Roth vs Traditional: The Bracket Arbitrage

The answer flips at least three times in a physician career

Roth versus traditional generates more lounge debate than almost any other money question, and most of the debate misses the structure of the problem. There is no permanently correct answer — there is a correct answer for a given year, and a physician career changes that answer at least three times. In residency, your marginal federal rate can sit at 12 percent, the lowest it will ever be, and Roth dollars are close to a gift. At peak attending income, the same contribution decision runs through a 35 percent marginal rate, and the traditional deduction becomes the obvious move. Then, in any low-income window — a sabbatical, unpaid leave, or the gap years between retirement and required minimum distributions — the answer flips back to Roth, this time through conversions. The stakes per decision are concrete: on a single $10,000 contribution, choosing correctly versus incorrectly swings roughly $1,000 to $1,100 of tax, and you make this decision every year for thirty-plus years. This module gives you the one comparison that settles all of it, worked with verified 2026 numbers.

bracket arbitrage

Choosing when retirement dollars are taxed — at contribution (Roth) or at withdrawal (traditional) — so the tax lands at the lower of the two marginal rates.

Strip away the marketing language and the two account types differ in exactly one respect: when the tax is charged. Traditional dollars skip tax at your marginal rate today and pay ordinary income tax at withdrawal; Roth dollars pay your marginal rate today and skip tax at withdrawal. The growth in between is a red herring, because the arithmetic is commutative: $10,000 that grows eightfold and is then taxed at 24 percent ends at exactly the same after-tax value as $10,000 taxed at 24 percent first and then grown eightfold. If the two rates are equal, the accounts tie to the penny. The only variable that breaks the tie is the rate itself — your marginal rate in the contribution year against your expected marginal rate in the withdrawal year. Lower rate wins. Everything else you have heard, from tax-free compounding to tax diversification, is either a restatement of this comparison or a hedge against not knowing the second number. Note that both ends are marginal rates, never effective rates: contributions come off the top of your income, and withdrawals stack on top of whatever other retirement income you have.

Why it matters: Every other framing of Roth versus traditional dissolves into this one comparison, which means the decision is recomputable in about a minute any year your income changes. Physicians who learn the comparison stop treating the choice as an identity and start treating it as an annual arbitrage — Roth in cheap years, traditional in expensive ones, conversions whenever a window opens.

The same $10,000 at 12 percent and at 35 percent: opposite answers, worked

A PGY-1 earning $64,000 and an attending earning $450,000, both single, both contributing $10,000 to an employer plan in 2026. Assume the withdrawal-year marginal rate is 22 to 24 percent — a typical projection for a physician-sized traditional balance in retirement.

Bottom line: The identical $10,000 contribution produces opposite verdicts — Roth in residency locks in 12 percent on dollars headed for 22 to 24, and traditional at peak earnings saves 35 cents on the dollar against a projected 24 — with roughly $1,000 to $1,100 of tax riding on each year you get it right.

The conversion window: six months off opens $122,875 of cheap bracket space

The third flip in the career arc is the low-income window, and it is the one physicians most often waste. Any year your income drops — sabbatical, unpaid parental leave, fellowship transition, or the gap years before required minimum distributions — the brackets above your reduced income sit empty, and a Roth conversion can fill them at rates you will never see again in a working year. The mechanics: you move dollars from a traditional account to a Roth account, the converted amount is added to that year's ordinary income, and the dollars are tax-free thereafter.

You are single and take a six-month sabbatical in 2026, so your income for the year is $95,000 instead of $450,000. You hold $600,000 of traditional 403(b) and IRA money from prior years, and next year you return to a 35 percent marginal rate. You have taxable cash available to pay a tax bill. What do you do with the window?

The state leg: deducting at 0 percent and withdrawing at up to 13.3

The federal bracket is only one leg of the arbitrage. Your state marginal rate at contribution and at withdrawal is a second comparison, and it follows where you live when you withdraw, not where you earned the money — federal law (4 U.S.C. §114, enacted as P.L. 104-95 in 1996) bars states from taxing the retirement-plan income of nonresidents. Consider an attending who practices in Texas, deducts traditional contributions at a 0 percent state rate for twenty years, then retires to California: every withdrawal now runs through California brackets that reach 12.3 percent, plus a 1 percent Mental Health Services Tax above $1,000,000 — 13.3 percent at the top. The deduction never saved a state cent, and the withdrawal pays one of the highest state rates in the country. That direction argues Roth. The reverse move flips it: a California attending deducting at roughly 9.3 to 11.3 percent state (confirm current Franchise Tax Board brackets) who retires in Nevada captures the deduction and pays 0 percent at withdrawal — traditional gets a bonus leg. A second, related failure: the flat claim that you will be in a lower bracket in retirement is not a law of nature. A physician who defers aggressively for thirty years can reach required-minimum-distribution age — 73 if born 1951 through 1959, 75 if born 1960 or later — with several million traditional dollars, and the forced distributions stack into the 24 to 32 percent brackets or higher.

How to avoid it: Run the state leg explicitly before defaulting to traditional: write down your current state marginal rate and the rate of the state where you honestly expect to retire, and add the difference to the federal comparison. Direction matters — moving toward a high-tax state strengthens Roth; moving away strengthens traditional. Then project your traditional balance to your RMD age at 6 percent growth; if the forced income lands in your current bracket, deferring more is not lowering your rate.

Check yourself: which two numbers decide it

The most common version of this decision fails on the choice of inputs, not the arithmetic. Confirm you can name the decision variable before you run it on your own paycheck.

One comparison, three flips, two legs

• The entire decision is marginal rate at contribution versus expected marginal rate at withdrawal; growth is commutative and cannot break a tie between equal rates.
• The answer flips across a physician career: Roth in residency at 12 to 22 percent, traditional at peak attending rates of 32 to 37 percent, and Roth conversions in low-income windows.
• In 2026, a single physician with $95,000 of sabbatical-year income has $122,875 of conversion room below the top of the 24 percent bracket, at an effective 23.6 percent instead of 35.
• The state leg follows your residence at withdrawal (4 U.S.C. §114): practicing in Texas and retiring in California argues Roth, while the reverse move strengthens traditional.
• The claim that you will retire into a lower bracket fails for physicians with large traditional balances, because RMDs at age 73 or 75 force six-figure ordinary income whether you need it or not.

Do this next: Write down two numbers today — your current marginal rate (federal plus state) and your honest projected marginal rate at withdrawal, including RMD income — and note which of the two is lower; recheck the pair any year your income moves by a bracket.