresholds some years; so I cross-check spending regimes I originally entered in TPAW, and adjust their specific $ amounts as indicated, even changing whole regimes if it makes sense. If necessary I iterate between Pralana and TPAW until they roughly match.

6. TPAW's scaled block bootstrapping is a more sophisticated simulator than I've found anywhere else. (Certainly better than the crude statistical monte carlos found in Boldin and Pralana.) So I use TPAW as my main stress tester, both to make sure the essentials are covered, and for the acceptable range of discretionary income. However, Pralana is useful for additional stress tests for specifics like inflation rate, taxation rates, ACA changes, cuts in social security etc. (My sense is Boldin goes even further in offering stress testing options.)

That's how I've been doing it. It's idiosyncratic to my own needs.

An issue I've struggled with is how best to simulate the total variable actuarial spending with these programs' deterministic assumptions. At first I used Pralana's actuarial spending model option, given it had the word "actuarial" in its name. Alas Pralana only models SPAW, and more importantly, doesn't allow tilting of SPAW's inherently rising spending curve. (The latter invalidates Pralana's taxation modeling and roth conversion recommendations, the main reason I'm using it.) I next tried Pralana's SWR option by aiming for the final median portfolio to hit legacy amount by death. In theory that would match up Pralana's SWR with TPAW; but results differed too much between the two. I lately have adapted the method of simply adding TPAW's general spending amount into Pralana as described in #4 as a constant yearly expense, which only works because I prefer a flat discretionary retirement budget. (But it's still fitting a square to a circle, frankly.)

How do others integrate TPAW with programs like Boldin and Pralana or other programs used to optimize roth conversions, aca, taxation?

[Raspberry-503]

I have subscriptions to Pralana Online, Boldin and Projection Lab
I also have links to Firecalc and FiCalc, Rich Borke or Dead and NestEggly
I also have my own spreadsheet that is a combination "time value of money" (along the line of Siamond's blog post, the TPAW spreadsheets, or BigERN's CAPE-based withdrawal spreadsheet) and amortized withdrawal (VPW, ABW, TPAW...)
I also regularly try to use RPM (the Retire Portfolio Model) although it seems I run into a brick wall every time because I try to use it for things it's not designed for.
What can you say, I'm a tools nerd

I would say I agree with the characterization that I use my spreadsheet/TPAW for strategic purposes. Do I heave enough to retire, am I comfortable with long-term prospect, what if Die early? Live to 105? etc...

When I retire next year I will also use my spreadsheet/TPAW in a very tactical way: it will dictate how much money I withdraw every month. In fact my spreadsheet them breaks it down into "withdraw XXX, save YYY for taxes, the rest goes into the checking, don't forget mortgage ZZZ and check the level f or emergency fund and card fund are at the expected level, then spend the rest"

I'm still trying to figure out which one of the other tools is best for tactical and short-term modeling: optimize taxes (which of the 3 tax buckets to draw from ), how much to convert to Roth this year, income vs ACA subsidies, income vs IIRMA, etc... This is definitely the stuff TAPW doesn't do (nor am I convinced it should ever try to do). Not of the other tools are perfect for the above but I'm not sure I want to do my own spreadsheet to model taxes, etc... when those are close to usable (Projection Lab has an issue in that it doesn't deal "properly" with year boundaries. If they fixed it, it's be my tool of choice)

The other tools don't really work well with the concept of amortized spending, so I don't try.

[idc]

Very interesting discussion, I am tinkering with it as we speak, and here is my approach.

I tried Pralana Gold (before they went online), Boldin, PlanVision and PlanVision of the paid ones and FireCalc, Fidelity Full View, VWP, TPAW. Currently only using ProjectionLab, TPAW and to some extent VWP. Like you I am using ProjectionLab for tactical purposes, and I really like how you can manage your expenses. Like you I also use TPAW for strategy and especially for projecting an optimal asset allocation and estimating ranges for legacy.

