Strategy for Withdrawing LTCG from the Taxable Account

Interesting that you don't like withdrawing unrealized gains last for stocks. Seems to me that it's the typical choice (so as to leave unrealized gains sitting in there untaxed until my heirs get them tax free). Withdrawing unrealized gains first is the one I cannot make sense of. I hope our experts can clarify!

@out-of-office I think you just answered my question 4: you might want to leave capital gains till last when you intend to give it to your heirs with a step up. However
if you have taxable high up in the withdrawal priority, Pralana will take unscheduled withdrawals from your taxable account so the two might need to be coordinated?

When would you choose to take CG first? When your spouse might the be one withdrawing, at a higher rate. Or when you don’t intend to leave money for heirs, or will do so with roths? If not doing Roth conversions, you might elect to use up the "head space" in the 0% bracket with stocks that have accumulated lots of gains. Or the net gain from 0>15% later might be more than one has with conversions.

It seems hard to make a blanket choice for the whole plan.

When to use “even” (proportional) remains a complete mystery.

Please forgive the double-posting, but for those viewing this thread and not this other one, here is an except about how the math works for the LTCG Withdrawal Strategies:

In the context of unrealized LTCG withdrawal strategy, proportional ('Distribute withdrawals evenly') means the % of the account balance, at the time of the withdrawal, that is unrealized LTCG....e.g. 10% or 30%.

Example: Investment account has an balance of $100K of which $30K or 30% is unrealized LTCG. (Some or all of the unrealized LTCG may have existed at plan start and entered on the Initial Balances page or accumulated from growth during the plan years.)

If a $10K scheduled or unscheduled withdrawal is made from that account and the withdrawal strategy for unrealized LTCG is:

• 'Evenly' aka Proportional: $10K will be withdrawn, 30% or $3K will be taxed as realized LTCG that year, the remaining balance will be $90K total / $27K unrealized LTCG.
• First: $10K will be withdrawn, all of which will be taxed as realized LTCG that year, the remaining balance will be $90K total / $20K unrealized LTCG.
• Last: $10K will be withdrawn, none of which will be taxed as realized LTCG that year, the remaining balance will be $90K total / $30K unrealized LTCG.

Posted by: @cstone

↑

Please forgive the double-posting, but for those viewing this thread and not this other one, here is an except about how the math works for the LTCG Withdrawal Strategies:

In the context of unrealized LTCG withdrawal strategy, proportional ('Distribute withdrawals evenly') means the % of the account balance, at the time of the withdrawal, that is unrealized LTCG....e.g. 10% or 30%.

Charlie, that is what I originally thought for many years! But then perhaps the manual wording needs to be clarified, in particular the sentence in question in the larger excerpt in my original post at the top:

If “Distribute withdrawals evenly” is selected, withdrawals will be made in proportion to the interest, realized gains, tax exempt and unrealized capital gains categories specified on the Taxation page

This made me think that Pralana would use the taxation growth categories and proportions (which didn't make sense to me) rather than the % of Unrealized gain, which makes total sense. I believe what Pralana calls "evenly" syncs up with the common provider choice of using "average cost" of your capital gains. Rare case where reading the manual caused me to change from correct to incorrect interpretation. 🙂

@jkandell You were correct for those many years! The language in the manual is not correct. My apologies for this error. I have rewritten it to coincide with your expectation and the example I provided in my prior post.

I, and many new subscribers, thank you for alerting me to this.

Posted by: @cstone

↑

I, and many new subscribers, thank you for alerting me to this.

@jkandell At least one old subscriber as well thanks you for bringing this topic up and having it clarified! I have puzzled for years how the definition of "Distribute withdrawals evenly" (up until now being defined as related to Growth Taxation) could have ever possibly resulted in a realistic projection for realizing capital gains. What could Growth possibly have to do with it? The corrected explanation of realized gains being based on account balance / unrealized gains percentage makes perfect sense. Thanks for asking this question, leading to this newfound clarity. I suppose understanding how capital gains are calculated would have been revealed if I'd studied the Tabular Projections output close enough. But I never did.

