QuantCalcResearchWithdrawal Order 2026

When Withdrawal Order Helps, and When It Backfires (2026)

The classic advice is to spend taxable accounts first, then traditional, then Roth — and a "tax-smart" tool can go further, picking the single-year order that pays the least tax right now. Measured path by path across full retirements, that single-year order helps some households and quietly costs others money. Here is where it lands, for nine of them.

QuantCalc Research · Generated 2026-07-22 · v2026.1 · CC-BY-4.0 dataset

Sequencing which account you draw from in retirement is one of the most-repeated levers in tax planning. The intuition is clean: defer the tax-deferred money, spend the already-taxed money first, and let the Roth grow. A tool can sharpen that into a rule — each year, split the withdrawal across traditional, Roth, and taxable to minimize this year's tax. The question this study asks is whether that single-year rule actually lowers taxes over a whole retirement. The answer is that it depends, sharply, and in a way a single headline number would hide.

The two-sided result. For a household still in its low-income years before Social Security and required distributions, the recommended single-year order lowers lifetime taxes by a paired median of $28,807 (Balanced three-way, retire at 62, plan to 80; band +$8,180 to +$41,535). For a traditional-heavy household deep in the required-distribution years, the same single-year order raises lifetime taxes by a paired median of $10,613 (Traditional-heavy, California, plan to 90; band −$22,376 to +$2,312). Every figure is a paired per-path median: we run each order on the identical simulated market paths, take the difference on each path, and report the median of those differences — not the gap between two separately-computed averages.

Run this household in the planner →

Opens the tax-aware planner prefilled with the Balanced three-way corridor household above (Florida, retire at 62, $500,000/$200,000/$300,000 traditional/Roth/taxable) — then change the numbers to see which side of the crossover you land on.

+$28,807
Paired median where sequencing helps most (pre-RMD corridor)
−$10,613
Paired median where it backfires most (RMD tail)
4,000
Market paths per household (paired, common random numbers)
9
Households (account mix × age × state × spending)

The full picture, household by household

Each household draws the same real spending every year under both orders, so portfolio survival barely moves between them — the tax difference is what changes. "Tax-aware lifetime tax" is the median lifetime tax under the single-year order; "Traditional-first" is the median under the plain draw-traditional-first rule. The paired median is the apples-to-apples number: the median, over market paths, of (traditional-first − tax-aware) on each path. Positive means the tax-aware order wins; negative means it loses.

HouseholdAges Trad / Roth / TaxableSpend/yr Tax-aware lifetime taxTraditional-first Paired medianp10 … p90 band
Balanced three-way
Pre-RMD corridor · FL
62–80$500,000 / $200,000 / $300,000$60,000$30,089$65,561+$28,807+$8,180 … +$41,535
Taxable-heavy
Pre-RMD corridor · CA
62–80$300,000 / $150,000 / $550,000$66,000$28,537$44,569+$16,364+$5,126 … +$25,004
Traditional-heavy
Pre-RMD corridor · FL
62–82$750,000 / $100,000 / $150,000$60,000$71,019$77,294+$5,445−$26,725 … +$21,427
Roth-heavy
Pre-RMD corridor · FL
62–82$200,000 / $600,000 / $200,000$60,000$4,532$16,691+$11,971−$2,204 … +$16,867
Balanced three-way
RMD straddle · FL
70–88$650,000 / $250,000 / $300,000$72,000$79,132$88,051+$9,124−$10,478 … +$43,753
Traditional-heavy
RMD straddle · CA
70–90$900,000 / $120,000 / $180,000$72,000$180,601$169,821−$10,613−$22,376 … +$2,312
Traditional-heavy
Deep RMD tail · FL
75–95$900,000 / $100,000 / $150,000$66,000$189,438$189,438+$0+$0 … +$0
Balanced three-way
Deep RMD tail · CA
75–93$700,000 / $250,000 / $250,000$72,000$126,474$126,149+$0−$702 … +$11,593
Traditional-heavy
High-balance (IRMAA-exposed) · CA
70–90$1,800,000 / $400,000 / $600,000$144,000$284,155$275,161−$7,129−$65,910 … +$80,987

Green rows: the tax-aware order lowers lifetime taxes. Red rows: it raises them. All dollars are paired per-path medians in today's money.

