Tax Planning

Should I Do Roth Conversions? The Question Is Dosage

Most people who do Roth conversions don't do enough of them. Here's a plain-English way to think about the right dose, and the short window where it tends to matter most.

By David Talley, CFP®, EA August 6, 2026 8 min read

Short Answer

The problem is almost never that somebody converted too much. It's that they converted too little, and a conversion done in a timid dose leaves the actual problem about half solved. Look at your traditional IRA and you don't fully own that number, because you and the government own it together and he hasn't taken his cut yet. A conversion is buying that partner out: you pay the tax now, on your terms, in a year you pick, at a rate you can actually read on a page. The years that make it worth doing usually open when your last paycheck stops and close when required distributions begin at 73, and in that window there tends to be room at the low end of your brackets that would otherwise go completely unused. So the whole topic comes down to one word, and the word is dosage. How much, in which years, up to which edge. Conversions are genuinely not always the answer, and if your bracket now looks about the same as your bracket later then the whole thing can be a wash.

I land on the same verdict here more often than any other one. The problem is almost never that somebody converted too much. It's that they converted too little.

I see the same version of it constantly. Someone hears about Roth conversions on a podcast, or a neighbor brings it up over the fence, and they go do a small one, a few thousand dollars maybe, and it feels responsible and they checked the box and that's exactly the trap, because a conversion done in a timid dose leaves the actual problem about half solved. You did the thing. You just didn't do enough of it to change anything.

Let me back up and talk about this the way I'd if you were sitting across from me. I'm not going to try to convince you that conversions are always right. Sometimes they're a wash. Six one way, half a dozen the other, and when that's true I'll say so and we'll skip them. Everything below is general education rather than a recommendation for your situation.

You And The Government Own That IRA Together, And The Window Is Short

Look at your traditional IRA, or that old workplace account you rolled over and stopped thinking about, and say it shows a number you're genuinely proud of. You don't fully own that number. You and the government own it together, Uncle Sam hasn't taken his cut yet, and he's going to take it eventually, either from you down the road or from whoever inherits it after you're gone. You have a partner. That partner never asked your permission.

I think of a Roth conversion as buying your partner out. You pay the tax now, on your terms, in a year you pick, at a rate you can actually read on a page, and from that day forward the money is yours. All of it. It grows without the partner, it comes out without the partner, and it lands in your kids' hands without the partner. You aren't avoiding the tax. You're choosing when to settle it.

Once you see it that way the question changes shape. It stops being "do I want to pay tax I could defer" and turns into something closer to "when would I rather settle up, now at a rate I can read, or later at one I cannot?"

I care most about a stretch of years that quietly does a lot of the work. It's easy to miss. It doesn't announce itself.

It usually opens the year your last paycheck stops and your income drops, and it usually closes when required minimum distributions begin, which for most people is the year they turn 73, at which point the government starts pulling money out of your retirement accounts whether you happen to need it that year or not (IRS: required minimum distributions FAQs). In between sits a window. Your taxable income in there may be the lowest it'll ever be again.

That gap is the whole opening. In those years there tends to be room at the low end of your brackets that would otherwise go completely unused, and a conversion can fill that room at a rate you may never see again for the rest of your life. Miss the window and the room is gone. It doesn't roll over. It doesn't wait on you. I'd much rather find that window with somebody early than explain it to them late.

I sometimes describe the whole thing as eating your broccoli. Nobody is excited to pay tax a year early. It isn't the fun part of the meal. But done in the right years, in a real dose, it tends to be the part that's good for you, so you do it anyway.

Filling A Bracket To The Top Without Spilling Over

Surface question Do I want to pay tax I could defer? That's where most conversion talk quietly fails people. It stops at "consider a conversion" and never once gets to the size of the bite.

Deeper question I think about it as bracket filling. Picture your tax brackets as buckets stacked from the floor up, each one holding income taxed at its own rate (IRS: federal income tax rates and brackets), and in a low-income year the lower buckets have real empty room in them, so the idea is to pour income into that empty room on purpose by converting just enough to fill a bucket to the top without spilling over into the next one that costs more. You have to watch the bucket while you pour. There are edges out there you don't want to cross by a single dollar, and some of them are true cliffs where going one dollar over changes something all at once, like what you pay for Medicare a couple of years down the line, while others are gentler slopes you can lean into a little. Knowing which edge is which, in your numbers, is most of the actual work.