I made a comment in the other related LTCG thread but have just now realized that that thread is related to how gains are allocated across Asset Classes during rebalancing. My comment was based on LTCG being realized during Unscheduled Withdrawals and should have been made in this particular thread. Frankly, I am not clear how the two LTCG topics are intertwined, though I expect they are closely related and complicated. At a minimum, I feel much better now about how LTCG are realized for withdrawals knowing that the "Even"/"Proportional" option does exactly what I would think it should do and should better mimic a brokerage's Avg Cost method for realizing gains. I personally use ETFs and the Spec ID method, selecting specific lots at sale, so Pralana's Avg Cost quasi-approximation doesn't work perfectly for me either. But it's sure closer than I understood it to be earlier.

Just brainstorming... As @ricke mentioned in the other LTCG thread, I also will likely never sell my oldest, highest-gain lots and will retain them for step-up in basis at my death. With that thought in mind, perhaps a new "Realized LTCG Percentage" could be associated with each Time Period in the plan (or better yet even another set of Time Periods dedicated solely to LTCG taxation). That would allow the user to control his/her own LTCG taxation (rather than have Pralana default to the proportional percentage) and in addition allow the user to vary that percentage over the life of the plan to mimic his/her personal intent of realizing gains in reality. @cstone has already articulated the horrendous complications related to taxation of capital gains so this might only complicate things. But it does seem, at this early morning hour at least, to place control of capital gains taxation more in the hands of the user while simultaneously relieving Pralana (maybe?) of algorithmically calculating the gains.

Posted by: @ilovemybeagles

↑

I made a comment in the other related LTCG thread but have just now realized that that thread is related to how gains are allocated across Asset Classes during rebalancing. My comment was based on LTCG being realized during Unscheduled Withdrawals and should have been made in this particular thread. Frankly, I am not clear how the two LTCG topics are intertwined, though I expect they are closely related and complicated.

I too am not quite sure how rebalancing relates, although I feel the answer to this thread answers the other thread. (I should have given this thread a better subject.)

Another issue that overlaps but I'm not sure exactly how is the Withdrawal Priority setting for unscheduled withdrawals. If you list "taxable" toward the top of the list, you are obviously more likely to have taxable drawn from, and vice versa. So that is another "lever" one has to model these things.

Just brainstorming... As @ricke mentioned in the other LTCG thread, I also will likely never sell my oldest, highest-gain lots and will retain them for step-up in basis at my death. With that thought in mind, perhaps a new "Realized LTCG Percentage" could be associated with each Time Period in the plan (or better yet even another set of Time Periods dedicated solely to LTCG taxation). That would allow the user to control his/her own LTCG taxation (rather than have Pralana default to the proportional percentage) and in addition allow the user to vary that percentage over the life of the plan to mimic his/her personal intent of realizing gains in reality. @cstone has already articulated the horrendous complications related to taxation of capital gains so this might only complicate things. But it does seem, at this early morning hour at least, to place control of capital gains taxation more in the hands of the user while simultaneously relieving Pralana (maybe?) of algorithmically calculating the gains.

This is an intriguing idea, and possibly takes us toward a potential step forward. The fact Pralana allows several accounts under taxable buttresses your suggestion: One account could be used for "taxable held off for step up", and the other one might be labeled "small gains". Given that Pralana can't keep track of lots, the user would need to update the basis on each bucket each year. That wouldn't be as burdensome as it seems, since people who manage LTCG probably do that kind of distinction anyway. And anyone else could just either not use those other accounts or use them for CDs and banks etc.