It is super easy to get into too many details, so I decided there are things I am comfortable ignoring for the purposes of planning and things I am comfortable with providing ranges of expenses that would not hinder my ability to deal with the unexpected. I am in Coast FIRE, I am deriving some income from consulting work, but I can retire any time. I am <60 years old. I don't have many lumpy expenses, I am lucky that I have kids out of the house, house paid off, etc. The only projected lumpy expenses of note are a few replacement vehicles, more below. Here is my workflow.

Assumptions:
1. I assumed that I will make Roth conversions, but I decided this is not something I need to include in my modeling. I will decide at the end of every year if and how much to convert based on circumstances. I did this because I am comfortable that Roth conversions will not affect my overall outcome, just the taxation for heirs, and that to me is a good problem to have. So I am not using any Roth in my ProjectionLab modeling. I tried that, I am comfortable on how to do it, but that to me is a fine tuning exercise, which I don't need to complicate my overall planning.
2. I am very comfortable with my required expenses. These are generous allowances for all day to day, health care and basic travel, based on how much I spent over the last 5+ years. They are projected to provide me a very comfortable life, without any reduction in the standard of living. I am not using this figure as a floor in TPAW, but rather as a check to ensure my P5 expenses do not go below this number when I play with the risk slider. I have no floor for any expenses otherwise.
3. As I have the house paid off, I assumed that any lumpy expenses I might need for housing to be included in the yearly budget, which I set to a generous value. I do not plan to downsize or otherwise change the housing situation.
4. I made no provision for the long term care in the terminal years. I assumed that to be too indeterministic and I am comfortable that the amount of assets I will have at the time will be adequate to cover any situation.

With the above in place I performed the following tweaks.
a. I projected a very generous allowance for travel in the go-go, slow go and no go years (in decreasing monthly amounts for each stage) on top of the basic travel budget. I set these as the expected expenses in ProjectionLab and as discretionary expenses in TPAW.
b. I set the lumpy expenses for cars at the approximate times when they are expected to occur in ProjectionLab and TPAW, but I set them as discretionary expenses in TPAW. I spread out these over 10 month periods, to make the scale easier to read in TPAW. Having them as discretionary will provide flexibility to purchase less expensive vehicles if market does bad and I like the ranges TPAW suggested.
c. I ran the "chance of success" for ProjectionLab and it provided 100% success rate for me, with a slow growth in real terms for the net worth.
d. I added a basic legacy figure in TPAW, not to include my residence.
e. I played with the risk tolerance slider in TPAW to optimize for making sure P5 does not fall under my required expenses until I arrived at a sweet spot for Asset Allocation now.

The main outcome for me is that I will never have to compromise my standard of living and the only adjustments I might need to make are on discretionary travel and the cost of the new vehicles I might purchase. The biggest uncertainty I solved though was getting an asset allocation I can reason with.

I am comfortable with an aggressive asset allocation, since I understand both the worst case scenario for me (P5) and the benefit for having a higher projected legacy.

I developed a high confidence in the plan, by playing with the advanced section, especially for the expected returns in TPAW, which to me was the fuzziest aspect of the planning. Even the most conservative scenarios are still allowing me enough comfort.

In summary, ProjectionLab provided me a very clear understanding I will do well under the 80% "things will go normal" scenario, and TPAW provided me excellent confidence I will still be comfortable, with good options and good legacy in the P5 worst case.

[Circle the Wagons]

I imagine I'm not the only person who uses a basic TPAW model supplemented by a program like Boldin, Pralana or ProjectionLab. I'm curious how others integrate the two types of programs.

In broad terms I use TPAW for most of the strategic decisions (e.g. AA glidepath, general withdrawal amounts) and Pralana for tactical projections of "core" expenses like medical expenses, taxation, charity, medical, social security, pension and ACA; and some strategic decisions like roth conversions.

