Roth Conversions Are a Calendar Problem, Not a Math Problem
A Roth conversion pays off when you use your weird, low-income years to move money from pre-tax to Roth at a discount. The real decision isn't "is a Roth better." It's "which exact years do I convert, and how much in each before I trigger something ugly?" Get the schedule right and the spreadsheet mostly stops being interesting.
A conversion is just a taxable transfer from your traditional IRA to a Roth: you pay ordinary income tax on the amount you move, and after that it grows and comes out tax-free. The math is a straight bet between your tax rate now and your rate later. The part that actually moves your net worth is when that tax falls on the calendar, and what else is hitting your return that year.
This bites hardest from roughly 55 into your early 70s if you've built a seven-figure pre-tax balance. You're staring at a narrow, closing window most planning software nods at but never really centers. I'll map the window, give you bracket logic that's accurate enough to act on, and hand you a one-page test you can build in an evening with a yellow pad and Claude.
I'll admit something. Early in my career I treated this as a modeling problem: elaborate spreadsheets, breakeven curves, Monte Carlo runs. I was solving the wrong thing. The clients who came out ahead weren't the ones with the prettiest models. They were the ones who simply converted in the right years. I stopped building models and started building calendars.
Why is this a calendar problem and not a math problem?
Because the same $100,000 conversion can cost wildly different amounts depending on which year you do it. Convert in a semi-retired year when your taxable income is $80,000, and most of it might be taxed in the low 20s. Convert the same $100,000 in the year you sell your business and stack a large capital gain on top, and you could be pushing into the mid-30s. Same IRA, same investments, completely different tax bill, driven entirely by the calendar.
Over a long retirement, Roth dollars usually beat pre-tax dollars if you've got real money trapped in IRAs. That part isn't controversial. The live question is "in which specific years, and how much each year before I trip something ugly?" That's a sequencing problem. Fill the cheap brackets in the cheap years, and stay out of the expensive brackets and Medicare cliffs in the expensive ones.
What are "gap years," and why should a 55-to-72-year-old care?
Your gap years are the odd stretch between the end of full-time W-2 income and the point where the government turns the income taps back on. For a lot of higher earners that's somewhere between retiring in your early 60s and age 73, when required minimum distributions (RMDs) begin, with Social Security filing somewhere in the middle.
Picture a couple who retires at 62, delays Social Security to 70, and leaves the IRA alone at first. Their taxable income at 63 and 64 can fall from, say, $260,000 while working to under $100,000, mostly dividends, a small pension, a bit of consulting. Then at 70 Social Security switches on, and at 73 RMDs start forcing tens of thousands out of that IRA whether they need the cash or not.
Those low-income years in their 60s are the cheapest conversion window they'll ever get. Do nothing, and the IRS turns the IRA into a forced-income machine in their 70s. Those RMDs can shove you into a higher bracket than you ever hit while working, drag up the taxable share of your Social Security, and bump you over a Medicare premium threshold two years later. Use the gap years and you pre-pay a chunk of that tax at a discount. Ignore them and you hand the government scheduling power over your income.
How much should you convert in a given year?
The practical rule, call it the Bracket-and-Cliffs Test: in a cheap year, convert until you fill the bracket you're comfortable paying, then stop before the next bracket jump or the next Medicare IRMAA threshold. You don't need a PhD model. You need a line you won't cross.
Rough current-law example for a married couple filing jointly: the federal 24% bracket runs up to around the high-$300,000s of taxable income, and it's set to tighten in 2026 as the 2017 tax cuts expire, which is itself a reason not to wait. You might decide, "We'll pay 24% federal on conversion dollars; we are not volunteering for 32%." In a low-income year, that can be a surprisingly large conversion. (Confirm the exact thresholds for the year you act, they move.)