Where sequencing genuinely helps

The common thread in the winning cases is a low-income corridor: the household retires before claiming Social Security and before required distributions begin, so its taxable income sits low for a stretch of years. Steering the year's withdrawal to keep income low — and to realize taxable dollars cheaply while brackets are empty — is exactly the room the single-year order exploits.

Balanced three-way · retire at 62, plan to 80 · Florida

Sequencing helps Balanced three-way · Pre-RMD corridor · Florida (no state income tax) · traditional $500,000, Roth $200,000, taxable $300,000 · $60,000/yr spending · Social Security $30,000/yr from age 67.

Here the recommended single-year order lowers lifetime taxes by a paired median of $28,807 — measured path by path, then taking the median of the per-path difference. The band runs +$8,180 to +$41,535; with the 10th percentile above zero, more than nine paths in ten come out ahead. This household spends much of its early retirement in a low-income window — before Social Security and before required distributions — and the order uses that room. Portfolio survival is 99.9% for this plan (mortality-adjusted 99.9%); both orders fund the identical real spending each year, so the tax difference is the story, not the survival difference.

Tax-aware draw (median, real $)Yr 1
age 62
Yr 2
age 63
Yr 3
age 64
Yr 4
age 65
Yr 5
age 66
From Traditional$0$0$0$0$0
From Roth$0$0$0$0$0
From Taxable$60,000$60,000$60,000$60,000$60,000
Tax paid that year$0$0$0$0$0
IRMAA surcharge$0$0$0$0$0

Taxable-heavy · retire at 62, plan to 80 · California

Sequencing helps Taxable-heavy · Pre-RMD corridor · California (taxes retirement income) · traditional $300,000, Roth $150,000, taxable $550,000 · $66,000/yr spending · Social Security $28,800/yr from age 67.

Here the recommended single-year order lowers lifetime taxes by a paired median of $16,364 — measured path by path, then taking the median of the per-path difference. The band runs +$5,126 to +$25,004; with the 10th percentile above zero, more than nine paths in ten come out ahead. This household spends much of its early retirement in a low-income window — before Social Security and before required distributions — and the order uses that room. Portfolio survival is 99.3% for this plan (mortality-adjusted 99.5%); both orders fund the identical real spending each year, so the tax difference is the story, not the survival difference.

Tax-aware draw (median, real $)Yr 1
age 62
Yr 2
age 63
Yr 3
age 64
Yr 4
age 65
Yr 5
age 66
From Traditional$0$0$0$0$0
From Roth$59,400$59,400$39,600$0$0
From Taxable$6,600$6,600$26,400$66,000$66,000
Tax paid that year$0$0$150$1,119$1,260
IRMAA surcharge$0$0$0$0$0

Traditional-heavy · retire at 62, plan to 82 · Florida

Sequencing helps Traditional-heavy · Pre-RMD corridor · Florida (no state income tax) · traditional $750,000, Roth $100,000, taxable $150,000 · $60,000/yr spending · Social Security $30,000/yr from age 67.

Here the recommended single-year order lowers lifetime taxes by a paired median of $5,445 — measured path by path, then taking the median of the per-path difference. The band runs −$26,725 to +$21,427; with the 10th percentile below zero, the benefit is real at the median but not on every path. This household spends much of its early retirement in a low-income window — before Social Security and before required distributions — and the order uses that room. Portfolio survival is 99.9% for this plan (mortality-adjusted 99.9%); both orders fund the identical real spending each year, so the tax difference is the story, not the survival difference.

Tax-aware draw (median, real $)Yr 1
age 62
Yr 2
age 63
Yr 3
age 64
Yr 4
age 65
Yr 5
age 66
From Traditional$0$0$0$0$24,000
From Roth$0$0$18,000$60,000$36,000
From Taxable$60,000$60,000$42,000$0$0
Tax paid that year$0$0$0$0$585
IRMAA surcharge$0$0$0$0$0

Roth-heavy · retire at 62, plan to 82 · Florida

Sequencing helps Roth-heavy · Pre-RMD corridor · Florida (no state income tax) · traditional $200,000, Roth $600,000, taxable $200,000 · $60,000/yr spending · Social Security $30,000/yr from age 67.

