FREE NEWSLETTER
EVERY ARTICLE ABOUT Roth conversions says the same thing: pay the tax from taxable money, not from the IRA.
That is good advice if you have taxable money.
Plenty of retirees do not. Their savings sit almost entirely in a traditional IRA, built from years of 401(k) contributions and a rollover at retirement, with little brokerage money and no cash reserve worth naming.
For them, “pay from outside money” is not advice. It is a condition they do not meet. Their real choice is a self-funded conversion or no conversion at all.
The familiar warning is that self-funding requires a gross-up. You withdraw money to pay the conversion tax, that withdrawal is itself taxable, so you have to withdraw a little more.
Less discussed is what else that withdrawal sets off. It makes more of your Social Security taxable, it raises your Medicare premiums two years later, and in a high-tax state it enlarges the amount that has to leave the IRA. Here is what those three cost, on one household.
A cautious conversion
Meet Dianne, a composite rather than a real person. She is 66, single, retired to Florida, with $1 million in a traditional IRA and nothing outside it. Social Security pays her $28,000 a year.
She is past 59 and a half and already on Medicare. Florida keeps state income tax out of the arithmetic for now.
She converts $40,000. Careful, modest, the size people choose when they are trying not to do anything dramatic.
The tax on that conversion, paid from a checking account she does not have, would be $4,234. Funding it from the IRA instead, the withdrawal that covers it is $5,125, and her tax rises to match.
That extra $891 is 17.4% of the $5,125 she withdrew. Her ordinary-income bracket is 12%.
Twelve of those points are the tax on the withdrawal itself, which is the gross-up everybody expects. The other 5.4 points come from $2,300 of her Social Security being pulled into taxable income by that same withdrawal.
Paying from cash, $21,500 of her benefit would be taxable. Self-funding, $23,800 is. The difference is caused by how she paid, not by what she converted, and nothing on her return will label it.
Not a smaller version of a large one
The effect is not linear. Run the same woman at $150,000 and 85% of her benefit, the statutory maximum, is already taxable before she funds the tax. The funding withdrawal drags in nothing further, and this particular cost is zero.
That is not an argument for converting more. It is an argument against assuming a small conversion is simply a smaller version of a large one. The one that looks cautious can carry the higher marginal rate.
What self-funding actually costs
Give Dianne the $150,000 conversion, still with no outside cash, and add the state question.
Still in Florida, $195,469 has to leave the traditional IRA to put $150,000 into the Roth. Move her to California, changing nothing else, and it is $219,751. That is $1.30 of IRA spent for every dollar reaching the Roth, against $1.47, and the gap is California income tax compounded through the gross-up. If a move across state lines is anywhere in your plan, the order of operations may matter more than the size of the conversion.
But $1.47 invites a conclusion it does not support, and I would rather correct that myself than let it travel. A conversion is taxable whichever pocket pays.
Of the $69,751 Dianne withdraws in California, roughly $44,639 replaces cash she would have spent anyway. The incremental cost of self-funding, in wealth given up, is $25,111. About 17% of the conversion, not 47.
The bill that arrives in 2028
At that $150,000 conversion, self-funding also costs Dianne $1,735 in higher Medicare premiums, a surcharge of $6,355 rather than $4,620, charged on top of the standard premium, for one year, and hers alone as a single filer.
The bill arrives two years late. Medicare sets her 2028 premiums from her 2026 income, so the cost is invisible at the moment she is deciding how to pay the 2026 tax.
When it can still make sense
Doing nothing is not free either. Money left in a traditional IRA comes out eventually, under required distributions, at whatever rates apply then, possibly to a survivor filing single, possibly to heirs facing a 10-year deadline.
Where self-funding still holds up, a few things tend to be true.
The rate gap is durable rather than a one-year accident. Whether it repays a cost this size depends on time, future tax rates and investment returns. A temporary dip in income is a thin foundation. A structural window, after retirement and before Social Security or required distributions begin, is more compelling.
