Help creating a poor-man's I-ORP

[SecondCor521]

Hi all,

I have a spreadsheet in which I attempt to predict and evaluate Roth conversions, RMDs, and IRMAA taxes.

Currently I have a spreadsheet which:

1. Goes from my current age (50) to 85 (end of my planning horizon).
2. Allows me to set my Roth conversions to anything I want each year.
3. Calculates my RMD from age 70.5 onward
4. Calculates my SS from age 70 onward.
5. Calculates my "rough" taxable income as "Conversions + RMD + 85% SS - Standard Deduction". (1, 2)
6. Calculates federal taxes and IRMAA taxes on my "rough" taxable income.

I don't use Excel, I use LibreOffice.

I'd like to ask LibreOffice's solver to tell me what the "best" series of Roth conversions are each year from 50 to 85 in order to minimize taxes. Basically, a poor man's I-ORP.

I have used I-ORP, and understand the principles on which it is based. But I can't see inside the calculations and so would like to replicate the results myself, however crudely.

I'm running into two problems.

First, I don't know how to define the goal properly. If I try to minimize taxes, then the solution is to obviously not Roth convert at all and do minimum RMD's. If I maximize spending, then the solution is to Roth convert a lot every year. If I maximize after tax spending, then I can do that but then I am ignoring the age 85 remaining tIRA balance, which it feels like I should account for somehow.

Currently what seems to be working somewhat is to calculate my effective tax rate each year as (income tax + IRMAA tax) / "rough" taxable income. I can then average those over the entire 50 to 85 period. Then I play around and try to minimize that number. So far the lowest I can get that is a bit under 20%, which seems pretty good.

Second problem is that if I try to use LibreOffice to solve for it the way I want to, it claims my model isn't linear and quits. I've googled that and the remedy I saw there isn't working for me.

Any thoughts or advice on either of these two problems would be great.

Thanks.

...

This year I think I'm probably going to kick the can down the road. I have a kid in an expensive college and he's getting financial aid, so I have a FAFSA cliff at an AGI of $49,999 for this year and next that it would be nice to be under.

So I think this year I'm Roth converting up to there, even though that leaves some non-refundable credits on the table. Some preliminary analysis suggests that this cliff isn't even that bad, especially if I go up to an AGI of $60K or $70K or so (it's effectively a hump, so once over it I can spread the pain out by converting more).

But if I just go to the FAFSA cliff, then I can ignore all this other analysis which honestly makes my head hurt trying to optimize.

...

Footnotes:

1. So I always assume 85% of my SS will be taxed; I know it could be less than this but it probably won't be in my case no matter what I do, and the extra complexity isn't worth addressing IMHO.

2. I say "rough" taxable income because of the flaw mentioned in footnote 1 plus the fact that I'll probably have other income at that point, and may have other deductions, etc.

[SevenUp]

The usual optimization statement goes something like "Maximize _____ by adjusting _______, subject to ________ constraints on _______."

If one is no longer eligible to make IRA contributions, I-ORP uses "Maximize disposable income by adjusting withdrawals and conversions, subject to non-negativity constraints on annual account balances and a minimum for total final assets."

Another common approach for retirement optimization is "Maximize total final assets by adjusting withdrawals and conversions, subject to non-negativity constraints on annual account balances and a specified spending schedule."


As for solving the optimization problem, I-ORP simplifies some things (e.g., taxation of social security and capital gains) so it can use Linear programming to get a fast global optimum solution.

If one uses the actual tax code (let alone introducing other variables such as when to start taking SS or a deferrable pension), it becomes a Nonlinear programming problem and getting a global optimum may be difficult and/or time consuming.

[SecondCor521]

Thanks for the reply, SevenUp.

I realized after I posted that I was being kind of lame and lazy. Hey, world, here's my problem, help me solve it for free! Blech. I'm going to blame it on having my screening colonoscopy earlier today and not being 100%.

I think I need to noodle on what exactly I want to maximize. I know vaguely that I want to minimize overall taxes paid, and I'd like to maximize disposable income. I value "evenness" of spending, but how to define that? That's probably a constraint, but a loose one. I also haven't accounted for TVM. Finally, having some left over is fine and nice, but not sure how much to weigh that. So I have to figure out which of those I want to maximize, and if more than one, to what degree.

I'm familiar with linear programming. Embarrassingly enough, I took and got a decent grade in a class on this sort of optimization as an undergrad at an Ivy League university, but don't remember any of it and couldn't solve a basic problem even if it was multiple choice and I knew the answer was either B or C.

I guess I wonder what exactly about my spreadsheet setup makes Excel think it's non-linear, but I'm not that motivated enough to try to figure that out until I know the answer to the first part of my problem above.

...

In the grand scheme of things, at this point I have a net WR of around 28 basis points, so anything I do that is halfway reasonable will work. I'm just still somewhat motivated to run up the score for my kids, even though they probably wouldn't know one way or the other. And them having more money than they already will have is possibly a negative.