Here the recommended single-year order lowers lifetime taxes by a paired median of $11,971 — measured path by path, then taking the median of the per-path difference. The band runs −$2,204 to +$16,867; with the 10th percentile below zero, the benefit is real at the median but not on every path. This household spends much of its early retirement in a low-income window — before Social Security and before required distributions — and the order uses that room. Portfolio survival is 99.9% for this plan (mortality-adjusted 99.9%); both orders fund the identical real spending each year, so the tax difference is the story, not the survival difference.

Tax-aware draw (median, real $)Yr 1
age 62
Yr 2
age 63
Yr 3
age 64
Yr 4
age 65
Yr 5
age 66
From Traditional$0$0$0$0$0
From Roth$0$0$0$24,000$60,000
From Taxable$60,000$60,000$60,000$36,000$0
Tax paid that year$0$0$0$0$0
IRMAA surcharge$0$0$0$0$0

Balanced three-way · retire at 70, plan to 88 · Florida

Sequencing helps Balanced three-way · RMD straddle · Florida (no state income tax) · traditional $650,000, Roth $250,000, taxable $300,000 · $72,000/yr spending · Social Security $36,000/yr from age 70.

Here the recommended single-year order lowers lifetime taxes by a paired median of $9,124 — measured path by path, then taking the median of the per-path difference. The band runs −$10,478 to +$43,753; with the 10th percentile below zero, the benefit is real at the median but not on every path. This household spends much of its early retirement in a low-income window — before Social Security and before required distributions — and the order uses that room. Portfolio survival is 100.0% for this plan (mortality-adjusted 100.0%); both orders fund the identical real spending each year, so the tax difference is the story, not the survival difference.

Tax-aware draw (median, real $)Yr 1
age 70
Yr 2
age 71
Yr 3
age 72
Yr 4
age 73
Yr 5
age 74
From Traditional$0$0$28,084$29,188$30,063
From Roth$0$0$7,916$6,801$5,931
From Taxable$36,000$36,000$0$0$0
Tax paid that year$0$0$2,172$3,171$3,431
IRMAA surcharge$0$0$0$0$0

Run the corridor household yourself →

Where it barely matters

When required distributions already force most of the year's cash out of the traditional account, there is little left to sequence — both orders converge.

Traditional-heavy · retire at 75, plan to 95 · Florida

About even Traditional-heavy · Deep RMD tail · Florida (no state income tax) · traditional $900,000, Roth $100,000, taxable $150,000 · $66,000/yr spending · Social Security $42,000/yr from age 70.

Here the two orders come out within $500 of each other at the median (+$0, band +$0 to +$0). Required distributions already force most of the draw out of the traditional account, so the order has little room to change the outcome either way. Portfolio survival is 100.0% for this plan (mortality-adjusted 100.0%); both orders fund the identical real spending each year, so the tax difference is the story, not the survival difference.

Tax-aware draw (median, real $)Yr 1
age 75
Yr 2
age 76
Yr 3
age 77
Yr 4
age 78
Yr 5
age 79
From Traditional$40,292$41,297$42,418$43,724$45,235
From Roth$0$0$0$0$0
From Taxable$0$0$0$0$0
Tax paid that year$5,013$5,325$5,673$7,457$8,194
IRMAA surcharge$0$0$0$0$0

Balanced three-way · retire at 75, plan to 93 · California

About even Balanced three-way · Deep RMD tail · California (taxes retirement income) · traditional $700,000, Roth $250,000, taxable $250,000 · $72,000/yr spending · Social Security $40,800/yr from age 70.

Here the two orders come out within $500 of each other at the median (+$0, band −$702 to +$11,593). Required distributions already force most of the draw out of the traditional account, so the order has little room to change the outcome either way. Portfolio survival is 100.0% for this plan (mortality-adjusted 100.0%); both orders fund the identical real spending each year, so the tax difference is the story, not the survival difference.

Tax-aware draw (median, real $)Yr 1
age 75
Yr 2
age 76
Yr 3
age 77
Yr 4
age 78
Yr 5
age 79
From Traditional$31,428$32,307$33,370$34,517$35,520
From Roth$0$0$0$0$0
From Taxable$0$0$0$0$0
Tax paid that year$3,387$3,696$4,061$5,115$5,444
IRMAA surcharge$0$0$0$0$0

Where the same rule backfires

The losing cases share the opposite setup: a large traditional balance and a long horizon that runs deep into the required-distribution years. This is where a single-year objective and a whole-life objective pull apart.