1. Pralana allows multiple accounts under Taxable. Users who intend to leverage LTCG can use these separate accounts for buckets with unique names e.g. "Stocks to leave for heirs", "High gains to harvest" "Loss Harvesting". These might literally be separate accounts or they might be mental buckets of different lots/basis/purposes within a single actual account.
2. Users would set the basis each year for each account. An accounts could even be used as placeholder for negative unrealized (saved losses).
3. Users would have a new ability to set the unscheduled taxable withdrawal priority for each taxable account and each time period: "First" "Last" (or "Never use"?), and "Medium priority" (the latter the default).
   (So overall order of withdrawal would be: Inherent growth of gains/dividends etc within the account first, of course. Then scheduled withdrawals, then "first" priority accounts, then "medium" priorities, and finally ""last priority" accounts last. With "never" for accounts saved for heirs)
4. It is likely that people will have different harvesting strategies at different time periods, so allowing users to set start year/end year for each account would be needed. (E.g. You might want to only gain harvest when you first retire; you might to save your losses for when you know a huge treasury bond is maturing.)
5. Users already set Account Withdrawal Priority (between Taxable, Roth, Tax-deferred, etc), that shapes when taxable is called, so that gives another lever to influences LTCG.

I also have the issue with large unrealized LTCG in my taxable accounts (all my assets). My plan is not to sell any of these unrealized LTCG in taxable (unless the gain turns into loose or step-up happened) and use SS + RMD + taxable dividends for expenses. However, Asset Allocation/Location may require moving assets to/from taxable accounts. Asset Allocation rebalancing may avoid selling taxable assets by manipulating only IRA/Roth assets, not Taxable. But Asset Location may require selling taxable unrealized LTCG. My Asset Locations are not ideal, but I do not want to fix it and pay taxes. What I am trying to do is to be realistic and to insert my not ideal Locations’ information into the tool and hope that the tool will have minimum movement of taxable assets. Since the accounts will change unproportionally in the future I need to adjust also future Locations accordingly to avoid extra taxes. This turns out to be a complex task. With the wonderful online tool improvement, I have now 12 assets to work with. Apparently, the order of the assets’ names is used in the Asset Allocation/Location algorithm. When I created the Assets list, I was not aware that the order is important and now I have difficulties changing the order without messing up the numbers that are associated with the names. For example, Bonds are the last in my list, so the algorithm almost does not move them, and most bonds are in IRA, few in Roth and nothing in taxable. This is good, almost ideal. However, REITS are in the middle of the list and therefore the algorithm adds some REITS into taxable which is not my intention. So, what I would like is

1. Understand how the algorithm works, and which parameters influence it.
2. Have the ability to reorder the asset names list (klick and drag) so the tool keeps the associated date with the name. (If the algorithm really uses the order.)

Adjusting 12 assets for Allocation/Location is not an easy task. So, I am not suggesting changing the algorithm but working with what we have.

@jkandell Intrigued squared. I like it!

My initial impression is that you have translated my humble suggestion into a more practical, workable solution. I had originally envisioned the user explicitly specifying a Realized Taxable LTCG Percentage per Time Period, thereby relieving Pralana of convoluted gyrations to impute LTCG taxable based on Cost Basis algorithmically allocated to each Time Period. There may yet be value in this approach. But I find a bit more precision and simplicity in your proposal. If I understand your development of the idea correctly, my "Time Period/explicit Realized Taxable LTCG Percentage" idea is replaced with your multiple "virtual" Taxable Account-centric idea where Cost Basis is maintained per Account, with LTCG taxable calculated by Pralana as 1-(Cost Basis/Account Balance) and levied in a prioritized manner across the multiple "virtual" Taxable Accounts.

This will inevitably make this a long post, but maybe this can be better visualized with a basic outline of our proposals, realizing that both ideas are bareboned at present and would require much fleshing out by the esteemed Mr. @cstone. (This basic outline can also serve as an invitation to you to correct and elaborate on anything I may have misunderstood from your suggestion.)

LTCG Taxation - Portfolio Time-Period Approach

In Pralana, User defines:

• Year 2025 - Realized Taxable LTCG Percentage 30% (this desired realization of taxable LTCG would be in place until 2029, as the current Time Periods work)
• Year 2030 - Realized Taxable LTCG Percentage 45%
• Year 2035 - Realized Taxable LTCG Percentage 70%

User Responsibility:

1. Inspect Pralana's Taxable Account Unscheduled/Scheduled Withdrawals in Tabular Projections and cross-reference them with the actual brokerage portfolio to see what kind of lots would have to be liquidated to satisfy each withdrawal.
2. Derive a "Realized Taxable LTCG Percentage", by Time Period, for each Pralana withdrawal (gains/cost basis)

This puts the onus on the User but with the benefit of Pralana providing higher fidelity projections, critically important given that LTCG taxation is a make-or-break factor in such things as ACA cliff modelling, Roth Conversions, IRMAA cliff avoidance, various phase-ins/outs, etc. (but then this same benefit of higher fidelity extends to your proposal as well).

