Which Account First in Retirement: Taxable, IRA, or Roth?

Margaret is 63, single, retired last year in Phoenix, Arizona, and sitting on a $1,000,000 portfolio split three ways: $300,000 in a taxable brokerage account, $600,000 in a traditional 401(k) rolled to an IRA, and $100,000 in a Roth IRA. She needs about $55,000 a year to live on and plans to delay Social Security until 70. The question every retiree in her seat asks: which account do I spend first?

The textbook answer — taxable, then traditional, then Roth — would have her drain the $300,000 brokerage account over roughly five years, paying almost nothing in tax. It feels efficient. But if she follows it blindly, that $600,000 traditional IRA keeps compounding to roughly $850,000 by age 73, and her first required minimum distribution alone is over $32,000, stacked on top of Social Security, pushing her into the 22% bracket and triggering a Medicare IRMAA surcharge. The better plan uses these low-income 60s as a tax-arbitrage window. Here is the full logic.

The default order, and why it works

The conventional sequencing rule has been the same for decades because the math behind it is sound:

1. Taxable brokerage first. When you sell holdings in a regular brokerage account, you owe tax only on the gain, not the whole withdrawal — and long-term capital gains are taxed at 0%, 15%, or 20% (IRC §1(h)). A retiree with low ordinary income can realize gains in the 0% LTCG bracket (taxable income up to $48,350 single / $96,700 MFJ for 2026). Spending this money first also stops the dividends and interest it throws off from inflating your taxable income every year.
2. Traditional IRA / 401(k) second. Every dollar out of a traditional account is taxed as ordinary income (IRC §408(d)). Leaving it for later lets it keep growing tax-deferred — but, crucially, it does not escape tax forever, because RMDs eventually force it out.
3. Roth last. A Roth IRA has no lifetime RMDs and qualified withdrawals are 100% tax-free (IRC §408A). Letting it compound tax-free the longest maximizes the most valuable dollar you own, and it is the ideal asset to leave heirs — they inherit it tax-free.

For a retiree with a modest pre-tax balance, this default is hard to beat. The problem appears only when the traditional bucket is large.

The trap the textbook order sets: the RMD and IRMAA cliff at 73

Required minimum distributions begin at age 73 for anyone born 1951–1959 (age 75 if born 1960 or later) under SECURE 2.0 Act §107. The RMD is your prior-year-end traditional balance divided by an IRS life-expectancy factor — 26.5 at age 73 (IRS Pub. 590-B, Uniform Lifetime Table), or about 3.77% of the balance, rising every year.

If Margaret never touches her traditional IRA in her 60s and it grows from $600,000 to roughly $850,000 by 73, her first RMD is about $32,000 — forced out whether she needs it or not, taxed as ordinary income, and stacked on top of the Social Security she finally started at 70. Two things detonate at once:

• Bracket creep. The RMD plus 85% of her Social Security can push her well into the 22% bracket (income over $48,475 single for 2026), and higher every year as the RMD divisor shrinks.
• IRMAA surcharge. Medicare premiums are means-tested on your MAGI from two years prior. In 2026, single MAGI over $103,000 raises the Part B premium from $185 to $259/month and adds a Part D surcharge — roughly $1,000+ a year of extra Medicare cost triggered purely by the size of a withdrawal she was forced to take.

The strict-sequencing retiree spends their 60s in a low tax bracket doing nothing, then gets ambushed by a high one in their 70s. The fix is to use those low-income years.

The smarter move: fill the low brackets in your gap years

Between the year you retire and the year RMDs and (if delayed) Social Security begin, you have gap years — a stretch of unusually low taxable income. For Margaret, retiring at 62–63 and delaying Social Security to 70, that is up to seven or eight gap years when her only income is brokerage gains. Her 12% bracket and even her 22% bracket are sitting nearly empty.

The tax-smart play is to deliberately pull from (or Roth-convert) the traditional IRA during these years to “fill up” the cheap brackets — even though the textbook says save it for later. For 2026 the bracket ceilings she is filling are:

Add the 2026 single standard deduction of $15,750 (plus the $1,600 age-65 addition once she turns 65), and Margaret can realize roughly $64,000 of total income — standard deduction plus the full 12% bracket — before a single dollar is taxed above 12%. Every year she leaves that headroom unused is a year of cheap conversion capacity she can never get back.