Traditional-heavy · retire at 70, plan to 90 · California

Sequencing backfires Traditional-heavy · RMD straddle · California (taxes retirement income) · traditional $900,000, Roth $120,000, taxable $180,000 · $72,000/yr spending · Social Security $38,400/yr from age 70.

Here the recommended single-year order raises lifetime taxes by a paired median of $10,613 versus simply drawing traditional first. The band runs −$22,376 to +$2,312. Minimizing this year's tax means drawing whichever bucket is cheapest right now — often the tax-free Roth or return-of-basis taxable dollars — which preserves the traditional balance. Over a long horizon that preserved balance grows and is then forced out as required distributions at higher rates, with IRMAA surcharges on top. Portfolio survival is 100.0% for this plan (mortality-adjusted 100.0%); both orders fund the identical real spending each year, so the tax difference is the story, not the survival difference.

Tax-aware draw (median, real $)Yr 1
age 70
Yr 2
age 71
Yr 3
age 72
Yr 4
age 73
Yr 5
age 74
From Traditional$0$0$38,445$39,817$41,230
From Roth$26,880$26,880$0$0$0
From Taxable$6,720$6,720$0$0$0
Tax paid that year$0$0$5,269$6,381$7,012
IRMAA surcharge$0$0$0$0$0

Traditional-heavy · retire at 70, plan to 90 · California

Sequencing backfires Traditional-heavy · High-balance (IRMAA-exposed) · California (taxes retirement income) · traditional $1,800,000, Roth $400,000, taxable $600,000 · $144,000/yr spending · Social Security $48,000/yr from age 70.

Here the recommended single-year order raises lifetime taxes by a paired median of $7,129 versus simply drawing traditional first. The band runs −$65,910 to +$80,987. Minimizing this year's tax means drawing whichever bucket is cheapest right now — often the tax-free Roth or return-of-basis taxable dollars — which preserves the traditional balance. Over a long horizon that preserved balance grows and is then forced out as required distributions at higher rates, with IRMAA surcharges on top. Portfolio survival is 99.9% for this plan (mortality-adjusted 100.0%); both orders fund the identical real spending each year, so the tax difference is the story, not the survival difference.

Tax-aware draw (median, real $)Yr 1
age 70
Yr 2
age 71
Yr 3
age 72
Yr 4
age 73
Yr 5
age 74
From Traditional$0$0$76,891$79,634$82,461
From Roth$86,400$86,400$19,079$16,366$13,525
From Taxable$9,600$9,600$0$0$0
Tax paid that year$0$0$9,343$11,153$11,605
IRMAA surcharge$0$0$0$0$0

Why a "tax-smart" order can cost money

The mechanism is not exotic. A rule that minimizes this year's tax draws whichever bucket is cheapest right now. Roth dollars cost nothing this year; taxable dollars sold near their cost basis cost almost nothing. So the single-year order tends to spend Roth and taxable early and leave the traditional balance untouched. That feels efficient — until the required-distribution years arrive.

A large, preserved traditional balance keeps compounding, and once required distributions begin the retiree is forced to pull it out, whether or not the brackets are full. Those forced dollars land on top of Social Security, push more of it into taxable income, and can cross an IRMAA threshold that raises Medicare premiums two years later. The low-bracket corridor years — the cheapest years to have drawn that traditional money — were spent on Roth instead, and that room is gone for good. Drawing traditional first does the unglamorous thing: it empties the tax-deferred account while brackets are cheap, so the forced-distribution years are smaller.

The horizon is the hinge

The clearest way to see the trade is to hold one household constant and stretch only its planning horizon. Below is the Balanced three-way corridor retiree from above, simulated at a spread of end ages. The benefit of the single-year order is large on a short horizon, shrinks as the horizon lengthens, and inverts once the plan runs far enough into the required-distribution tail — on the identical market paths and seed each time, so nothing changes but the years.

AgesYears Tax-aware lifetime taxTraditional-first Paired medianp10 … p90 band
62–7816$19,973$61,843+$37,483+$19,743 … +$45,942
62–8018$30,089$65,561+$28,807+$8,180 … +$41,535
62–8220$40,540$63,272+$19,257−$5,749 … +$35,225
62–8523$57,867$64,909+$2,189−$33,297 … +$22,847
62–8826$78,081$65,017−$15,254−$71,274 … +$8,407
62–9028$92,482$65,175−$28,062−$102,282 … −$416

Same household, same market paths — only the planning horizon changes. The benefit of the single-year order falls monotonically and crosses zero as the required-distribution tail lengthens.

