The 4% rule has a tax blind spot
TL;DR. The 4% studies asked how much you can withdraw. Most retirees need to know how much they can spend. Unless the savings are entirely in a Roth IRA, those are different numbers.
The 4% rule is still a good answer to the question it was built for: how much can a retiree take from savings each year and still last 30 years. William Bengen tested it against market history in 1994, the Trinity study confirmed it in 1998, and researchers have been updating it ever since, including Morningstar’s latest look. Take 4% of your savings in the first year of retirement, raise that dollar amount with inflation, and the money has lasted 30 years in nearly every stretch of history anyone has checked. Their “safe” numbers differ a little because they test slightly different stock/bond mixes, years of history, and fees. For withdrawals, that work is still the right reference.
What they all measured is a safe withdrawal rate: how much can leave the account. For a Roth IRA, a withdrawal is what you spend. For a traditional IRA, a 401(k), or a brokerage account, it is not. Tax is paid from the same savings, so to spend 4% you have to withdraw more than 4%.
That second figure is a safe spending rate, and it is the one most retirees actually need. Many retirees hold much of their savings in 401(k)s and traditional IRAs, not Roths.
We asked that spending question for 30 years, age 60 to 90, through 1,000 simulated market futures. Every percentage below is a link. Open it, change the state, the mix, or the spending, and you are looking at your own version of the same test. Nothing is uploaded, and you do not need an account.
The tax-free starting point
One person, age 60, planning to 90. $1,000,000 in a Roth IRA. Spending $40,000 a year in today’s dollars, raised with inflation. Half stocks, half bonds. No Social Security, no pension, no other savings.
Across 1,000 simulated futures, the money lasted in 93.1% of them.
The 50/50 mix is deliberate. It was one of the allocations the classic studies tested.
If you know the Trinity study, you may remember 95% for this mix, not 93%. The gap is about method, not taxes. The study replays the few dozen 30-year stretches US markets have actually had. Our figure comes from 1,000 simulated futures, which can produce market sequences that historical replay cannot.
Same 4% spending, six ways
The same person, the same 4% spending, six places the money could sit. Two household sizes: $1 million spending $40,000, and $5 million spending $200,000. Both are 4%.
| Where the money is | $1M, spending $40,000 | $5M, spending $200,000 |
|---|---|---|
| Roth IRA, any state | 93.1% | 93.1% |
| Brokerage, bought recently, Texas | 92.9% | 87.2% |
| Brokerage, bought long ago, Texas | 92.7% | 80.0% |
| Traditional IRA, Texas | 89.1% | 75.5% |
| Brokerage, bought long ago, Oregon | 86.3% | 68.8% |
| Traditional IRA, Oregon | 83.3% | 53.8% |
Each figure is the share of 1,000 simulations in which the money lasted to age 90. “Bought long ago” means the account is worth ten times what was paid in, so 90% of every sale is profit. “Bought recently” means none of it is profit yet.
Texas has no state income tax, so those rows are federal tax only. Oregon adds a state tax on this income, including capital gains, which is why those two rows are worse.
Traditional IRAs and 401(k)s
Every dollar out of a traditional IRA or 401(k) is taxable income. To spend $40,000, you have to take out more than $40,000, and the extra is gone.
| To spend | $40,000 | $200,000 |
|---|---|---|
| From a Roth IRA, withdraw | $40,000 (4.0%) | $200,000 (4.0%) |
| From a traditional IRA in Texas | about $43,000 (4.3%) | about $252,000 (5.0%) |
| From a traditional IRA in Oregon | about $46,000 (4.6%) | about $292,000 (5.8%) |
A typical first year, in today’s dollars, on 2026 tax tables.
In Texas, the IRS takes about $3,000 a year at the $40,000 spending level, so the 4% spending rate is a 4.3% withdrawal rate. In Oregon the combined bill is about $6,300, a 4.6% withdrawal rate. At $200,000 of spending the gap is wider: about $52,000 of federal tax in Texas, about $92,000 combined in Oregon, which is a 5.0% withdrawal in one place and a 5.8% withdrawal in the other. Thirty years of taking more out is why the traditional IRA rows sit below the Roth.
This is the case the famous studies skipped, and the one that matches how most people actually hold retirement savings.
Brokerage accounts and cost basis
A brokerage account that is 90% unrealized profit sounds expensive. At $40,000 a year of spending in Texas it is not: the chance of success is 92.7%, almost the same as the Roth. At that income the federal tax on long-term capital gains is 0%, and it stays 0% until taxable income passes $49,450 for a single filer. Dividends cover part of the spending before anything has to be sold, and what does get sold is taxed at nothing. This household’s federal bill is about $100 a year.
Oregon does not have that 0% bracket. It taxes the gain as ordinary income, so the same “bought long ago” household drops from 92.7% to 86.3%.
At $200,000 a year the old gain starts to matter even in Texas: 80.0%, against 87.2% for an account bought recently. The profit in every sale is now taxed at 15%, and a 3.8% surtax on investment income kicks in once income passes $200,000. Move that household to Oregon and it falls to 68.8%. The embedded gain everyone worries about is mostly harmless in a 0% federal bracket, and costly once spending is large enough to leave that bracket, or once a state taxes gains as wages.
$200,000 is not five copies of $40,000
On the Roth row, $1 million or $5 million, the answer is 93.1%. Without taxes, the 4% rule does not care how big the numbers are.
Taxes do, because they are progressive: a bigger withdrawal is taxed at a higher rate, not only a bigger amount. Every other row gets worse as the household gets richer, and the traditional IRA in Oregon falls hardest. At $5 million that household withdraws about $292,000 to spend $200,000. Roughly $92,000 a year goes to taxes. The chance of running out goes from 7% on the Roth to 46%. The 4% rule budgeted the $200,000 of spending. It did not budget the tax on top.
What this is, and what it isn’t
None of this is a recommendation, and 93% is not a target. It is what this model does with these inputs: 2026 federal and state tax tables, 1,000 market paths built from US returns since 1928, one person, all savings in a single account, no Social Security or other income. Real retirements mix accounts, and the order you draw from them matters. Add a benefit, change the mix, or spread the money around, and the figures will move.
The engine always uses the standard deduction. It applies a state’s personal exemption credit where the tables have one, and not other credits. Many states also exempt some retirement income; Oregon does not, for IRA withdrawals. The tax page lists which state rules are in the model and which are not.
The number worth watching is not our 93% against someone else’s 95%. It is the gap between two versions of a plan. Ten points between a Roth and an Oregon IRA at $1 million is a tax effect. Thirty-nine points at $5 million is the same effect, larger, because the tax rate went up.
Try it with your own numbers
Every plan above opens in the app, free, with no account. Change the one thing you doubt: move the Oregon household to Florida, split the money between a Roth and an IRA, add Social Security. If you would rather start from your own situation, the quick start asks six questions and builds a real plan from them.