Roth conversion vs. simply spending the traditional account

There are two ways to drain the traditional bucket early, and they serve different goals:

• Spend it for living expenses. If she needs the cash, taking $30,000–$40,000 from the traditional IRA in a gap year — instead of more brokerage — uses up cheap bracket space and shrinks the future RMD base.
• Roth-convert the excess. If her living expenses are covered by the brokerage account, she can convert traditional dollars to Roth up to the top of her target bracket. She pays ordinary-income tax now (conversion deadline is Dec 31, not April 15), but the money then grows tax-free, never faces an RMD, and never counts toward future MAGI or IRMAA. Recharacterization was eliminated by the TCJA, so a conversion is permanent — size it carefully.

One caution on conversions: pay the conversion tax from the brokerage account, not from the IRA itself. Withholding the tax out of the converted amount shrinks how much lands in the Roth and, before 59½, can trigger a 10% penalty on the withheld portion.

Worked example: strict sequencing vs. blended on Margaret’s $1M

Compare two paths for the same $600,000 traditional IRA, ignoring growth for clarity to isolate the tax effect.

The blended path has Margaret pay tax at 12–22% in her 60s instead of dodging it — but it cuts her traditional balance at 73 nearly in half, halves the forced RMD, keeps her in the 12% bracket once Social Security starts, and keeps her MAGI under the $103,000 IRMAA threshold. Over a 25-year retirement, the lifetime tax difference for a profile like this routinely runs into the $50,000–$100,000+ range, plus the IRMAA savings. Arizona has a flat 2.5% state income tax, so the conversions cost her a little at the state level too — still far cheaper than the federal-bracket spike she avoids.

When the strict order actually is the right answer

Blended withdrawals are not universal. Spend taxable-first and leave the rest alone if any of these fit you:

• Small traditional balance. If your pre-tax IRA/401(k) is under roughly $400,000–$500,000, your RMDs will likely stay inside the 12% bracket anyway — there is no spike to defuse.
• No real gap years. If you have a pension or claim Social Security early, your bracket may already be full; conversions then cost 22%+ with no arbitrage.
• Large unrealized brokerage gains you want stepped up. Highly appreciated taxable holdings get a basis step-up at death (IRC §1014), so for assets you intend to leave to heirs, not selling them can beat selling — spend IRA money instead and let heirs inherit the brokerage tax-free.
• Charitable intent. At 70½ you can use Qualified Charitable Distributions (up to $108,000 in 2026) to satisfy RMDs tax-free — which changes the conversion math.

What most people miss

Three blind spots cost retirees the most:

• Treating the order as fixed instead of annual. The right answer is not “empty account A, then account B.” It is: every year, look at your bracket headroom and pull from whichever account fills the cheap brackets without spilling into expensive ones. A good year might mix a brokerage sale for living costs and a Roth conversion up to the top of the 12% bracket.
• Ignoring the widow’s-penalty trap. For married couples, the surviving spouse files single the year after the first death — with brackets and the IRMAA threshold roughly half as wide, on a similar income. A large traditional balance that was tolerable filing MFJ becomes punishing filing single. Gap-year conversions while both spouses are alive defuse this directly.
• Forgetting capital gains stack on top of ordinary income. Roth conversions raise your ordinary income, which can push your long-term capital gains out of the 0% bracket and into 15%. Model the conversion and any planned brokerage sales together, not separately — the order you realize them in matters.

The decision lever

The single variable that decides your withdrawal order is the size of your traditional (pre-tax) balance relative to your bracket headroom in your 60s. If that balance is small, follow the textbook order — taxable, traditional, Roth — and keep it simple. If it is large (roughly $600,000+ for a single filer, $1M+ for a couple) and you have low-income gap years before age 73, flip the script: intentionally tap or Roth-convert traditional dollars now to fill the 12% and 22% brackets, shrink the RMD base, and protect yourself from the IRMAA cliff and the widow’s penalty. Pull your prior-year-end traditional balance and your projected RMD at 73 today — if that future RMD pushes you above where you sit now, you have gap-year work to do this year, because every empty bracket you leave unfilled expires on December 31.