[pb4uski]

I think you would want to optimize age 85 ending balances, perhaps with tax-deferred balances adjusted for taxes if all taken out at age 85.

Alternatively, you could set spending high enough that there was minimal ending balances at age 85 and then optimize ending balances.

It seems to me that if you optimize ending balances then by default you optimize taxes all else being equal.

[SevenUp]

Quote:

Originally Posted by SecondCor521 I know vaguely that I want to minimize overall taxes paid....

Perhaps not. E.g., take an investment that will double in 10 years. It is currently held in a traditional account. Marginal rate to convert it to Roth now is 28% and expected marginal rate to convert to Roth in 10 years is 15%. Do you want to minimize tax paid, or maximize amount remaining after tax?

[SecondCor521]

Thanks again for the responses.

Quote:

Originally Posted by pb4uski I think you would want to optimize age 85 ending balances, perhaps with tax-deferred balances adjusted for taxes if all taken out at age 85. Alternatively, you could set spending high enough that there was minimal ending balances at age 85 and then optimize ending balances. It seems to me that if you optimize ending balances then by default you optimize taxes all else being equal.

After reading the above I think what I might do is set a low ending balance for my tIRA (like $100K) and then try to maximize after-tax spending subject to that $100K constraint. In other words, what is the most tax efficient way to get all my tIRA money taxed between now and 85. I probably won't pull that much out, but at least it seems like it would be instructive.

Quote:

Originally Posted by SevenUp Perhaps not. E.g., take an investment that will double in 10 years. It is currently held in a traditional account. Marginal rate to convert it to Roth now is 28% and expected marginal rate to convert to Roth in 10 years is 15%. Do you want to minimize tax paid, or maximize amount remaining after tax?

I want to maximize amount remaining after tax. In this particular subset of my financial life, for every dollar in my IRA I can either (a) leave it in the IRA untouched, (b) take it out and send it to Uncle Sam, or (c) take it out and spend it. Depending on when and how much affects the ratio between (b) and (c), but any dollar taken out will be either (b) or (c), so in that sense I view it as a zero sum game there. If I constrain (a) to $100K as mentioned above, then minimizing (b) will, by definition, maximize (c).

...

I'll probably apply some sort of TVM to my spreadsheet, as I do somewhat value money in my 50's and 60's more than in my 70's and 80's.

And I think I'll fiddle with stuff manually so I (a) can keep withdrawals even to a certain degree, and (b) get a better gut feel for the trade-offs.

...

I am hopeful that I am on a wide, flat area of the optimization curve / n-dimensional polygon / whatever. I did have a hard time getting my average tax rate below about 19.8% when I was playing around before, and any changes didn't raise it much - maybe to 21% or something - so there is a little bit of evidence for that. I think that's already past the point where it matters. It is also already past the point where I can plan my life to spend exactly $xx,yyy.zz a year.

...

It does seem that with moderately aggressive conversions that I can stay out of the 3x% federal income tax rates. If I pay 24% or less then I'm OK with that. And hitting or trying to avoid hitting the third IRMAA tier when I am 79 is definitely a first world problem.

Thanks again.

[SevenUp]

Quote:

Originally Posted by SecondCor521 In this particular subset of my financial life, for every dollar in my IRA I can either (a) leave it in the IRA untouched, (b) take it out and send it to Uncle Sam, or (c) take it out and spend it.

Interesting - that's different from the usual situation in which how much "(d) convert it to Roth" gets all the attention.

[Animorph]

I wrote my own computer program to optimize my withdrawals and Roth conversions. No, it is not available to others.

I am optimizing after-tax spending, mostly with a fixed final year net worth in order to provide a fair comparison.

I assume I will spend down taxable accounts first, until they're empty. Traditional IRA withdrawals are essentially my one variable for each year of retirement. If the tIRA withdrawal is not needed for expenses I assume the excess is Roth converted. I assume Roth accounts will be used to meet expenses when SS and tIRA withdrawals are insufficient.

I have a spending budget for each year that remains fixed (adjusted for assumed inflation). I then add an "additional spending" amount, a single value that is inflation adjusted each year. That number is adjusted by the program to hit the final net worth target.

The program optimizes each year's tIRA withdrawal to maximize the additional spending amount. It calculates a step in each year's withdrawal that should increase the additional spending number. The steps start out fairly large and then decrease in size as find an optimum value. I may be simultaneously optimizing 50-60 years of withdrawals, so this is a complex and delicate process.

I think this will be too complex for a spreadsheet, but additional assumptions might simplify the calculations significantly. For example, you could try just a few different possibilities to optimize each year, like filling a tax bracket, hitting $250k AGI, just avoiding AMT, or avoiding the ACA subsidy cliff. When I evaluate my results I mostly see that it is filling a tax bracket or hitting some other tax threshold (federal or state).

A spreadsheet that allowed you to specify tIRA withdrawals, spend a certain after-tax amount each year, and calculate a final net worth would allow you to optimize the yearly withdrawals manually. I have found that once you are anywhere close to optimum there is usually little to be gained after that. While I like my program, it is severe overkill.