There is runway, because shrinking the portfolio to change its tax character needs years of tax-free growth to earn back.
Age 59 and a half is cleanly behind you. The converted amount is not subject to the 10% additional tax on early distributions. A separate distribution taken to pay the tax generally is, unless an exception applies.
And enough is left afterward. Check the balance after the funding withdrawal, not after the conversion. Dianne’s California IRA drops to $780,249, and whether that funds the next 30 years matters more than whether the conversion was tax-efficient.
A practical warning about withholding
Withholding from the conversion is not a cheaper way to pay the tax. It reduces what reaches the Roth.
Elect 24% on Dianne’s $40,000 conversion, as this illustration does rather than as any custodian requires, and $9,600 goes to the IRS while $30,400 lands in the Roth, against $4,234 actually owed. The cash comes back next spring as a refund. The Roth room does not come back at all.
Restating the rule
Pay the conversion tax from outside cash if you have it. That remains the best answer.
It is not an answer for the retiree whose savings are almost entirely in a traditional IRA. For that person, “never pay from the IRA” skips the actual decision, which is whether a self-funded conversion, with its full marginal cost, beats leaving the money where it is.
Before deciding, count the gross-up, the Social Security effect, state tax and the Medicare bill two years later. Sometimes that arithmetic still says no. It beats applying a rule written for somebody with a different balance sheet.
________________________________________________________________________________
John Urban is the founder of RetireSmartIRA, a retirement tax-planning app. Earlier, he founded GT Nexus, a supply-chain software company acquired by Infor in 2015. He lives in Northern California with his wife, Kathy, and enjoys time with family, travel, reading, Bay Area sports, and the occasional deep dive into the fine print of the tax code.
I enjoyed reading your informative Humble Dollar article John. Thank you for posting.
After reading your article and the many comments I was thinking back upon my final working years as a tax preparer and the problems clients were having regarding the rule changes on inherited IRAs. Instead of writing a comment or forum post on Humble Dollar I happily found your excellent article titled “The Inherited IRA 10-Year Rule: RMDs and Tax Impact in 2026” on your website. In that article you addressed many changes from the first SECURE Act and how it impacts those who have or will inherit either a traditional or Roth IRA. Another thank you, I recommend reading your article.
I did have a few other thoughts on these topics. A short summary of my thoughts of other matters to consider follows-
Many spousal beneficiaries inheriting a IRA may be past their own full retirement age. Unfortunately sometimes a knowledgeable preparer or advisor are first engaged past the time that RMDs should have been distributed and sometimes where RMDs have not been taken for multiple years. The preparer has now been recommend to the client as a result of a letter from the IRS addressed to the widow(er) years after the death of a spouse. Plan on a lot of time to resolve the issues, have necessary distributions made, file all needed amended returns and request abatement of the penalty(s) with the appropriate form and expect to include a long statement of the reasonable cause to reduce the penalty from the 25% level to the 10% level or even a complete abatement.
When thinking about the ten year requirement planning for non-spouse beneficiaries to empty the inherited IRA I have seen still working beneficiaries with earned income who had room to increase 401(k) and/or IRA contributions and then schedule out receiving inherited IRA distributions while simultaneously increasing their contributions for ten years thus effectively extending the inherited deferrals in their own retirement accounts for the beneficiaries life plus 10 years while keeping their cash flow and taxable income approximately the same.
Getting your primary and contingent beneficiary designations as you want them is crucial to achieve your goals for passing your IRAs as you intend them to be distributed. I believe this should be a formal every year best practice and again every time there is a major life event.
If there are multiple beneficiaries of your IRA the beneficiaries need to under the requirements as explained in IRS Pub 590-B related to separate accounts –
“A single IRA can be split into separate accounts or shares for each beneficiary. These separate accounts or shares can be established at any
time, either before or after the owner’s required beginning date. Generally, these separate accounts or shares are combined for purposes of determining the required minimum distribution. However, these separate accounts or shares won’t be combined for required minimum distribution purposes after the death of the IRA owner if the separate accounts or shares are established by the end of the year following the year of the IRA owner’s death.”