LTCG Taxation - Multiple "Virtual" Taxable Account Approach (again, clarification of any misunderstanding welcomed)

Assume $100,000 Total Cost Basis across all Taxable Accounts, with a $1,000,000 Total Account Balance.

In Pralana, User defines:

• "Virtual Account 1: Low Cost Basis" - Cost Basis $10,000 ($800,000 LTCG, 80% of portfolio Gains) - your "Postpone gain" terminology
• "Virtual Account 2: Moderate Cost Basis" - Cost Basis $60,000 ($190,000 LTCG, 19% of portfolio Gains) - your "No priority" terminology
• "Virtual Account 3: High Cost Basis" - Cost Basis $30,000 ($10,000 LTCG, 1% of portfolio Gains) - your "Gain Harvest" terminology

Additionally, a second sublevel of Withdrawal Priority is defined for "Taxable Investments", essentially replacing the current "LTCG Withdrawal Strategy".

LTCG Withdrawal Strategy

• Virtual Account 1 - Withdrawal Priority 3 (liquidate low basis investments last / realize high LTCG taxes last)
• Virtual Account 2 - Withdrawal Priority 2
• Virtual Account 3 - Withdrawal Priority 1 (liquidate high basis investments first / realize low LTCG taxes first)

Additionally - and I'm not sure whether this was your intent - the sublevel Withdrawal Taxable Investment Withdrawal Priority could also be defined by Year, as is currently done with the toplevel "Account Withdrawal Priority", for example:

LTCG Withdrawal Strategy - by Year

• 2025 - Virtual Account 1 - Withdrawal Priority 1
• 2025 - Virtual Account 2 - Withdrawal Priority 2
• 2025 - Virtual Account 3 - Withdrawal Priority 3

• 2030 - Virtual Account 1 - Withdrawal Priority 3
• 2030 - Virtual Account 2 - Withdrawal Priority 2
• 2030 - Virtual Account 3 - Withdrawal Priority 1

• 2035 - Virtual Account 2 - Withdrawal Priority 2 (Virtual Accounts 1 & 3 have run dry)

(Note: As I read back over your post, I think I glossed over your "Postpone gain"/"Gain Harvest"/"No Priority" designation for each Taxable Account and having Pralana automatically determine the Withdrawal Priority. What I've outlined above (explicitly dragging/dropping Withdrawal Priority in the desired order) is then an acknowledged embellishment on your original suggestion. But I decided to leave it documented as another possible alternative.)

User Responsibility:

1. For each Taxable "Virtual Account", update the current Cost Basis every Jan, adjusting for actual remaining portfolio balance and desired realization of Taxable LTCG going forward into the plan.
2. Review Taxable Account Withdrawal Priority of each Virtual Account (and Year?) and adjust as needed to match desired Realized Taxable LTCG, based on the actual cost basis/gain composition of the portfolio and projected Pralana withdrawals

Welp, that's my current understanding and contrast/comparison of our current proposals. This rehashing may or may not have been a productive exercise. But it helped me to clarify my own thinking and to lean more toward your approach, while also perhaps blending in the "Time Period" based aspect of my idea into yours. It also has made it clear to me that there needs to be a simple "one-click" option for users who have no desire to manage their realized LTCG to match the actual composition of their Taxable portfolio (and to, thereby, acknowledge the very real trade-offs in the fidelity of Pralana's projections).

