Safe withdrawal rates: beyond the 4% rule
Key takeaways
04 · ideas- 4% rule has 95%+ success rate over 30 years of historical data
- Early retirees (retiring at 40) may need 3.0–3.5% to last 50+ years
- Flexible spending rules increase safe withdrawal rates by 0.5–1%
- Sequence of returns risk matters most in the first 10 years
Basic level
The essentials · 1 min
The 4% rule says: if you withdraw 4% of your portfolio in year 1 and adjust for inflation each year, your money should last at least 30 years. Based on historical US stock and bond returns since 1926, this has worked ~95% of the time.
Simple example: With $1,000,000 saved, you withdraw $40,000 in year 1. If inflation is 3%, you withdraw $41,200 in year 2, $42,436 in year 3, and so on.
The "failure" cases were mostly retirements that started in 1929 (market crash) or 1966 (high inflation decade). Even these required hitting extraordinary bad luck twice in a row.
Rule of thumb: To retire comfortably, save 25× your annual expenses. At $50,000/year expenses → need $1,250,000.
Intermediate level
How it applies · 2 min
**US-specific accounts:** References to 401(k), IRA, Roth, and similar accounts are specific to the US tax system. Equivalent vehicles vary by country. The underlying mathematical concepts apply universally.
Where the 4% rule came from
The Trinity Study (1998) backtested withdrawal strategies from 1926 to 1995. Portfolio: 50% S&P 500, 50% bonds. Key findings:
| Withdrawal rate | 30-year success rate |
|---|---|
| 3% | 100% |
| 4% | 98% |
| 5% | 80% |
| 6% | 68% |
| 7% | 47% |
The 1998 study used 1926–1995 data. Subsequent updates (through 2022) show the 4% rule still holds for 30-year retirements.
When 4% is too aggressive
Early retirement (40–50 years): A 50-year retirement needs a lower rate. Historical data suggests 3.0–3.5% for 50-year periods. Retiring at 40 with $1M: withdraw $30,000–$35,000/year.
International stocks: US-heavy portfolios had higher returns. Global portfolios suggest 3.5% is safer.
High valuations (CAPE > 30): The Shiller CAPE ratio above 30 historically predicts lower 10-year returns. Some researchers recommend 3.0–3.5% when valuations are stretched.
The sequence of returns: A 30% crash in year 1 of retirement is far more damaging than the same crash in year 20. Early bad returns force you to sell shares at low prices to cover expenses, depleting principal permanently.
Dynamic withdrawal rules
Fixed 4% is conservative. Flexible rules allow higher initial withdrawals:
The Guyton-Klinger rules: Cut spending 10% if the portfolio drops significantly, increase by less than inflation in good years. Allows ~5% initial withdrawal rate with high success.
The floor-and-upside approach: Set a "floor" (minimum spending, covered by Social Security or bonds), and let variable spending from equities adjust with markets.
Advanced level
The math, step by step · 2 min
The mathematics of retirement sustainability
For a portfolio P_0 with withdrawal W_0 growing at inflation rate g, invested at return r:
Portfolio after t years:For the portfolio to survive (never reach zero), we need P_t > 0 for all t \leq T.
The critical insight: sequence of returns variance matters as much as average returns. Two portfolios with identical 30-year average returns but different sequences produce different outcomes. Monte Carlo simulation — running thousands of random return sequences — gives more realistic estimates than historical simulation.
Safe withdrawal rate as a function of time horizon:
| Horizon | Equity/Bond | SWR (90% success) |
|---|---|---|
| 20 yr | 50/50 | 5.4% |
| 30 yr | 60/40 | 4.0% |
| 40 yr | 70/30 | 3.5% |
| 50 yr | 80/20 | 3.2% |
The Kitces "rising equity glidepath": Starting retirement with 30% equity and gradually increasing to 60-70% over the first 10 years paradoxically improves outcomes. The lower equity allocation in early years reduces sequence risk when it matters most. After the portfolio survives the dangerous early years, increasing equity captures upside.
Where E(t) is equity allocation at year t. Research suggests E_{initial} = 30\%, E_{final} = 75\%, T = 10 years.