Where 4% actually came from
It is the answer to a question almost nobody is asking: what would have survived the single worst retirement in a century, taken by a person who never once adjusted?
In 1994 William Bengen ran a plain experiment. Take a portfolio of US large-cap stocks and government bonds. Retire in 1926, then in 1927, then in every year after that. Withdraw a fixed percentage in year one, raise it by inflation every year after, and never change course. How high could that first percentage go and still last thirty years?
The answer, for the unluckiest start in the record, was 4.15%. Every other start year tolerated more. That is the entire origin of the number.
Four things happened to it on the way to becoming a rule
It got rounded down for marketing. 4.15% became "the 4% rule", shaving off a sixth of a percentage point for no reason other than that 4 is a rounder number than 4.15.
It got trimmed again for safety. The number is a worst case that already assumes the worst case, and people trimmed it anyway — to 3.5%, to 3.25%, and in parts of the FIRE community into the 2s. Each trim was individually reasonable-sounding and collectively they compound into a materially smaller life.
It got told as if it never failed. The popular version is that 4% always worked. It didn't. The Trinity Study — the other paper everyone cites — reported success rates in the mid-90s, not 100%. Run it on this site's data and a flat 4% plan fails four of the sixty-nine complete 30-year windows since 1928, all of them mid-1960s starts. That is a good result. It is not a guarantee, and it was never advertised as one by the researchers.
It got treated as a constant of nature. It is an artifact of data choices. Bengen's bond sleeve was Ibbotson intermediate-term government bonds. This site uses a Treasury blend matched to the duration of Vanguard's Total Bond Market fund, and gets 3.80% for that same worst year instead of 4.15%. Same question, different index, a third of a percentage point of disagreement. Anyone quoting the figure to two decimal places is quoting a data vendor, not a law.
What the rest of the record says
Focus on the worst year long enough and you forget it is one year. Across every complete 30-year window since 1928, on this site's data:
- The worst start, 1966, sustained 3.80%.
- The median start sustained about 6.5%.
- The best start, 1982, sustained nearly 9.9%.
So a rule calibrated to 1966 tells the median retiree to live on a little over half of what they could have spent. Not for a year — for the whole retirement. That is the cost of the rule, and it is never quoted alongside it.
1966 is also not the disaster people picture. It wasn't a crash. It was fifteen years of inflation quietly eating a portfolio while the retiree kept taking inflation-adjusted raises out of it.
So is the rule useless?
No — it is a good answer to its own question. If you want to know whether you can retire without ever having to think about money again, the 4% rule tells you, and the answer is worth having.
It is a bad answer to the question most people are actually asking, which is "how much can I spend?" Because the assumption doing all the work is not the 4%. It is the clause that says you will never adjust. Remove that one assumption — grant that a real person facing a 40% drawdown will notice and take less that year — and the sustainable number moves a long way up.