It must be acknowledged - and I am confident anyone reading this would agree - that it is so easy to toss out these (hopefully) reasonable-sounding proposals and to conclude that the only thing standing between idea and implementation is Charlie's time. @cstone, we understand completely that there are 1,000+ complications for any suggested enhancement and that you have myriad factors to consider when making the simplest change to Pralana and its sophisticated algorithms. I believe @jkandell would agree that we are throwing out these ideas for potentially making Pralana even more powerful and to make our planning even more true-to-life and trustworthy. But most, if not all, of us out here have a deep appreciation for the complexity involved in such proposals. The last thing we would want is for our "helpful suggestions" to result in a dash for the hills and the 2-member Pralana team suddenly reverting back to 1-member. 🙂 These particular suggestions would seem to overlap greatly with your currently-underway efforts with account rebalancing and the substantial recursive, same-year taxation effort. This can wait (but it would be absolutely wonderful someday as LTCG taxation is kind of an underserved area in Pralana today while a major consideration in real life.)

[Edit: Tried to clean up some formatting - the Editor version doesn't match the Display version]

It is gratifying to see so much thoughtful discussion on these and other topics. These discussions help me immensely.

I do not pretend to comprehend all of the discussion points in these most recent posts.

Thinking about @jkandell: "Pralana allows multiple accounts under Taxable....These might literally be separate accounts or they might be mental buckets..."

I vote for mental buckets! or some other way to achieve this goal.

Adding support for a user-defined variable number (2, 5, 12?) of Taxable Investment Accounts introduces a large amount of additional complexity in many areas, like:

• Tabular projections that have columns for each account. And account statements.
• Withdrawal priority and handling unscheduled withdrawals from perhaps multiple taxable accounts.
• Optimize Withdrawal Priority: Currently, Pralana evaluates withdrawal order permutations for 4 accounts (taxable, IRA1, IRA2, roth) = 4 factorial (24) permutations of withdrawal priority (plus proportional) = 25 per time period for 2 time periods is 25^2 = 625 permutations (full scenario recalcs).

  Add 2 more accounts and its 6 factorial = 720 per time period, add 1 for proportional and its: 721^2 = 519,841 full scenario recalcs

  This is why the inherited IRAs, HSA and 529 are not included in the withdrawal priority optimization permutations.

• Mode 2 asset allocation algorithm.
• Supporting scheduled withdrawals from a user-defined number of source or destination accounts.
• More complexity = more calcs = slower Monte Carlo and Historical Analyses. Some of the optimizations run multiple, iterative Monte Carlo/Historical Analyses.

Given that there are about 400 open feature requests and usability improvements, the hard reality is that the reengineering required to support multiple Taxable Investment accounts is not likely to happen in the foreseeable future.

Thinking about @ilovemybeagles: LTCG Taxation - Portfolio Time-Period Approach

Enhancing the LTCG Withdrawal Strategy along these lines is certainly more do-able. There is a placeholder for this on the Feature Voting page.

Alas, it is Aug 2025 and Pralana doesn't yet have a meaningful Home page or Dashboard....and 'beautification' was to be the priority for this year.

@cstone I appreciate the impossibility of the permutations of 10 accounts with 20 time spans with 5 LTCG options. 🙂

If I can press a little, given at least one of the three current strategies for taxable withdrawal choices are invalid, so something has to be done.

First, I note that multiple (unrestricted?) taxable accounts already exist in Pralana, and the 3 first/last/even taxation choices are already there. So there's already a lot of potential permutations. The main increase in permutations are the time spans.

I wonder if my proposal with some heavy restrictions like:

• Only two accounts, namely users have 1 "regular" taxable account and 1 optional "special purpose" second account to be used for e.g. postponing withdrawal for heir, or for "gathering" tax loss harvest to be used later in scheduled withdrawal,
• restricting that second account to at most 1 specific time spans, and
• a 3 choices of priority in withdrawing from that second account (only)between "Draw from this first", "Draw from this last", "draw evenly between this and main taxable accoutn"

Would that be workable? That's 1 X 2 X 3 permutations on the second account, but with a lot more power than users have now, I think.

Moreover each account gets taxed at average cost (i.e. Unrealized/Total--what is now called "even") as a simplification over what's there now, since it eliminates the two other choices? The big change is user setting the priority the second account is drawn from rather than how its taxed.

I am not quite sure how the proposal people are voting on would work. Do you think that's a more workable solution given the realities?