In practice I iterate results between the two, playing to the strengths of each. Below I outline what's worked for me, in the hopes it might help others. But I'm really curious how others approach things. (I hope this topic doesn't stray too far out away from the main thread; I'll move it out if it does.)

1. Entering income streams is almost identical in both programs. I know my projected employment, pension, social security, house sale, and total portfolio. I struggle in both programs on entering lumpy liability matched individual TIPs; but in practice it's merely the tedious task of entering a bunch of winfalls in both. (I suppose i could simply exclude the bonds and liabilities on both sides, but that's also tedious.)

2. Pralana is useful to project yearly essential expenses--especially medical, housing costs, education costs; and total taxation. A graph of the Pralana expense summary enables me to visually notice essential spending regimes that I then enter into TPAW under "Extra Expenses". (For instance, my essential expenses are pretty constant age 62 until my son finishes college at my age 68, with another regime age 68-80.) I separate out some of my Go-Go early retirement expenses in TPAW's (underappreciated) "Discretionary Expenses" subsection, used for things that you will want to splurge on if the market does well, but are willing to cut back on if the market does poorly—e.g. vacation travel.

3. With my complete income and essential expense streams thus entered, TPAW then gives me the ability to derive (a) my General (discretionary) spending amount, and (b) Asset allocation glidepath (with risk aversion and merton share in the background). This is pretty much by-the-book TPAW. Since I prefer a constant (flat) general spending "income" on top of my quite lumpy essential expenses, I end up using the General Spending card more than I use the main spending card. I flatten the general spending curve by tinkering with spending tilt and "preference for the future". (This step will be easier once Ben builds in the ability to see essential vs discretionary expenses on the main spending home card.)

4. I then enter TPAW's "General Spending" $ amount back into Pralana as a constant discretionary expense under the "specified spending only" SWR option. (See below for issues with doing so.) I enter an Asset Allocation glidepath into Pralana that roughly matches what I derived in TPAW (from Merton share).

Having a complete income and expense picture in Pralana then enables me to leverage Pralana's strengths in (a) optimizing account withdrawal order, (b) optimizing taxation, e.g. roth conversions versus ACA subsidies etc. These results temper and often override the basic TPAW monthly guidance card.

5. #4 will often result in spending changes due to roth conversions and changing ACA subsidies, or crossing taxation thresholds; so I cross-check spending regimes I originally entered in TPAW, and adjust their specific $ amounts as indicated, even changing whole regimes if it makes sense. If necessary I iterate between Pralana and TPAW until they roughly match.

6. TPAW's scaled block bootstrapping is a sophisticated simulator than I've found anywhere else. (Certainly better than the crude statistical monte carlos found in Boldin and Pralana.) So I use TPAW as my main stress tester, both to make sure the essentials are covered, and for the acceptable range of discretionary income. However, Pralana is useful for additional stress tests for specifics like inflation rate, taxation rates, ACA changes, cuts in social security etc. (My sense is Boldin goes even further in offering stress testing options.)

That's how I've been doing it. It's idiosyncratic to my own needs.

An issue I've struggled with is how best to simulate the total variable actuarial spending with these programs' deterministic assumptions. At first I used Pralana's actuarial spending model option, given it had the word "actuarial" in its name. Alas Pralana only models SPAW, and more importantly, doesn't allow tilting of SPAW's inherent rising spending curve. (The latter, especially, invalidates Pralana's taxation modeling and roth conversion recommendations, the main reason I'm using it.) I next tried Pralana's SWR option by aiming for the final median portfolio to hit legacy amount by death. In theory that would match up Pralana's SWR with TPAW; but results differed too much between the two. I lately have adapted the method of simply adding TPAW's general spending amount into Pralana as described in #4 as a constant yearly expense, which only works because I prefer a flat discretionary retirement budget. (But it's still fitting a square to a circle, frankly.)

How do others integrate TPAW with programs like Boldin and Pralana or other programs used to optimize roth conversions, aca, taxation?

