Tax-Optimized Withdrawals, Part 2: Conventional vs Proportional Order

By Tom, CPA · Updated October 25, 2026

TL;DR: The conventional order (taxable, then tax-deferred, then Roth) is simple but often bunches taxable income late in life. Blending withdrawals so taxable income is smoother from year to year usually lowers lifetime tax. In our 35-year illustration, the conventional order paid roughly twice the federal tax of the blended approaches.

Educational only, not tax, legal, or investment advice. Examples are illustrations, not predictions. Rules and limits change; verify current figures and talk to a qualified professional about your situation.

"Taxable first, Roth last" is the rule of thumb you will see everywhere. It works beautifully, right up until it trips you down the stairs at age 75 carrying a giant required distribution in each arm.

This part of the series explains where the conventional order came from, what a proportional or blended approach does differently, and why the answer depends on one thing above all: how evenly your taxable income is spread across your retirement years.

CPA Warning Label: Withdrawal order is a multi-year tax decision. Running one year's math in isolation is like judging a movie by its trailer: entertaining, often wrong.

The conventional order and why it exists

The conventional sequence says: spend your taxable brokerage money first, then your tax-deferred accounts (traditional 401(k) and IRA), and save Roth money for last. The logic is sound on its face:

• Taxable first because selling only taxes the gain, often at 0% or 15%, and because dividends from that account are taxed every year anyway. Spending it down shrinks an ongoing tax drag.
• Tax-deferred second to let it grow without annual tax as long as possible.
• Roth last because it grows tax-free forever and has no lifetime required distributions.

That is the "maximize deferral" philosophy. Deferral is nice. But deferral is not the same as savings. If deferring a dollar today means it is taxed at a higher rate later, you have not saved anything; you have just taken out a loan from your future self at a terrible rate.

The conventional order's weakness is its shape. Early retirement years under this plan often have almost no taxable income: you are living on brokerage sales with mostly-basis and 0% gains. Then the brokerage account runs dry, and suddenly all spending comes from the IRA, which is 100% ordinary income. Then Social Security starts. Then required minimum distributions kick in. Taxable income goes from a pancake to a skyscraper.

Proportional withdrawals explained

A proportional approach takes each year's spending from all buckets in proportion to their balances. If 40% of your money is in taxable, 50% in tax-deferred and 10% in Roth, a $60,000 withdrawal would be roughly $24,000, $30,000 and $6,000.

The effect is a steadier stream of ordinary income every year. Some of the IRA is taxed in your early years, when you would otherwise waste the standard deduction and the 10% and 12% brackets. That shrinks the IRA, which shrinks later required distributions.

A close cousin is the bracket-targeting approach: draw (or convert) from tax-deferred accounts each year up to a chosen line, such as the top of the 12% bracket, and fill the rest of spending from taxable or Roth money. Instead of a fixed ratio, you use a fixed tax target. Part 3 covers bracket-filling in detail.

Neither is magic. Both are attempts to flatten the income skyline so more dollars are taxed at low rates.

Worked example over 35 years

Illustration. Assumptions (deliberately simple, not a prediction):

• Married couple, both 55, filing jointly; spending $80,000 a year in today's dollars.
• Starting balances: $600,000 taxable (half of it cost basis), $900,000 traditional IRA, $100,000 Roth IRA.
• 4% real return every year, no volatility; 2026 federal brackets and standard deduction held constant in real terms (they are indexed for inflation).
• $48,000 a year of combined Social Security starting at 67; required distributions starting at 75 using the IRS Uniform Lifetime Table.
• Federal tax only; tax is paid from the taxable account, then Roth; model runs through age 89.

What happened? Under the conventional order the couple paid almost nothing for about a decade, then spent their 70s and 80s with required distributions plus 85% of Social Security taxable. Their tax roughly doubled compared with either blended approach, and they still ended with nearly a million dollars of pre-tax money that someone will eventually pay tax on.

The two blended approaches landed close together. Bracket-filling converted everything to Roth, so its ending balance is fully tax-free; the proportional plan's ending balance looks bigger but still carries an embedded tax bill on the $631,000 IRA. If heirs eventually pay 22% on that IRA, the two are within a few percent of each other. Bracket-filling also paid some tax inside the Social Security tax torpedo zone once benefits began, which is why it did not run away with the contest.

The lesson is not "proportional always wins." It is that any approach that smooths taxable income beat the stair-step shape of the conventional order in this model.

Where conventional order backfires

The conventional order tends to hurt most when:

• The tax-deferred bucket is large relative to the taxable bucket. Big IRAs become big required distributions.
• Retirement starts early. A 50-year-old retiree may have 15 to 25 low-income years before Social Security and RMDs, and the conventional order wastes most of them.
• A spouse is likely to be widowed. The survivor eventually files as single with narrower brackets while the IRA is still large. See Part 7.
• Medicare premiums are in play. Large late-life distributions can cross IRMAA thresholds, which in 2026 begin above $218,000 of modified adjusted gross income for joint filers and $109,000 for others (CMS).

It holds up better when the taxable account is huge relative to the IRA, when heirs are in low brackets and will inherit the IRA, or when health-insurance subsidies depend on keeping income very low before 65. In that last case, deliberately recognizing IRA income can cost more in lost premium tax credits than it saves in future tax. As of October 2026, the enhanced premium tax credits have expired, and households above 400% of the federal poverty line are not eligible for any credit (CMS), which makes that cliff very real.

Which bucket pays the tax bill

One quiet detail swings these comparisons: where the tax money itself comes from. When you convert or withdraw from an IRA, paying the resulting tax from the taxable account is usually the most efficient choice. The full converted amount stays in the Roth growing tax-free, and the taxable account, which carries an annual tax drag on dividends, gets smaller. Paying the tax by withholding from the IRA itself shrinks the converted amount and, before 59½, the withheld portion can be treated as an early distribution.

This is also why the conventional order's "spend taxable first" instinct is not wrong, just incomplete. A brokerage account is the best place to pay conversion taxes from. If you burn through it entirely in the first decade on living expenses, you lose that tool right when it could be most useful. A blended plan keeps some taxable money alive longer as a tax-payment fund, a cash cushion for bad markets, and a source of 0% gains in low-income years.

Choosing a default rule

If you want a starting rule you can actually follow, a bracket-target approach is the easiest to run in practice:

1. Each January, decide how much ordinary income you want to recognize this year. Common targets: the standard deduction only, top of the 10% bracket, or top of the 12% bracket ($100,800 of taxable income for joint filers in 2026; $50,400 single).
2. Check two guardrails: the ACA subsidy line if you buy Marketplace insurance, and the IRMAA line if you are 63 or older (IRMAA looks back two years).
3. Take or convert tax-deferred money up to the target.
4. Fund remaining spending from taxable money, then Roth.
5. In Q4, revisit with actual numbers and add any gain harvesting that still fits at 0%.

That is a default, not a law of nature. Run your own numbers with the Roth conversion calculator, and see the annual withdrawal checklist for the calendar version.