If the decedent was still working past age 70.5 and still contributing to a traditional IRA there are potential limitations on QCD’s.
Per IRS Pub 590-B(2025)-
Offset of QCDs by amounts contributed after age701/2.
Beginning in tax years after December 31, 2019, the amount of QCDs that you can exclude from income is reduced by the excess of the aggregate amount of IRA contributions you deducted for the taxable year and any prior year that you were age 701/2 or older over the amount of the IRA contributions that were used to reduce the excludable amount of QCDs in all earlier years.
Again, thanks John for your article.
Best, Bill
Bill,
Thank you for reading, and for the generous words about the inherited IRA piece.
The still-working beneficiary is the case I keep coming back to. Pairing inherited distributions against increased 401(k) deferrals turns a ten year forced drawdown into something much closer to a rollover, with cash flow and taxable income barely moving. The binding constraint is deferral capacity. Someone inheriting a large IRA can rarely absorb a tenth of it under the elective deferral limit each year, so it works best on mid-size accounts where the beneficiary has real room. Where it does fit, the ten year rule stops being a tax event and becomes a timing exercise.
Your point on separate accounts deserves more attention than it gets. That deadline passes quietly and nobody sends a reminder.
Thanks for the article reference, Bill. The website seems loaded with information.
A question for HD readers about advisability of Roth conversions. Most advice revolves around future tax rates, but I’ve never seen advice based on the actual deferral. Suppose you were in the Fed 33% marginal tax rate in 2015, deferred the 401k max ( 59k which includes employer & catch up) then convert this to Roth in 2026. If it’s done anywhere below a 33% Fed tax rate, isn’t this a win regardless of future tax rates? It seems like that never is mentioned as a reason to go ahead and convert. Of course, if future rates are lower that’s even better, but that doesn’t negate the benefits of delaying taxes for 11 years while compounding the amount. And it’s not based on future unknowns but actual data from previous transactions. Just wondering if I’m missing something in this analysis?
David, I’ve written about this before, and provided some simple examples to demonstrate that the primary driver in the decision is the difference in tax rates between initial investment and the tax rate at withdrawal. If the tax rates are the same, then the amount available to withdrawal is the same for a Trad or Roth IRA. If the rate at initial investment is 32%, and the rate at withdrawal is less, than the Trad account wins. It doesn’t matter if you then convert it to an IRA or spend it.
For example, say you invest $100,000 in a Trad IRA and over a period of time it doubles to $200,000. If the marginal rate at withdrawal is 32%, then you net $200,000 x (1-0.32) =$136,000.00.
If the marginal rate at initial investment is also 32%, then the initial investment is $100,000 x (1-0.32) =$68,000.00. If it is invested in the same manner as the Trad account and doubles, you net 2 x $68,000 = $136,000.
If, however, the marginal rate at withdrawal is lower than initially, say 22%, then the Trad account wins. The amount at withdrawal is $200,000 x (1-.22) =$156,000.00. As you say, you still win whether you convert this to a Roth or spend it. Once you withdrawal from the Trad account, there is not future tax event for that specific block of money.
I’ve ignored some potential complications like holding period and 10% penalty for early withdrawal, but I believe the concept is sound. The rub is predicting what your future marginal rate will be. That’s one reason it makes sense for young people who are at low to very low marginal rates to consider a Roth over a Trad account.
I wish I had done some conversions 10 or 15 years ago, in the period I was coasting toward retirement and let my income decline. But there is an age you reach at which you can no longer expect the Roth tax free benefits will offset the tax payments and other cost impacts from conversion. IMHO, I am now well past that age, so I’ll live with what I’ve got and live with the tax implications. It was a fairly simple calculation to reach that decision.