I use a similar approach which I posted about here:

viewtopic.php?p=8266926#p8266926

TPAW is the core. Pralana is used for the tax engine and any decisions related to it (e.g., Roth conversions). You can "curve fit" Pralana to the AA path from TPAW, with up to five points on the curve.

[GaryA505]

In TPAW is there a way to simulate having 50% of assets in a SPIA and 50% in stock?

you can enter the SPIA as an income stream of XX/per month. Check "not adjusted for inflation" unless your SPIA has a rider to keep up with inflation. I'm not super familiar with SPIAS, so if the rider is a fixed number rather guaranteed CPI-adjusted COLA, it might be a little inexact.
You then enter your portfolio value on top of the SPIA, since it's the only variable component. If you are 100% stock you may want to go in "advanced" and select SPAW and specify the AA now and in the future. the default TPAW assumes you let TAPW dictate your AA. (it's possible that with 50/50 it will tell you to be 100% stock in your portfolio either way)

Yes, I tried it and it did go to 100% stock.

[jmk]

(deleted.)

[jmk]

I have subscriptions to Pralana Online, Boldin and Projection Lab I also have links to Firecalc and FiCalc, Rich Borke or Dead and NestEggly
I also have my own spreadsheet that is a combination "time value of money" (along the line of Siamond's blog post, the TPAW spreadsheets, or BigERN's CAPE-based withdrawal spreadsheet) and amortized withdrawal (VPW, ABW, TPAW...) I also regularly try to use RPM (the Retire Portfolio Model) although it seems I run into a brick wall every time because I try to use it for things it's not designed for. What can you say, I'm a tools nerd

I'll admit I don't understand why you need to use VPW, ABW and TPAW, given that TPAW synthesizes the other two, and offers even more functionality. By my assessment there is literally nothing APW/VPW can do that isn't possible in TPAW. If I did all that you do it would feel more like an anxiety disorder than investment planning.

I suppose ERN spreadsheet is the exception, as a cross-check, since his method is entirely different from both TPAW (VPS/ABW) and from deterministic programs like Boldin and Pralana. ERN calculates the (variable) SWR percentile of success based on previous CAPE regimes, looking only at the number of historical past runs 1871-present that succeeded. TPAW has similar "historical returns" option, with an indirect CAPE adjustment to the discount rate of the pmt equation, but within an annuity model.)

[Raspberry-503]

I'll admit I don't understand why you need to use VPW, ABW and TPAW, given that TPAW synthesizes the other two...

If my initial post wasn't clear, my own spreadsheet combines TMV and Amortization (was you call actuarial), just like TPAW does .ABW, VPW... were used both to develop and cross-check my spreadsheet. And yes, I agree that I am more or less building my own version of TPAW, which has been extremely useful in understanding how TPAW works (I won't trust a tool I don't understand). If anything it allows me to approximate "what would have happened using TPAW in 1966" to the extent I have CAPE historical data (I haven't tried to to re-derive historical value for TPAW's regression of future-looking stock values, although I think it's possible).

To me it's interesting to understand the differences between VPW and ABW, even though they are "basically the same" (I know longinevst would take exception to this statement, but I stand but it, VPW is a special case of ABW).

As I said, when I retire I will use my spreadsheet and TPAW, which have historically returned very similar results for similar expected returns input, and will only train my wife in the use of TPAW should I become unable to take care of it.

I agree that it you're not one to spend hours to understand why TPAW and your spreadsheet diverge and enjoy the process and stimulation, use TPAW!

[jmk]

I use a similar approach which I posted about here:
viewtopic.php?p=8266926#p8266926
TPAW is the core. Pralana is used for the tax engine and any decisions related to it (e.g., Roth conversions). You can "curve fit" Pralana to the AA path from TPAW, with up to five points on the curve.

Wow, you not only utilize exactly the same iterative process, but even use Pralana. If I wasn't already happily married I'd ask you on a date.