Why this study is different

Most withdrawal-order coverage quotes a single "you could save $X" figure. Two things make that fragile, and this study addresses both. First, the number here is a paired per-path median: each order is run on the same simulated markets and the difference is taken path by path, so the figure is one a real retirement could actually experience — not the distance between two separately-ranked medians, which no single path realizes. Second, the traditional-first comparison is an independent simulation with its own account balances, not a one-year hypothetical bolted onto the tax-aware path — so it is fair to say "a traditional-first retiree would pay this instead."

How to read the tables. Every dollar is a paired per-path median in today's money. The "p10 … p90 band" is the 10th and 90th percentile of the same per-path difference, so it shows the spread of outcomes across market paths, not a confidence interval on an average. A band whose low end stays above zero means the tax-aware order wins on more than nine paths in ten; a band straddling zero means the median verdict is real but individual paths can go either way.

What it means

Download CSV (per household + horizon crossover) Download JSON (full stats, order snapshots, bands)

CC-BY-4.0 — free for any use including republication and journalism, with attribution to QuantCalc Research.

Find your own crossover

Whether sequencing helps or backfires turns on your account mix, your horizon, and your state — and it can flip. QuantCalc's PRO tax-aware simulation runs both orders on the same market paths for your numbers and shows the paired difference with its band, year by year, so you see which side of the line you are on rather than trusting a rule of thumb.

Open the PRO tax-aware simulation →

Methodology

Engine. Each household is run through QuantCalc's tax-aware retirement simulation over 4,000 market paths. In each year the tax-aware order splits the required withdrawal across traditional, Roth, and taxable to minimize that year's total tax cost (federal, state, IRMAA, and any early-withdrawal penalty), subject to the required-distribution floor. This is a single-year, tax-cost-minimizing order — a recommended order for that year, not a whole-life plan — and the study measures exactly the gap between those two things.

Paired baseline. The traditional-first comparison is an independent simulation carrying its own account balances, cost-basis lots, and Medicare-income history, drawn traditional-first each year, and run on the same market and inflation path as the tax-aware order (common random numbers). Because it is a genuine parallel run rather than a one-year counterfactual on the tax-aware balances, its lifetime tax is a defensible "what a traditional-first retiree would actually pay."

Headline statistic. On each path we compute (traditional-first lifetime tax − tax-aware lifetime tax) in today's money, then report the median of those per-path differences and the 10th and 90th percentiles as a band. This paired median is not the difference between the two orders' separately-ranked median lifetime taxes — that difference is a quantity no single path realizes, and the two can even disagree in sign.

Tax detail. The engine applies 2026 statutory federal brackets, the standard and senior deductions, Social Security provisional-income taxation, state income tax by residence, IRMAA tiers on a two-year lookback, and required distributions on the SECURE 2.0 schedule. Taxable-account sales use per-lot HIFO with correct short- and long-term character. The 2026 constants are pinned identically between the engine and the calculator's own tax code by a parity test.

Verification. The per-year order comes from optimize_withdrawal, cross-checked in the rig backend/tools/withdrawal_verify.c against an independent replica of its search grid and a hand-computed 2026 federal-tax anchor, and the paired-median statistic itself is re-derived independently there and asserted to equal the engine's. Every figure on this page is emitted directly by the engine via backend/tools/withdrawal_study_dump.c; the generator recomputes the bands and the crossover monotonicity and aborts on any mismatch, so the page cannot drift from the data.

Reproducibility. The first persona was simulated twice and returned a bit-identical paired median (difference 0) and matching lifetime-tax medians for both orders. The simulation is seeded and serial, so re-running reproduces every figure.

Model-dependence. These are modeled results under a specific set of 2026 tax rules, a modeled return distribution, and stated household assumptions — not personalized advice. Your brackets, state, health-coverage timing, and longevity all move the verdict. Consult a qualified tax professional before acting.

Changelog

Dataset license: CC-BY-4.0. QuantCalc is an independent retirement-planning research project. Model-dependent results under 2026 tax rules; not financial, tax, or legal advice.

Get new research and dataset updates