The only other consideration is if you want to potentially leave a portion of your portfolio tax free to your children by you paying the taxes up front for them by doing the conversion. That was one of my considerations in us converting.
Same here. We dont want our kids getting stuck with those taxes during their highest earning years.
Excellent article and information. I find no easy resolution for someone like me at 80 years old. It just seems too tax costly to take the leap to say sell $600,000 of stock for IRA conversion, every items increases, IRMMA, higher tax bracket, etc, I just cannot make it work. Anyone have advice, please advise. Thank you.
Roth IRA for you seems like a solution looking for a problem! You are blessed. Enjoy your fortune as is, I say.
From what I’ve read, Roth conversions very rarely make sense for an 80 year old. The main benefit of doing it is long term tax free growth, and at that point the time horizon is likely short and RMDs are well under way, so the benefit really isn’t there. If I were in your place, I’d take Roth conversions off my list of things to think about.
Another important cost of Roth conversions that is rarely addressed is the opportunity cost of the money used to pay the conversion tax.
If someone had $50,000 of taxable money available to pay the tax, that $50,000 isn’t just “spent” — it’s $50,000 that can no longer remain invested and compound. The comparison shouldn’t only be today’s tax cost versus the potential future tax savings. It should also consider what that tax money could have grown to, on an after-tax basis, if it had remained invested.
I’d be interested in seeing a Roth-conversion analysis that includes the opportunity cost of the tax payment itself, along with the future tax savings.
Magoo, this is the objection I hear most often, and it’s the one that changed the most for me when I worked through the arithmetic.
The money paid from a taxable account isn’t consumed. It’s relocated. Convert $200,000 at a 25% rate and pay the $50,000 from savings: before, you held a $200,000 traditional IRA carrying $50,000 of deferred tax, plus $50,000 in taxable. After, you hold $200,000 in a Roth and nothing in taxable. Your after-tax net worth is identical. What changed is that all of it now sits in a tax-free account instead of three quarters of it.
So the $50,000 keeps compounding. It just compounds inside the Roth, where it’s no longer paying tax on dividends and rebalancing along the way.
That drag is the actual gain from paying with outside money, and it begins immediately rather than requiring a long holding period.
The real costs are elsewhere, and they’re worth watching. If you raise the $50,000 by selling appreciated shares, the capital gains tax on that sale is a genuine cost, and a reason an older investor might prefer to wait for the step-up. If you pay the tax from inside the IRA instead, you really have shrunk the invested balance, which is what the article is about. And if your rate at withdrawal turns out lower than the rate you converted at, you paid too much for the shelter no matter where the money came from.
William Wang, a law professor who writes on this, has a framing I’ve found clarifying: a traditional IRA is a joint venture in which the government owns a share equal to your tax rate, so the conversion tax isn’t really a cost, it’s the price of buying out that share. Once you see it that way, “the money stops compounding” becomes “the money bought something.”
Professor Wang’s is a smart way to look at it.
Putting it that way, I see it now. I was always thinking of it as holding onto the IRS’s share, $50K tax money, and allowing it to grow in my brokerage account. However, I wasn’t accounting for the fact that the government’s 25% claim on the Traditional IRA grows right along with the IRA. If both grow at the same rate, the two effects offset each other and it’s a wash.
I actually ran the numbers using your $200K/$50K example, assuming an 8% return for 10 years, and the math comes out exactly as you described. Thank you for taking the time to explain—it finally clicked for me!!
Or, for instance, using the same taxable money to bridge to a later social security start date. Lots of moving parts. I’m going to do some serious what-if scenarios in Boldin, which does a decent job of showing how various changes interact.
I recently read a book called The Roth Conversion Formula by Alex O’Shea. The book provides 10 case studies using 2026 tax law. It was an easy and understandable read.