It sounds like you did not have the same issue "curve fitting" TPAW and Pralana I described in my post. I started by simply adjusting the SWR "miscellaneous" (labeled "general") expense until death portfolio matched 50% percentile in Pralana's Monte carlos. But lately I've landed on using Pralana's "total essential expenses" as my "essential expenses" in TPAW. Then deriving the (flattened) discretionary spending yearly amount in TPAW, and then adding that back in to Pralana as my constant "miscellaneous non-essential expense", and letting things fall where they lie. (The median ends up being 63% success or so, but as is clear I don't care much about that particular metric.)

I trust you noticed that Pralana uses geometric return inputs (which it converts to an approximate arithmetic average behind the scenes), whereas TPAW wants the arithmetic mean. It's easy to convert one to the other, but they are a few percentages apart with equities.

[Circle the Wagons]

I use a similar approach which I posted about here:
viewtopic.php?p=8266926#p8266926
TPAW is the core. Pralana is used for the tax engine and any decisions related to it (e.g., Roth conversions). You can "curve fit" Pralana to the AA path from TPAW, with up to five points on the curve.

Wow, you not only utilize exactly the same iterative process, but even use Pralana. If I wasn't already happily married I'd ask you on a date.

It sounds like you did not have the same issue "curve fitting" TPAW and Pralana I described in my post. I started by simply adjusting the SWR "miscellaneous" (labeled "general") expense until death portfolio matched 50% percentile in Pralana's Monte carlos. But I've landed on deriving the (flat) general spending in TPAW, using that as my flat miscellaneous expense in Pralana ("use specified expenses only"), and letting things fall where they lie. (The median ends up being 63% success or so.)

I trust you noticed that Pralana uses geometric return inputs (which it converts to an approximate arithmetic average behind the scenes), whereas TPAW wants the arithmetic mean. It's easy to convert one to the other, but they are a few percentages apart with equities.

Great minds think alike. It's good to hear others have come to the same conclusion that these two tools are the superior options out there and are synergistic. TPAW Planner, specifically, is utterly unique in its approach and elegance.

Reading your post upstream again, we have some differences in approach. I set an expense target and plug it into Pralana to refine, e.g., for taxes. The result goes into TPAW as general "regimes" (your term). I'm not asking either tool how much to spend. I don't rely on Pralana's spending strategies except as a sense check (MC consumption smoothing).

TPAW gives the AA that corresponds to the downside risk around spend that is most acceptable. The elegance is in allowing the user to rapidly visualize required risk tradeoffs across decades.

The AA goes back into Pralana to further refine spending outlook and taxes. It's unfortunately rather crude (I've asked them for more time periods), but if you revisit the plan regularly to tweak, I believe it allows you to tack very closely to a long-term efficient path.

[Circle the Wagons]

I trust you noticed that Pralana uses geometric return inputs (which it converts to an approximate arithmetic average behind the scenes), whereas TPAW wants the arithmetic mean. It's easy to convert one to the other, but they are a few percentages apart with equities.

Yes, good catch. I had this wrong for months before catching it. I confirmed with the developers that they convert the input to arithmetic before running MC. Suggested they better label the ROR input page too.

[ZMonet]

I'm a big fan of TPAW Planner and plan to use it for our retirement withdrawal strategy. I'm confused about the legacy settings, though. I've input a $3 million legacy target, with $750,000 of that being the estimated value of our home. When I run the simulations, the 50% outcome shows a final estate value of only about $200,000. I expected it to be closer to $2.25 million ($3 million - $750,000). Could someone explain how the legacy calculations work?

[ConstantChrysalis]

I'm a big fan of TPAW Planner and plan to use it for our retirement withdrawal strategy. I'm confused about the legacy settings, though. I've input a $3 million legacy target, with $750,000 of that being the estimated value of our home. When I run the simulations, the 50% outcome shows a final estate value of only about $200,000. I expected it to be closer to $2.25 million ($3 million - $750,000). Could someone explain how the legacy calculations work?

