How much should I convert to Roth in the gap years before RMDs?
Filling the 12% bracket each gap year paid off at every balance we tested, most reliably on large IRAs. For a married couple both 65 in 2026 (born 1961, so RMDs start at 75), with $36,000 a year of Social Security from 67 and a plan to age 90, converting to the top of the 12% bracket in each of the 10 gap years (65–74) cut lifetime federal income tax by a median $38,224 on a $1,000,000 IRA, $92,490 on $1.5M and $146,598 on $2M, winning in 76.4%, 92.2% and 92.6% of 20,000 return paths. On a $500,000 IRA the median saving was $11,870, but converting lost in 37.5% of paths. Every year is priced by QuantCalc's federal tax engine on 2026 law projected by statute, federal income tax only, in 2026 dollars — see the full results table and the methodology.
Updated 2026-10-12: re-priced on the federal tax engine. Until this update the study priced every year on the 2025 brackets with a $30,000 standard deduction, its own simplified Social Security rule and RMDs from 73. That overstated the tax on, for example, $100,000 of IRA income for a couple aged 66 ($7,923 against the engine's $5,807 on 2026 law) and gave this household two fewer gap years than SECURE 2.0 does. The headline moved from a median $31,741 to $38,224 at $1M, and the $500,000 result changed sign (from a median loss of $12,939 to a median saving of $11,870). The earlier figures are kept in the JSON summary under previous_version.
Key numbers — each row links to the section that carries the full table. Updated 2026-10-12.
| Key number | Value | Where on this page |
|---|---|---|
| Median lifetime federal income tax saved by filling the 12% bracket, $1,000,000 IRA (25th–75th pct: +$2,042 to +$99,631) | +$38,224 | Full results |
| Median saving, $1,500,000 IRA | +$92,490 | Full results |
| Median saving, $2,000,000 IRA | +$146,598 | Full results |
| Median saving, $500,000 IRA (converting loses in 37.5% of paths) | +$11,870 | Full results |
| Share of return paths where converting wins: $500k / $1M / $1.5M / $2M | 62.5% / 76.4% / 92.2% / 92.6% | Full results |
| Conversion rule tested | each gap year 65–74, convert to the top of the 12% bracket | Methodology |
| Deterministic flat-5% check, $1,000,000 IRA | $44,640 lifetime saving | Methodology |
"Do Roth conversions in the gap years" is one of the most repeated pieces of retirement-tax advice: the window between retiring and the first required minimum distribution (RMD), when taxable income is unusually low. The pitch is that filling the low brackets now avoids a larger RMD-driven tax bill later. The advice is usually given without a number attached, so we attached one: the entire lifetime federal income tax bill for a representative married couple under two strategies on the identical portfolio and the identical returns — do nothing, or convert just enough each gap year to fill the 12% bracket.
Headline finding: A married couple with a $1,000,000 Traditional IRA who does nothing pays a median $38,224 more in lifetime federal income tax than one who fills the 12% bracket with Roth conversions during the gap years (both 65 in 2026, RMDs from 75, plan to 90). The gap is $92,490 at $1.5M and $146,598 at $2M. At $500,000 the median saving is $11,870, but converting loses in 37.5% of return paths: the advice holds on average at every balance tested, and becomes more reliable as the balance grows.
Why doing nothing gets expensive: the RMD bracket trap
Leave a large Traditional IRA untouched and two things compound against you. First, the balance keeps growing: a $1,000,000 IRA earning 5% real grows to about $1,628,895 by age 75. Second, RMDs are forced: the IRS Uniform Lifetime Table (Publication 590-B) requires a rising fraction of the balance every year, starting at 1/24.6 at 75. On that balance the first RMD is about $66,215 in 2026 dollars, and on top of it 85% of the couple's Social Security becomes taxable (Publication 915; with Social Security alone, none of it is), putting their last dollars in the 12% bracket. Every year after, the required fraction grows.
The gap years are the escape hatch. Before RMDs and before Social Security, a couple living off taxable savings can have near-zero ordinary income. Under 2026 law the standard deduction, the extra deduction for each spouse 65 or older (§63(f)) and the temporary $6,000-per-person senior deduction (26 U.S.C. §151(d)(5)(C), 2025 through 2028) sit under the 10% and 12% brackets, so in 2026 this couple can convert $148,300 before leaving the 12% bracket; once Social Security starts and the senior deduction ends, the room is about $105,644 a year in 2026 dollars. Converted dollars then grow tax-free and are exempt from RMDs, shrinking every future forced withdrawal.
The trap is that inaction feels free. Doing nothing has no upfront tax bill, so it looks costless. But the bill is only deferred, and deferral lets the balance grow into larger RMDs. For this couple the window to convert at 10% and 12% is open for 10 years and closes when RMDs begin at 75.
Headline chart: lifetime tax cost of inaction, by IRA size
Median extra lifetime federal income tax from doing nothing vs filling the 12% bracket
Positive bars = converting saved money (doing nothing was costlier). 20,000 Monte Carlo return paths per IRA size; same couple, same returns, only the strategy differs. 2026 dollars.
The median saving rises with every step in balance: $11,870 at $500,000, $38,224 at $1,000,000, $92,490 at $1,500,000, $146,598 at $2,000,000. At $500,000 the conversions empty the IRA within the first few gap years at 10% and 12%; whether that beats waiting depends on returns, because in weak-return paths the smaller IRA's later RMDs stay small and the tax paid early is not recovered by age 90.