Has anyone experimented with the author’s RetireSmartIRA and compared it to other programs? I went to the website and it looks interesting. As with any source, how do you know the information is accurate in the details it returns?
I enjoyed the author’s article, although nebulous in some parts and will require reading several times over. His other articles have also been good to read.
Olin, that is a fair question. I have an obvious interest in the specific answer, so let me answer the general one.
I would not trust any planning tool just because it produces a detailed number. The important question is whether you can see enough of the work to understand what drove that number.
For me, there are a few things worth looking for:
I have not read the O’Shea book, so I cannot offer a comparison. But those are the standards I would use for any retirement-planning tool.
And thank you for the “nebulous.” I know my writing can get dense when I am deep in a topic. I will try to do better.
Best regards,
John
John, you raise many good additional thoughts to consider. Thanks!
I believe the reasoning is thorough, and the article brings out various aspects involved in the consideration of conversion. It also illustrates that results vary depending upon “starting conditions.” I did not check each number and don’t think that is necessary.
I also checked out the Boldin site suggested below. Once again, and as expected, results vary depending upon how you set it up.
I would start with a simpler analysis tailored to a limited problem statement with fewer variables. That gives a better handle on the impact of each variable and project tax consequence.
V Saraf, if you experiment with both Boldin and the author’s app, please share your thoughts if you don’t mind.
Olin, thanks for asking.
John Urban drilled down above, in his response to you, to a large extent. Hope that helped.
As you know, these apps are basically tools. One has to consider his finances, goals, expectations and intangibles such as longevity and change the inputs to the model. More you specify, more the answer gets “kind of clear” in number but more vague in how it got there, unless you go into its mechanics of the calculation. It rolls in author’s or the model’s assumptions unless you change them.
Simpler, in my opinion, is to make basic calculations yourself on a spreadsheet, and experiment with one variable at a time on a reasonable setup that matches your finances and plans. There, you get a sense of direction in which you want to move. John, in his article, elaborated well, aspects one should consider in feeding the app and getting to a decision.
Hope this helps Olin. Sorry, if it doesn’t. I had to subscribe to Boldin so that I could see what it does, and asked for cancellation soon after. Not that it is not good; I just don’t see a need to use app for myself.
Understood. Thanks for elaborating.
Great analysis but, as I think has been pointed out somewhere previously, let’s not forget about the survivor’s tax penalty and the downstream tax advantages for heirs that also come into play beyond just one’s current tax situation for MFJ esp w children.
Excellent point. Sooner or later, every couple runs into this. I am very thankful that I gradually did Roth conversions to move from 90% Tira to about 12%. As it is, I will still pay the first tier premium for IRMAA without any further Roth conversions.
I do not think people consider the reality of going from MFJ to S enough in their tax planning.
Mark, you are right to raise both. The article prices what self-funding costs and stops there, and the two things you name sit on the other side of that ledger.
On the survivor penalty I wrote a piece here in July, and the part worth knowing is that it hits middle-income couples hardest, not affluent ones. On heirs, the reason children matter is timing: under the 10-year rule an inherited traditional IRA usually comes out during their peak earning years, so the rate that applies is theirs, not yours.
Thanks for your comment,
John
When I considered doing Roth conversions my reasoning was this not necessarily in this order:
1) I wanted to convert all of my wife’s smaller (1/3) of our portfolio so I only had to deal with RMDs from one account for simplicity’s sake.
2) Decreasing the amount of our RMDs for tax purposes considering that our combined Social Security payments will cover the vast majority of our total spending, not just our fixed costs. As a result if the conversion were not performed when RMDs started we would be forced to withdrawl even more than we would spend.
3) Because of the last reason the money which is converted may never need to be spent. If my wife lives to 100 (women two recent generations lived past 103) with compounding for 35 years this balance could be immense.
4) If the above comes to fruition and my children do not withdraw any money for the 10 years currently allowed by the IRS that would result in 45 years of TAX FREE compounding.