The 50% value should be approximately $3 million, so it sounds like the planner can't meet your legacy target with the current settings and assumptions. Try throwing in another large income source temporarily and see what happens.

[Ben Mathew]

When I run the simulations, the 50% outcome shows a final estate value of only about $200,000. I expected it to be closer to $2.25 million ($3 million - $750,000).

As ConstantChrysalis said, this is because there aren't enough funds to fund the $2.25 million legacy target. Note that essential extra expenses get prioritized over legacy. If you have categorized a lot of extra expenses as "essential," try making some of them "discretionary" instead and see if the legacy outcome gets closer to target. Note also that the median outcome will be less than the expected outcome due to the portfolio balance being right tailed. So even if there are sufficient funds, the median legacy outcome will be less than the target legacy outcome.

[ScubaHogg]

Ben, can you talk a little about what exactly adding “discretionary expenses” does in the background?

Essential expenses are simple enough, but I’m not clear how discretionary expenses are any different than just the normal available to spend amounts at any time point

Are they treated a bit like a time preference?

[Raspberry-503]

Essential Expenses are funded as safe bonds, they are guaranteed and highest priority
Discretionary expenses are funded from the risk portfolio, the mix of stock and bonds which is more volatile but has more upside

Of course it doesn't mean that you necessarily hold actual bonds for those. For example your essential spending could happen to match your social security and "cancel each other out"

I think they are no different than any other "normal amount" except that you decide when to spend it, for example you want $20K a year for the first 15 years of retirement for extra travel money, it asks the tool to schedule disbursement to account for this "step" distribution, if money allows.

[jmk]

To me it's interesting to understand the differences between VPW and ABW, even though they are "basically the same" (I know longinevst would take exception to this statement, but I stand but it, VPW is a special case of ABW).

VPW is definitely a special case of ABW. Though I obviously prefer TPAW, I was originally going to recommend my wife use VPW after I pass given it's so simple (simplistic?). But the idea of only amortizing income while completely ignoring expenses seems fundamentally unwise to me. I suppose VPW's presumption is to have your core expenses covered automatically by an annuity, liability matching, or other arrangements (i.e. excluded from the model), so she'd just use VPW with her discretionary accounts (as an improvement on RMD method). That will essentially generate a similar result to what TPAW would have.

But...here's the rub: it if I go that route my wife can even more easily just run a PMT equation I've left her in excel following my directions. I can type that up in half a page, shorter than VPW's instructions. She won't even have to know about bogleheads

In short: TPAW is best of class ABW. But for use by an elderly spouse after we pass with little financial knowledge or inclination, automation of essential expenses + PMT equation for discretionary rather than VPW. Oh, and I'll leave her a photo of Merton she can tape to my tombstone for nostalgia.

[ScubaHogg]

Discretionary expenses are funded from the risk portfolio, the mix of stock and bonds which is more volatile but has more upside

…

I think they are no different than any other "normal amount" except that you decide when to spend it, for example you want $20K a year for the first 15 years of retirement for extra travel money, it asks the tool to schedule disbursement to account for this "step" distribution, if money allows.

Yeah, this is my guess. It’s just disburses from the risk portfolio at the specified higher rate for the specified time. Like a very detailed spending tilt

But I don’t know that’s right for sure

[Wannaretireearly]

Enjoying the discussions on using Pralana for expenses and taxes, and then feed into TPAW planner. I also use and like projection lab.

If there are other discussions/threads on how folks use/cross pollinate these tools that would be interesting to me.

[Ben Mathew]

I think they are no different than any other "normal amount" except that you decide when to spend it, for example you want $20K a year for the first 15 years of retirement for extra travel money, it asks the tool to schedule disbursement to account for this "step" distribution, if money allows.

