Retirement stress test
Will the 4% rule work for me?
The published 4% rule measured an inflation-adjusted withdrawal on historical retirement cohorts. This calculator can run a similar question only in Historical cycles mode, and only inside 1975–2024. Shuffled years are a different statistic.
This is the same setup as the calculator’s “Classic 4% check”: $40,000 on $1,000,000, 50% stocks, 0% tax, spending raised with inflation, Social Security off. The link still carries the form’s unused $2,000 and $1,500 benefit fields. They are not in the result while the boxes are off.
- Sample
- 1975–2024, 50 years
- Assets
- S&P 500 and 10-yr Treasuries
- Not in the sample
- 1966 and 1973–74
- Framing
- An illustration, not advice
Bengen’s 4% figure, and the Trinity study that followed, asked whether a portfolio survived 30-year historical retirements when the first withdrawal was 4% and later withdrawals rose with inflation. The stock and bond history in those studies includes the bad starts of the late 1960s and the 1973–74 crash. FIREcalc’s table starts in 1975. It does not include 1966 or 1973–74. A success rate here is not a full-history Trinity study, even when every window in this file survives.
The textbook setup, on this sample
$1,000,000, spending $40,000 after tax, which is 4% of the portfolio. Stocks 50%, 10-year Treasuries 50%. Tax rate 0%, so the portfolio draw is $40,000, not a grossed-up amount. Spending rises with CPI starting in the second year. Horizon 30 years. Social Security off. Historical cycles only.
A 30-year retirement has 21 complete windows here, 1975–2004 through 1995–2024. On this textbook plan, 21 of 21 windows still had a balance above zero at the end of every year. That is a real result of this engine and this sample. It is not evidence that 4% survived every US retirement in the literature, because the hardest starts are before the table.
- Portfolio
- $1,000,000
- First-year spending
- $40,000
- Stocks / Treasuries
- 50% / 50%
- Tax
- 0%
- Social Security
- Off
- Windows
- 21 of 21
The same inputs as shuffled years, seed 246813, kept a balance through all 30 years on 984 of 1,000 paths. That 98.4% is not the 4% rule. Shuffled years pull each year independently from the table. They do not replay the decade that actually followed a retirement date. Use them as a different stress test, and do not quote them as Trinity.
The ways a “4%” test stops being 4%
- A tax rate. Simple tax turns spending into spending ÷ (1 − rate). At 15%, $40,000 becomes about $47,059 from the portfolio, which is 4.7%.
- Social Security or a pension. Other income reduces what the portfolio pays. The success rate goes up and is no longer a portfolio-only withdrawal test. The unused $2,000 and $1,500 fields on the share link do nothing until you check the boxes.
- A longer horizon. A 40-year retirement has 11 complete windows in this table. A 50-year retirement has one. An early-retirement 4% test is a thinner sample, not a stricter version of the 30-year study.
- A different mix. The non-stock sleeve is that year’s 10-year Treasury total return. It lost 17.83% in 2022, the same year stocks lost 18.04%. It is not a total-bond index, and there is no fee.
Named stress packs on the textbook plan
The retirement results include six named packs. Each replays one real stretch, in order, through the same engine: spending, the 50/50 mix, 0% tax, inflation, and Social Security off. A pack is one path. It is not a success rate. On this textbook plan, none of the six packs ran out of money. Several of them are shorter than 30 years because the table ends in 2024, so “did not run out” is not a 30-year verdict for those packs.
- Stagflation squeeze, 1977–2006, 30 years. The balance in today’s dollars fell to 61% of the start in 1981, then finished ahead of where it began. This is the late-1970s stretch only. It is not 1966 or 1973–74. Open it.
- Rate wall of ’81, 1981–2010, 30 years. Stocks fell in year one while inflation was still high. The path finished ahead. Open it.
- Bond rout of ’94, 1994–2023, 30 years. Treasuries lost money and stocks went nowhere in the first year. The path finished ahead. Open it.
- Dot-com hangover, clipped to 2000–2024, 25 years. Three down stock years at the start. The path did not run out, and it finished smaller in today’s dollars than it started. A 30-year dot-com retirement does not fit in the table. Open it.
- GFC crash, 2008–2024, 17 years. 2008 is the worst stock year in the table, −37%. This pack cannot answer a 30-year question. The path did not run out inside the years the file has. Open it.
- 2022 rate shock, 2022–2024, 3 years. Stocks and Treasuries fell together. The path finished smaller and did not run out. Three years is not a retirement test. Open it.
Those links include the textbook inputs. A challenge link such as /?pack=dotcom names the pack only. The browser then runs it with the plan already saved on that device, which might be someone else’s numbers if you opened their share link first.
If your spending, tax, mix, or horizon differ, open the classic check and change one field. The 21-of-21 result belongs to the textbook case, not to whatever plan is on your screen after you edit it.
Questions people ask next
Did the 4% rule “work” in FIREcalc?
On the textbook inputs, in Historical cycles mode, 21 of 21 complete 30-year windows in the 1975-2024 sample stayed above zero. That sample omits 1966 and 1973-74, so it is not the published Trinity or Bengen study.
Why not use shuffled years for a 4% test?
Shuffled years are a different statistic. Each year is an independent draw. The 984 of 1,000 result on these inputs is not a historical-cohort success rate.
What is a stress pack?
One named stretch of the real table, replayed in order, for the plan’s length or until 2024. It is one path, not a success rate. The dot-com, GFC, and 2022 packs are shorter than 30 years.
Does a 15% tax still count as 4%?
No. Simple tax grosses the withdrawal up by dividing by 1 minus the rate. A 4% test in this tool uses a 0% tax rate, or you have to type a smaller spending number so the pre-tax draw is 4%.