5) We could pay the income taxes at the 12% maximum tax bracket out of inherited money in a taxable account, which is basically free money.
6) We live in NH so no income taxes
I did not even think of the widows’ tax, an extra benefit.
John, thanks for an interesting article. The tax interactions are complex and challenging for many of us. I wondered how Dianne’s modest conversion might impact the $6,000 additional senior deduction. Using Dinkytown’s 1040 calculator I found that she could increase her IRA withdrawal to $51,200 and still get the full $6,000 deduction. At that point her taxable income is $400 over the limit for the 12% bracket (into the 22% bracket). So a $50,000 IRA withdrawal would keep her in the 12% bracket, remain eligible for the senior deduction, avoid 2028 IRMAA, and cause a total tax bill of $5,713, leaving $44.287 to convert to her Roth.
One final thought that s beyond the scope of this article. This leaves her $28,000 to live on – equal to her total pre-tax SS benefits. I think Dianne would be a good candidate for a comprehensive retirement analysis, including understanding her situation and goals for retirement and legacy bequests. The Roth conversion might make sense, but it is not obvious without further information and analysis.
Rick, thank you. I ran your numbers through the same projection engine I used for the article, and they hold up.
A $44,287 conversion, funded from the IRA, requires about $5,710 more from the IRA to pay the tax. Total IRA outflow is essentially $50,000, just as you calculated.
The $51,200 number is worth noting. Both the conversion and the withdrawal used to pay its tax count as ordinary income. The number to watch is total IRA dollars leaving the account, not just the conversion amount. At $51,200 of IRA outflow plus $23,800 of taxable Social Security, Dianne lands at the $75,000 threshold for the full senior deduction. Past that point the $6,000 falls six cents for every dollar of income, so a dollar over the line is taxed as $1.06, which turns her 12% into roughly 12.7%.
At $50,000 she stays inside the 12% bracket, and IRMAA is not in play at any of these levels.
Your larger point is equally important. I gave Dianne no spending needs so the article would focus attention on the mechanics of a self-funded conversion. That makes the tax effect easier to see, but it does not tell us whether she should convert. A 66-year-old living on $28,000 of Social Security while taking roughly $50,000 from her IRA needs the broader retirement analysis you describe: spending, longevity, portfolio risk, future RMDs, and any legacy goal.
The conversion might still make sense. But it is certainly not obvious from the tax calculation alone, especially when a Roth contribution during working years may have avoided the self-funding issue in the first place.
John
John, I love your posts. Thanks for illustrating the importance of doing a comprehensive look at the math before jumping into conversion mania. It would be interesting to analyze the effects of funding Roth accounts during one’s working years, versus doing conversions after age 59.5. I suspect some would have been better off doing the former.
Dan, thank you. I think you’re right, and for one group I would put it more strongly than “suspect.”
A Roth contribution while you are working pays its tax from your paycheck, which is outside money by definition. No gross-up. No additional Social Security taxation. No Medicare surcharge two years later. The mechanism in the article never starts.
Whether that beats converting later still turns on the rate you paid, the years left to compound, whether a low-income window opens between retirement and required distributions, and one point that gets ignored: whether the tax savings from the traditional contribution were actually saved rather than spent.
I wrote about the post-59½ case because I received emails from people who knew the pay-from-taxable-money rule but had no taxable money to follow it. They wanted to know what it costs to do it anyway. That was the gap I wanted to close.
But your version of the question deserves its own piece as well.
Thanks as always for adding to the conversation!
John
Bolden has an excellent projection tool for analyzing ROTH conversions. One feature that helped us decide not to convert was the time factor, which I feel Bolden handled quite well. Bob Berger is also a fan of Bolden, for what it’s worth.
https://www.boldin.com/
I’m on the same path, also using Boldin. Elaborate on “time factor”?. Thanks.
Excellent article.
Thank you, John
Thank you Gary!