Yeah, this is my guess. It’s just disburses from the risk portfolio at the specified higher rate for the specified time. Like a very detailed spending tilt

But I don’t know that’s right for sure

Yes, that's right. It can be useful to think about this using the bucket view. We can think of our wealth as having been split up into different buckets to fund different goals: retirement spending at age 65, 66, ..., 100, and legacy are the basic goals. The asset allocation in each bucket would be determined by the amount of risk that we want to take with that goal. The funds stay in the bucket and grow, and when the time arrives, the goal gets funded by whatever's in the bucket.

Extra expenses are just additional buckets. In principle, each goal can have its own asset allocation, but we've restricted it to two options (1) essential goals are funded by 100% bonds, and (2) discretionary goals are funded by the same asset allocation as general retirement spending (which is determined by the risk slider). If the risk slider gives an asset allocation (for the total portfolio) of 30/70, then discretionary goals get funded with the same 30/70. So it is effectively a "detailed spending tilt."

[RajaDada]

Hi - A couple questions regarding the legacy setting of the planner:

1. I realize this question has been asked before. I have tried to look at the answers but I am still confused. My wife and I are near retirement. We have about $4M, and ideally, we would like to leave a legacy of $4M. Given our pensions and annual spending needs, this does not seem unrealistic. However, in order to get the tool to display $4M in the 50th percentile legacy card, I have to enter a legacy target of about $6M. Out of curiosity, I doubled the value of our current portfolio, but it did not alter the value of the 50th percentile by much - it was still around $4M. Why can't I get the target to align better with the 50th percentile outcome?

2. I am having a harder time articulating this second question. The legacy input is static (I think). It does not automatically adjust as my portfolio balance does. Do I need to update that input regularly to achieve the original goal? That is, if I input some amount as the target in 2025, is it going to be the same amount in 2035 or 2045? Or do I have to make an inflation adjustment to the target over time? Sorry if this is confusing. I am confused trying to articulate it.

As always, many thanks for the tool! It gives me a much greater sense of security as we head into retirement.

Best, Noel

[dcabler]

To me it's interesting to understand the differences between VPW and ABW, even though they are "basically the same" (I know longinevst would take exception to this statement, but I stand but it, VPW is a special case of ABW).

VPW is definitely a special case of ABW. Though I obviously prefer TPAW, I was originally going to recommend my wife use VPW after I pass given it's so simple (simplistic?). But the idea of only amortizing income while completely ignoring expenses seems fundamentally unwise to me. I suppose VPW's presumption is to have your core expenses covered automatically by an annuity, liability matching, or other arrangements (i.e. excluded from the model), so she'd just use VPW with her discretionary accounts (as an improvement on RMD method). That will essentially generate a similar result to what TPAW would have.

But...here's the rub: it if I go that route my wife can even more easily just run a PMT equation I've left her in excel following my directions. I can type that up in half a page, shorter than VPW's instructions. She won't even have to know about bogleheads

In short: TPAW is best of class ABW. But for use by an elderly spouse after we pass with little financial knowledge or inclination, automation of essential expenses + PMT equation for discretionary rather than VPW. Oh, and I'll leave her a photo of Merton she can tape to my tombstone for nostalgia.

Yep - that's my plan B. I've pre-run the PMT function, with a fixed "rate" and provided a table of percentages for each year. I have a simple spreadsheet where all she has to do is enter the dollar value of the accounts used for discretionary spending and out pops an annual budget for discretionary spending.

Cheers.

[jmk]

Ben, thinking of the upcoming liability matching element. From earlier comments it seems like TPAW will recommend a duration for bonds in order to immunize the input expenses. In calculating the duration of liabilities, will the tool also take into account the liability-matched assets like individual TIPs? I'm sure I'm not the only user who has a mix of individual bonds and bond funds which collectively cover one's floor. I currently try to keep the weighted average duration of the ladders and funds as a whole at the requisite level; is that how TPAW will work too?