Why it matters I see the opposite mistake far more often than I see somebody overfill. People convert a timid little amount, nowhere near the top of the bucket they're already standing in, and at that pace it can take twenty or thirty years to move any meaningful share of the account. You don't have twenty or thirty low-income years. You may have a handful. If the dose is too small you run out of window long before you run out of account, and the problem you set out to solve only ever gets half solved. I'd put this whole topic in one word, and the word is dosage. How much, in which years, up to which edge.

Five Things That Set The Dose

So how do you actually size it? I keep coming back to five things, and every one of them can move the answer in either direction.

How much cheap bracket room is sitting empty this year

Look at your projected taxable income against where the bracket lines fall, and the gap between the two is the room you can fill. I care about that number more than almost any other one here, because the room is the entire opportunity and it's a different size every single year. The tradeoff is that you're committing real cash to tax today for a benefit that shows up over decades, so in a year where the room is thin, the friction may not be worth it.

Where the money to pay the tax is coming from

Look at your cash and your taxable brokerage account. Paying the conversion tax from outside the IRA is what lets a conversion do most of its work, because every dollar of the converted balance stays invested and keeps growing on the other side. It changes the answer because withholding the tax out of the IRA itself shrinks the very balance you were trying to move, and before age 59 and a half that withheld piece can also be treated as an early distribution (IRS: exceptions to tax on early distributions). The tradeoff is liquidity. Cash spent on tax is cash you no longer have.

The edges that sit above the bracket line

Look at what else in your life is keyed to income. Medicare Part B and prescription drug premiums carry an income-related surcharge that's set from your tax return two years earlier, so a conversion at 63 is the one that shows up in your premium at 65 (SSA: Medicare premiums, rules for higher-income beneficiaries). If you retired before 65 and you're buying your own health coverage, marketplace help is keyed to income too. These edges can make one extra dollar expensive in a way the bracket table never shows you, and that's what moves the answer. The tradeoff is that dodging every edge forever usually means converting almost nothing, and that has a cost of its own.

How many low-income years you actually have left

Count them. If required distributions start at 73 and you retired at 62, you may have a decade to work with. If you retired at 70, you have about three years and one of them is half gone. I size the dose per year as roughly the amount you want to move divided by the years you have, so fewer years means bigger annual bites. The tradeoff is that bigger bites push you into higher brackets, and a short runway sometimes means accepting a rate you'd rather not pay, or accepting that you won't move all of it.

Who inherits the account, and what their brackets look like

Look at your beneficiaries. Most non-spouse beneficiaries now have to empty an inherited retirement account within ten years, and those ten years tend to land right in the middle of their own highest-earning stretch. It changes the answer because a conversion at your rate today can be cheaper than a distribution at their rate later. The tradeoff is that you'd be paying a known tax now to solve a problem that belongs to somebody else, on assumptions about their future income that neither of you can really be sure about. I wouldn't let this one be the whole reason.

What I Would Want To See Before Sizing A Conversion

None of this can be answered from a rule of thumb. These are the numbers the dose is actually built out of.

Your projected taxable income this year against where the next bracket line falls, because the gap between them is the room you can fill
Whether the tax can be paid from cash or a taxable brokerage account instead of out of the IRA itself
What else in your life is keyed to income: a Medicare surcharge set from your return two years back, and marketplace help if you retired before 65
How many low-income years sit between your last paycheck and the year required distributions begin
Roughly how much of the account you want to move, divided by the years you have to move it
Who inherits the account, and what their own brackets are likely to look like during their ten years to empty it
Whether most of your retirement money already sits in a Roth or a taxable brokerage account, which can make this somebody else's problem rather than yours
Whether writing the check would take the cash reserve you actually need to live on

When I Tell People To Skip It

I want to be square with you here, because conversions are genuinely not always the answer, and if your bracket now looks about the same as your bracket later then the whole thing can be a wash, and paying tax early for no gain is just paying tax early. I lean in some years. I skip others. If most of your retirement money already sits in a Roth or in a taxable brokerage account, the problem I've been describing may not be your problem at all. If you intend to leave the account to charity, the tax you'd prepay may never have been owed by anybody. And if writing the check would take the cash reserve you actually need to live on, my answer this year is no.

