Educational information, not individual financial advice.
Key Takeaways
In retirement, which accounts you draw from — and in what order — is one of the biggest levers on your lifetime tax bill and how long your money lasts. Two retirees with identical portfolios can pay very different lifetime taxes based on sequencing alone.
A withdrawal from a traditional (tax-deferred) account is taxed as ordinary income. So to net a given amount, you have to withdraw more. To net $80,000 at a 20% effective rate:
Gross withdrawal = $80,000 ÷ (1 − 0.20) = $100,000 — with $20,000 going to taxes.
Roth withdrawals, by contrast, are tax-free, and taxable-account withdrawals are taxed only on the gains. The mix matters.
The textbook sequence is taxable → tax-deferred → Roth last: spend the least tax-advantaged money first, let tax-advantaged accounts keep compounding, and preserve the Roth (no required distributions, best to leave to heirs).
It's a fine default, but a strict version has a flaw: leaving the pre-tax balance untouched lets it grow into a large RMD "tax bomb." Once required minimum distributions begin (age 73/75), that balance can force out big taxable withdrawals — possibly pushing you into a higher bracket and triggering IRMAA Medicare surcharges.
The sophisticated move is to smooth your tax rate across retirement instead of front-loading low-tax years and back-loading high-tax ones:
The gap years between retirement and the start of Social Security and RMDs are the prime window: your income is low, so filling the 12% and 22% brackets cheaply can keep you out of the 32% bracket later. Fidelity's modeling has found proportional strategies cutting lifetime taxes substantially versus naive sequencing.
Horizons grosses up withdrawals for taxes and lets you model withdrawal order and gap-year Roth conversions, so you can see the lifetime-tax and longevity impact of your sequencing.
Why is the strict 'spend taxable first, Roth last' order often not the most tax-efficient?
Try it in your scenario
Try this next
Required Minimum Distributions (RMDs)
More related reading