[Ben Mathew]

Hi - A couple questions regarding the legacy setting of the planner:

1. I realize this question has been asked before. I have tried to look at the answers but I am still confused. My wife and I are near retirement. We have about $4M, and ideally, we would like to leave a legacy of $4M. Given our pensions and annual spending needs, this does not seem unrealistic. However, in order to get the tool to display $4M in the 50th percentile legacy card, I have to enter a legacy target of about $6M. Out of curiosity, I doubled the value of our current portfolio, but it did not alter the value of the 50th percentile by much - it was still around $4M. Why can't I get the target to align better with the 50th percentile outcome?

This is because the distribution of outcomes is right tailed, which makes the median outcome lower than the expected outcome. Spending and legacy inputs become the target expected outcome, so the median outcomes will be lower. To get the median legacy outcome to be $4 million, you'll have to increase the legacy target to above $4 million.

2. I am having a harder time articulating this second question. The legacy input is static (I think). It does not automatically adjust as my portfolio balance does. Do I need to update that input regularly to achieve the original goal? That is, if I input some amount as the target in 2025, is it going to be the same amount in 2035 or 2045? Or do I have to make an inflation adjustment to the target over time? Sorry if this is confusing. I am confused trying to articulate it.

Yes, only the portfolio balance is auto-updated. All other inputs including the legacy target have to be updated manually at regular intervals to keep the plan up-to-date.

Note that the simulation itself does scale discretionary and legacy goals in response to portfolio performance. So the simulation results reflect the fact that these goals would be scaled as we move forward in time depending on portfolio performance. But we don't auto-update it in the plan itself. We could in principle do this—auto-update essential goals to reflect inflation, and discretionary and legacy goals to reflect portfolio performance. We haven't done it because we feel it could lead to confusion. If we can figure out a good UI that isn't confusing to use, we might try to do this. But this would be pretty far down the to-do list.

As always, many thanks for the tool! It gives me a much greater sense of security as we head into retirement.

Glad to hear you're finding the tool useful!

[Raspberry-503]

Is the scaling the same factor as the portfolio growth?
If in Jan I design a plan that has $1000 discretionary income per month for travel for the first 19 years when my portfolio was $2M, and we have an incredible year, and by November my portfolio is $3M (1.5x): when I go into the tool in Nov it will recommend I draw more than in Jan, because there is now an extra million to spread around, but should I manually adjust my plan to set discretionary expenses to $1500 to comparatively spend more on gogo travel?I would love to set some initial conditions and have the tool do the scaling for me.

Or asking another way, in Jan the projected 95 percentile says I will be able to withdraw X in November. If we assume the return until Nov are exactly the ones predicted for the 95 percentile, will the amount the tool tells me to send in Nov be what I predicted in Jan, or will it be lower because the projection assumed the discretionary expenses would grow, but didn't change it by Nov because I don't know what the "rules" are to decide what to change it to.

[Ben Mathew]

Ben, thinking of the upcoming liability matching element. From earlier comments it seems like TPAW will recommend a duration for bonds in order to immunize the input expenses. In calculating the duration of liabilities, will the tool also take into account the liability-matched assets like individual TIPs? I'm sure I'm not the only user who has a mix of individual bonds and bond funds which collectively cover one's floor. I currently try to keep the weighted average duration of the ladders and funds as a whole at the requisite level; is that how TPAW will work too?

The initial implementation of duration matching would calculate the duration of the spending being funded by bonds—so the target duration of bonds, broken down between essential/floor (LMP) vs discretionary/general (risk portfolio). The user can specify if they will be duration matching the floor and/or the risk portfolio, and the simulation will reflect that choice. So the risk reduction from duration matching can be seen in the simulation.

If the user specifies that the floor is being duration matched, the simulation will assume that it is being matched with a bond ladder. Modeling the impact of using bond funds instead of a bond ladder is not something we're tackling in this round. But we hope to get to it eventually given that it is a common way of duration matching.