The difference lives in the details of your income, your other accounts, and the edges specific to you. The concept is simple. Fitting the dose to your life is the work.

Follow-Up Questions

What does a survivor's tax bill have to do with converting now?

I'll keep this brief, because it deserves its own conversation, but it's often the thing that finally gets people to eat the broccoli. I want you to think about a married couple filing jointly today, with two sets of brackets and a joint standard deduction, and now imagine as gently as I can put it that one of them passes first, the way one of them eventually will. The survivor keeps most of the same income. The same required distributions, too. But the following year they're usually filing as a single person on single brackets that are far less roomy, so it's the same money, sometimes even less money, taxed as though it were more. I call that the widow's tax. It's one of the quiet reasons couples plan conversions now, together, while there are still two sets of brackets to work with. It's a kindness done in advance for the version of the household that'll still be here later. I've written about the widow's tax on its own.

What would this look like for a couple in their early sixties?

Let me walk through an illustration, and it's only an illustration. Imagine a couple who retire in their early sixties and won't face required distributions for roughly a decade. In their working years they sat near the top of a fairly high bracket. Then the paychecks stop, and Social Security hasn't started, and those forced distributions are still years out, so their taxable income falls a long way, and that drop is the opportunity. Doing nothing in those years isn't the safe choice it feels like. It can quietly mean a bigger forced distribution later, more of their Social Security counted as taxable income (IRS Publication 915), and a survivor eventually filing single on all of it. Filling their brackets on purpose in those quiet years, in a real dose rather than a token one, is what a plan is trying to capture. Whether it's right for them, and exactly how much, depends entirely on their own numbers. That's the point. The dose is personal.

Can I undo a conversion if I convert too much?

No, and this one matters. Recharacterizing a conversion used to be allowed and then it was taken away, so since January 1, 2018 a conversion from a traditional, SEP or SIMPLE IRA into a Roth can't be reversed (IRS: retirement plans FAQs regarding IRAs). Once it's done, it's done. I think that's the single best argument for converting late in the year, when you can actually see what your income turned out to be, instead of guessing at it in March.

Do I have to pay the tax out of the IRA itself?

Generally no, and I'd rather you didn't. If you have cash or a taxable brokerage account to cover it, paying from outside keeps the whole converted balance working for you on the Roth side. Paying from inside shrinks the balance you were trying to move, and before 59 and a half the withheld piece can be treated as an early distribution with a penalty attached. I'd usually have somebody convert a smaller amount rather than skip the year entirely.

Will a conversion push up what I pay for Medicare?

It can, and the timing is what blindsides people. The income-related surcharge on Part B and prescription drug coverage is set from your tax return two years back, so the conversion you do at 63 is the one that shows up in your premium at 65 (SSA: Medicare premiums). I look at that before we convert anything. It moves in steps, so crossing a line by one dollar costs the whole step, and by itself that's still not a reason to skip a conversion that's otherwise clearly worth doing.

Does a Roth force me to take money out later?

Not during your lifetime. The IRS puts it about as plainly as it ever puts anything, which is that you can leave amounts in your Roth IRA as long as you live (IRS: Roth IRAs). I think that's a real part of the appeal. It means the money isn't being pulled out on somebody else's schedule and taxed alongside everything else in a year you didn't choose.

Sources

Related Talley Wealth Resources

If this question is on your mind, these pages are natural next reads:

Next step

Find Out Whether You Are Leaving Something On The Table

If you're somewhere in the years between your last paycheck and your first required distribution, or you're looking ahead at that window and wondering whether you're leaving something on the table, that's exactly what a no-pressure Explore Call is for. It's a short conversation, no preparation needed, and if we aren't the right fit I'll tell you so directly.

Schedule an Explore Call

For discussion purposes only. This is not individualized legal, tax, or investment advice. Estate, tax, retirement, and investment decisions should be reviewed against the full facts and with the right professional before implementation.