Are Roth conversions overrated? What the research actually says

Last reviewed July 2026 · 6 min read

Roth conversions occupy a strange spot in retirement planning: half the internet treats them as a cheat code, and a serious strand of academic work says the typical payoff is smaller and slower than advertised. Both camps are looking at the same arithmetic. What they disagree about is magnitude — and that disagreement turns out to be useful, because it tells you exactly what to check in your own plan.

The settled part: it’s marginal-rate arbitrage

A conversion moves pre-tax dollars to a Roth and adds the converted amount to this year’s ordinary income^[4]. Left unconverted, those same dollars come out later at whatever rate applies then. So the core principle is not controversial: when today’s marginal rate is clearly below the money’s future marginal rate, converting wins mechanically. The subtle part is that “future rate” is not just your own bracket next year. Pre-tax balances face required minimum distributions starting at 73, or 75 for those born in 1960 or later^[5]; a surviving spouse eventually files single, hitting higher rates on similar income; and most non-spouse heirs must drain an inherited IRA within 10 years^[5], often stacking it on top of their own peak-earning-years salary. The money’s future rate is the rate of whoever finally pays the tax.

The contested part: how big is the win?

Edward McQuarrie’s research is the strongest case for tempered expectations. Working through conversion outcomes year by year, he finds the benefits are “often small and slow to arrive” — a conversion almost always pays off eventually, but “eventually” can mean life spans extending past 90, with the converted money left undisturbed the whole time^[1]. He also pushes back on planning folklore: paying the conversion tax from outside funds is less decisive than commonly claimed, top brackets don’t benefit disproportionately, and the biggest wins go to those who can convert partly at a 0% rate^[1].

The practitioner canon reads the same math more optimistically. William Reichenstein and William Meyer, in the Journal of Financial Planning, show conversions adding real value even for higher-income retirees when done in the first few retirement years, before RMDs begin — ideally paired with delaying Social Security^[2]. Michael Kitces’s analyses reach a similar place: a systematic series of partial conversions sized to fill the low brackets each year, on the logic that a low bracket left unused this year is gone forever^[3]. Note what is and isn’t in dispute: both camps agree conversions at rates below the future rate are profitable. They differ on how often real households have a big spread, and how long the payoff takes to show up.

When the spread is big enough to matter

The cases that clear McQuarrie’s bar look like this: a bomb-sized pre-tax balance whose RMDs will land in the 24%+ brackets, and gap years — after the last paycheck, before Social Security and RMDs — where the standard deduction and the 10–12% brackets sit empty. Converting at 12% against a 24%-or-higher future rate is a spread that survives modest assumptions. Add the survivor-single problem and heirs facing the 10-year rule in their own top brackets^[5], and the arbitrage can be genuinely large. One interaction to watch: long-term capital gains ride at 0% below a taxable-income ceiling^[7], and conversion income stacks beneath gains on that same ladder — a conversion can silently push gains that would have been free into the 15% band, raising its true marginal cost.

The counterforces

The skeptics’ strongest evidence is what sits on top of the brackets. Before 65, ACA premium tax credits are generally unavailable once household income exceeds 400% of the federal poverty line — the temporary expansion that suspended that rule applied only to 2021–2025^[8]. Near the cliff, one extra converted dollar can forfeit thousands of dollars of subsidy: an implicit marginal rate in the hundreds of percent. After 65, Medicare premiums are tiered by MAGI — for 2026, surcharges begin above $109,000 single / $218,000 joint and push Part B from a standard $202.90 to as much as $689.90 per month^[9].

Funding matters too. Paying the tax out of the converted amount shrinks the transfer, and before 59½ the portion kept back for taxes is itself a taxable early distribution^[4]. Charitable dollars are the clearest anti-case: from age 70½, qualified charitable distributions send up to $108,000 a year (2025 limit, indexed) from an IRA to charity with the income never taxed at all^[6], and pre-tax money left to charity at death is likewise never income-taxed — converting dollars that were headed to charity anyway means paying real tax on money that would have faced a 0% rate. And the quiet possibility behind McQuarrie’s modest averages: many households’ retirement bracket is simply lower than their working bracket^[1]. For them, deferral already won, and a conversion merely hands part of the win back.

Verdict: it’s an empirical question, per household

“Overrated” is the wrong axis. Roth conversions are a marginal-rate trade whose payoff ranges from strongly positive (big pre-tax balance, empty gap-year brackets, survivor and heir rates looming) to modest (small spreads, long break-evens^[1]) to strictly negative (converting charitable dollars, tripping the ACA cliff). None of that can be settled by doctrine — only by running your own numbers with the cliffs in the model.

Try it in Deorbit Plan

In the Strategy panel, set Conversion strategy to Fill bracket (retired years) and pick a bracket top, then compare against No conversions in the Compare view. The simulator models the ACA subsidy cliff, IRMAA tiers, RMDs, and survivor filing status, so the readout is your household’s answer — not the literature’s average. The Strategy Lab’s sweep (open the Lab) runs every conversion strategy against every claiming age on identical market draws, with cliff-margin warnings when a strategy converts too close to a threshold.

Educational content only — not financial, tax, or investment advice.

See how this plays out with your own numbers. Try it in the simulator →