@cstone Well color me as red as a beet. The feature you referenced on the Feature Voting page would seem to be the equivalent of “my” proposed enhancement above. My apologies to the original requestor.

I don’t know how I missed this. I even voted for that feature! Granted, I am not exactly sure of the details of what that feature request entails. But I voted for it nonetheless since I figured any enhancement better supporting realized LTCG income would be an improvement over the current offering where “First” and “Last” can never occur in the real world (no offense). [Edit: Too harsh. I am concerned only with stock/bond funds. There may very well be other types of Taxable assets where “First”/“Last” might be applicable. But we, alas, cannot choose to realize our gains “First” or “Last” when selling stocks/bonds.]

Thanks for considering our many supplications. With 400 open requests your plate is full indeed.

@jkandell, @ilovemybeagles and all:

On the Initial Account Balances pages, all the account types allow up to 10 of your own 'user' accounts, except inherited IRAs which allow 1 each. As you know, within each of the 11 Pralana account types, the inputs for the 'user' accounts are aggregated and they are otherwise not used.

In the short to intermediate term, we will stick with 11 account types and find a way to allow maximum flexibility for the LTCG strategy improvements, starting with:

• realizing LTCG on annual rebalancing in the taxable account
  • Part 1: allocate initial unrealized LTCG per asset class using the asset allocation
  • Part 2: create a new input form for user to specify the distribution of the initial unrealized LTCG by asset class.
• Enhanced LTCG Withdrawal Strategy per that Feature Voting item

"Enhanced LTCG Withdrawal Strategy": Below is the original feature request. Others have made similar requests. I have not started on this yet and so the details are TBD.

LTCG Withdrawal Strategy
This setting has a surprisingly large impact on the modeling -- both short-term (e.g., near year taxes) and long-term (e.g., legacy balances and embedded gains).

None of the three settings is ideal:

- gains last is the ideal goal but infeasible given gains are linked to basis at the lot level, and especially as lots age and accrue embedded gains
- gains evenly is too conservative and wouldn't reflect the use of any tax smart withdrawal strategy
- gains first would usually only make sense with a purposeful gain harvesting strategy for a specific period of time

Could we use a more granular percentage approach and have the ability to adjust the % in different time periods?

@jkandell asked: "I am not quite sure how the proposal people are voting on would work. Do you think that's a more workable solution given the realities?"
Charlie: Yes, though I do not completely understand what everyone is trying to achieve. Let's try to find a solution that does not require:

• additional Pralana accounts
• tracking lots, # shares, price per share by asset class
• introduction of short term capital gains into the mix.

Also, keep in mind that someday 'soon' Pralana Online will calculate asset income (interest and divs) as a % of account/asset balance instead of % of annual growth/appreciation.

I appreciate the discussion, but none of it appears to address my situation.

In my taxable account I have (in the order in which I want to withdraw them): 1. Municipal bonds, 2. Moderately appreciated stock, 3. Highly appreciated stock. However, I can't find any way to model this withdrawal order in Pralana online!

In my case, withdrawing the municipal bonds first (and replacing them with taxable bonds in the tIRA) is key. To model this, I have resorted to a rather ugly workaround: I observe the withdrawals (net of interest and dividends) in each of the next several years, and then I define four one-year time periods in which I gradually reduce the target allocation of municipal bonds (all intaxable) to correspond to how much I expect to have left after these withdrawals. But this has SOO many problems still, including that it needs iterative refinement, it is not robust to changes in rate-of-return assumptions (for any or all of the asset categories) or to total portfolio growth, and it doesn't play nice with Roth optimization (which is my main reason for using Pralana in the first place).

I would love to be able define my withdrawal strategy for Taxable by prioritizing asset classes. I would be happy with "average" LTCG assumptions. I'm also willing to forego distinguishing between moderately vs highly appreciated stock since I could (I think) define a new asset class, AppreciatedStock, and prioritize it "last".

P.S. As I understand it, this will still work with Monte Carlo analysis; and since I defined a Muni asset class I can't use Historical Analysis anyway; if anyone knows a way I can use Historical (by defining Muni and AppreciatedStock as "behaving like" bonds and stocks respectively) please let me know.