Full results: four IRA sizes
Each row is 20,000 Monte Carlo paths on the real return. "Median saving" is the typical lifetime federal income tax saved by converting versus doing nothing (positive = converting wins). "Conversion wins" is the share of return paths where converting produced the lower lifetime bill. All figures are 2026 dollars.
| Traditional IRA | Lifetime tax: do nothing | Lifetime tax: convert | Median saving | 25th–75th pct saving | Conversion wins |
|---|---|---|---|---|---|
| $500,000 | $59,168 | $46,398 | +$11,870 | −$12,929 to $37,363 | 62.5% |
| $1,000,000 | $149,779 | $115,987 | +$38,224 | $2,042 to $99,631 | 76.4% |
| $1,500,000 | $271,391 | $180,785 | +$92,490 | $33,632 to $179,167 | 92.2% |
| $2,000,000 | $419,729 | $270,882 | +$146,598 | $61,740 to $236,758 | 92.6% |
Three things stand out:
- The cost of inaction scales with balance. A $2M IRA owner who does nothing pays a median $419,729 in lifetime federal income tax versus $270,882 if they convert, a $146,598 gap.
- Roth conversions are not a sure thing on smaller balances. At $500,000, converting lost money in 37.5% of return paths, even though the median outcome favoured it.
- The win rate rises with balance. Conversions beat doing nothing in 76.4% of return paths at $1M and 92.6% at $2M. The bigger the future RMD problem, the more reliably the gap years pay off.
CC-BY-4.0 — free for any use including republication and journalism, with attribution to QuantCalc Research.
What the model leaves out
The study counts federal income tax only. Each omission below is a real cost or benefit that would move the answer for a particular household:
- No state income tax. In a state that taxes IRA withdrawals, both the conversions and the RMDs carry state tax too.
- No IRMAA. A conversion or an RMD can raise the Medicare Part B and D premiums charged two years later (42 U.S.C. §1395r(i)); neither strategy is charged for it here.
- No heirs, and the plan stops at 90. Whatever is left in the Traditional IRA at 90 is still pre-tax money, while converted dollars are not; most non-spouse beneficiaries must empty an inherited IRA within 10 years (26 U.S.C. §401(a)(9)(H)). The study counts neither.
- Future law. Each year is projected from 2026 law as written, including the senior deduction's end after 2028; legislation after 2026 is not modelled.
Methodology
Engine: roth-gap-engine-2.0.0 · every year priced by QuantCalc's federal tax engine (the website's copy of the C engine, held to it exactly) the way the planner's tax-aware Monte Carlo prices a year · fixed seed 20261012 · 20,000 Monte Carlo return paths per IRA size · plus a deterministic flat-5% case whose every year is published and re-priced through the C engine on each release.
Household: married filing jointly, both spouses 65 in 2026 (born 1961), retired, $36,000 a year of Social Security (2026 dollars, COLA-indexed) from 67, plan to age 90. Spending is funded outside the IRA in both strategies, so the IRA's only withdrawals are conversions and RMDs.
Tax law: 2026 federal law (Rev. Proc. 2025-32; the senior deduction of §151(d)(5)(C)), projected to each tax year the way each amount's statute moves it, on a 2.5% inflation assumption (the planner's API default). Social Security is taxed under §86 with its fixed thresholds. Each year is priced in that year's dollars and deflated back to 2026 dollars. Federal income tax only.
RMDs: from age 75, the applicable age for a 1961 birth year under SECURE 2.0 (26 U.S.C. §401(a)(9)(C)(v)), on the IRS Uniform Lifetime Table (Publication 590-B) applied to the prior year-end balance.
Strategies: Do nothing — no conversions; the Traditional IRA grows untouched until RMDs. Convert — each gap year (65–74), convert just enough to bring taxable income to the top of that year's 12% bracket (found by bisection on the engine's own taxable income, since Social Security taxation makes it nonlinear); the conversion tax is paid from outside the IRA and converted dollars grow tax-free and are exempt from RMDs. Both strategies face the identical return path, so the difference is the strategy alone.
Returns: annual real return drawn Normal(5%, 10%). The deterministic check uses a flat 5% real and gives a $44,640 lifetime saving at $1M ($160,627 doing nothing, $115,987 converting).
The conversion target, the top of the 12% bracket, is deliberately moderate; filling the 22% or 24% bracket would convert more on the largest balances but brings more case-specific judgment. The dataset and the deterministic ledger are published above for anyone who wants to check a year.
What would conversions save you?
The answer depends on your balance, your bracket, your state, and your spending. Model your own conversions against your own RMD schedule instead of a rule of thumb.
Plan Your Retirement Taxes →Related research
- → 510,000-path 51-State Retirement Monte Carlo — the state-tax companion; California adds $31,115 of state income tax over 30 years for its couple, on top of the federal picture here.
- → The 2026 ACA Cliff for Early Retirees — the pre-Medicare bridge, where conversion income can collide with ACA subsidy thresholds.
- → Social Security Claiming Age Under Uncertainty — the claiming-age decision that interacts with conversion timing.
Frequently asked questions
Last updated 2026-10-12. Dataset license: CC-BY-4.0. Tax law: Rev. Proc. 2025-32 (2026 amounts), 26 U.S.C. §§86, 151, 401(a)(9); IRS Publication 590-B Uniform Lifetime Table. QuantCalc is an independent retirement-planning research project. Not affiliated with the IRS or any government agency. Not financial, tax, or legal advice.