Say both spouses are 64, retired, with $1.4 million in a traditional IRA and another $600,000 across taxable and Roth accounts. They draw $70,000 from cash and have $20,000 of interest and dividends, so taxable income lands near $90,000 after deductions. That leaves roughly $200,000 to $250,000 of room before the next federal bracket. Converting $150,000 to $200,000 this year keeps them inside their target band, and repeated five or six years running, it moves the bulk of that IRA to Roth at an average federal rate in the low 20s, instead of letting RMDs slam them into the 32% bracket and higher Medicare premiums in their 70s.
How do you know whether a given year is good or bad?
Once you see it as a year-by-year puzzle, the signals get clear. You're hunting for cheap years to lean in and expensive years to back off.
| Signal this year | What it really means | Conversion move |
|---|---|---|
| Income drops (retirement, sabbatical, partial year, business loss) | Ordinary income fell; bracket space just opened | Convert aggressively up to your chosen ceiling |
| Big income spike (business sale, option exercise, large bonus) | You're already in a high bracket | Convert little or nothing; wait for a quieter year |
| You're in the 63–72 gap years, no RMDs, Social Security not yet on | Likely your last sustained cheap window | Plan multi-year conversions; don't dribble them out |
| Market pulls back sharply and the IRA balance drops | Same shares, temporarily lower value | Convert while prices are down, more shares move per tax dollar |
| Projected income is brushing an IRMAA threshold | One extra dollar can raise Medicare premiums two years out | Stop short; leave a cushion below the cliff |
The pattern to avoid is the one I've watched too many times: three or four low-income years drift by while everyone says "we'll look at conversions next year," and then RMDs start and the cheap window is gone. Almost nobody regrets modest, bracket-filling conversions in their early 60s. Plenty of people regret waking up at 73 with a $2 million IRA the government now empties on its schedule, not theirs.
What are the two deadlines that quietly decide everything?
Two dates do more work than any spreadsheet.
- December 31, the hard wall. Unlike an IRA contribution, a conversion must be completed inside the calendar year. There's no "do it by April 15" grace period. The income lands in the year the money actually moves, so a conversion you "meant to do in December" but executed on January 2 belongs to a different tax year, possibly a higher-income one. Most missed conversions die right here, with people who waited to see how the year shook out and ran out of runway.
- Two years before Medicare, the IRMAA lookback. Medicare premiums at 65 are set from your income two years prior. A fat conversion at 63 can raise your Part B and D premiums at 65. Often still worth it, but only if you saw it coming. The people who get ambushed are the ones who never mapped the calendar.
Where does Claude actually earn its keep here?
This is a sequencing problem, and sequencing, multi-year, what-if reasoning, is exactly what AI is good at scaffolding. I don't mean trust it with your return. Bracket thresholds and IRMAA tiers change every year, and a hallucinated number here is expensive. I mean use it to build the decision frame you then take to your CPA.
Give Claude your ages, your traditional IRA balance, a rough taxable-income estimate for each of the next several years, and the years you expect Social Security and RMDs to begin. Ask it to lay out a year-by-year conversion schedule that fills a target bracket in each gap year and flags any year that brushes an IRMAA threshold. You'll get a draft calendar in ten minutes that would've cost an afternoon of spreadsheet work, plus a sharp list of questions for your advisor. The judgment stays with you and your CPA; verify every threshold against current IRS figures before you act. The value is the structure, not the arithmetic.
Is it ever too late to convert?
Rarely. Even after RMDs begin at 73, you can still convert amounts above the required distribution in years that have bracket room, and there's a real estate-planning case for it, since Roth dollars pass to heirs free of the income tax that would otherwise hit an inherited traditional IRA drained within ten years. The window narrows with age, but "too late" almost always means "more expensive," not "closed."
So here's the move. Before this December, sit down and sketch a five-year income map, every year from now until RMDs start, with your honest best guess at taxable income in each. Circle the low years. Those are your conversion years, and the size of each one is simply whatever fills your target bracket without tipping into the next one or over an IRMAA cliff. That single page, built now and revisited each fall, is the whole strategy. The math was never the hard part.