A near-retiree with a large portfolio and a plan to stop working next year asks a forum what happens if the market falls in the first year. The replies are kind and abstract: keep a few years of expenses in cash, build a bond tent, write an investment policy statement. Nobody walks the poster's own numbers through an actual crash. One reply notes that “nobody will ever say yes” — the question has no answer people are willing to write down.
This story writes it down. Take a 1,000,000 portfolio at retirement, withdraw 40,000 a year, and retire on the eve of four real crashes: September 1929, January 1973, October 2007 and January 2022. Hold zero, two or five years of withdrawals in savings at 3%. Compare each with the future in which nothing happens and the market grows 7% a year. Ten years of each, in price terms, no dividends.
Sequence risk is not “the market might fall.” It is “the market might fall while I am selling.” The order of returns matters only because withdrawals are in the way.
Retiring into 2007 — year ten, four ways
S&P 500 price path from the 9 October 2007 closing peak (−56.78% to the 9 March 2009 low, regained 28 March 2013), then 7% a year. Price only: the dividends a real portfolio paid through 2008–13 are not included and would have helped.
The crash costs a decade. The bucket buys back part of it.
Retiring into 2007 with everything in equities and no reserve, the portfolio bottoms at 404k — less than half — and ends year ten at 756k, having paid every withdrawal. In the no-crash future the same person ends near 1,397k. The crash did not ruin the plan; it cost about 641k of ending wealth and a few very bad years of statements.
Two years of withdrawals in savings raises the low point to 429k and the year-ten value to 812k. Five years raises them to 501k and 876k. The bucket works the way the replies say it does — it stops the retiree selling equities at the bottom — but look at the other column: in the no-crash future the five-year bucket ends at 1,359k instead of 1,397k. The reserve is an insurance premium paid every year the crash does not come, and most years it does not.
| Retiring into | Savings bucket | Low point | Year 5 | Year 10, crash future | Year 10, no-crash future | Withdrawals missed |
|---|---|---|---|---|---|---|
| 1929 · Wall Street crash | 0 yrs | 0k | 111k | depleted in month 106 | 1,397k | 48k |
| 2 yrs | 0k | 131k | depleted in month 114 | 1,391k | 21k | |
| 5 yrs | 127k | 244k | 135k | 1,359k | — | |
| 1973 · Oil shock & stagflation | 0 yrs | 507k | 555k | 639k | 1,397k | — |
| 2 yrs | 524k | 577k | 674k | 1,391k | — | |
| 5 yrs | 583k | 614k | 732k | 1,359k | — | |
| 2007 · Global financial crisis | 0 yrs | 404k | 670k | 756k | 1,397k | — |
| 2 yrs | 429k | 708k | 812k | 1,391k | — | |
| 5 yrs | 501k | 751k | 876k | 1,359k | — | |
| 2022 · Inflation & rate-hike bear market | 0 yrs | 753k | 987k | 1,147k | 1,397k | — |
| 2 yrs | 769k | 1,004k | 1,171k | 1,391k | — | |
| 5 yrs | 798k | 1,003k | 1,169k | 1,359k | — |
1929 is the case the buckets cannot fix — in price terms
Retire into September 1929 on these assumptions and the price path does what no bucket survives: −86.22% to June 1932, and no return to the old level for 25.0 years. With no reserve the portfolio is empty in month 106. With two years in savings, empty in month 114. With five, 135k by year ten, still paying.
That row needs its own caveat, and it is the most important sentence on this page. The replay uses the price index. The companion story on recovery times shows that with dividends reinvested the 1929 investor was back to even in 184 months rather than 300, and in purchasing power in 86, because dividends kept flowing and prices deflated. A retiree drawing 4% in the 1930s was also drawing dividends of that order. The price-only 1929 row is therefore a worst case that is worse than what happened to real money; it is here because it is the historical shape the forum fears, not because it is a fair account of a 1929 retiree.
1973 is the quieter warning. The price path fell 48.2% and came back in 7.5 years, and the no-bucket retiree ends year ten at 639k — solvent, but with less than the 1,397k of the no-crash future, and having lived through inflation this replay ignores. 2022 barely registers: 1,147k against 1,397k with no bucket at all.
What the forum answers get right, and what they leave out
“Keep two to five years in cash” is good advice, and the table shows why: it converts a forced sale at the bottom into a sale at a better price later. What the advice leaves out is the cost in the other future, which the table also shows, and the fact that the bucket does nothing for a decline that lasts longer than the bucket. Five years of reserves were gone by the sixth year of 1929 and of 1973 alike; only the market's own recovery, or a cut in spending, carries the retiree after that.
“Nobody will ever say yes” is also right. No replay can promise that a plan survives, because the next path is not in this table. What the replay can do is turn a feeling — “I have no time to recoup” — into a set of year-by-year numbers that a person can react to before the year arrives: how low is too low, at what point spending gets cut, whether the reserve is a bucket or a bond tent, and what is done with it in the future where nothing happens.
Put your own numbers on the path
The calculator runs the same two futures with your portfolio, your assumptions and your rule for moving money in and out of savings. It does not model withdrawals — that is why this replay exists as a separate, tested helper — but it will show you what the same crash does to a portfolio you are still adding to, or one you have moved partly to savings.
What this does not prove
- Price paths only. Dividends, which mattered enormously in 1929–45 and materially in 2007–13, are excluded; so are fees, taxes and inflation. The 1929 row in particular overstates what happened to real money.
- Withdrawals are a fixed nominal 40,000 a year. Real retirees cut spending in bad years; that alone changes the 1929 and 1973 rows.
- 7% a year is an assumption for the no-crash future and for the years after each observed cycle. It is not a forecast, and the comparison between futures depends on it.
- Four selected US episodes are not a probability distribution. A future sequence can be worse than 1929 in price terms, or milder than 2022.
- The savings bucket earns a constant 3%. Deposit rates in 1930 and 2009 were not 3%; the calculator's savings research models that, this replay does not.
- Nothing here is a recommendation to hold a bucket, a bond tent, or neither.
Method and sources
Each episode's normalized daily-close path (weekly-sampled with exact peak, trough and recovery anchors for long cycles) is replayed month by month: market move, then savings interest, then the withdrawal, taken from savings while any remains and then from equities. Intraday highs and lows are excluded. Research snapshot cutoff: 14 September 2026.
- S&P 500 and composite predecessor historical data — the 1929, 1973, 2007 and 2022 price episodes.
- Yardeni / S&P / Haver bear-market tables — corroboration of the 2022 decline and the conventional thresholds.
- Companion story: how long each crash took to come back in price, with dividends, and after inflation — the dividend caveat for 1929.
- Markets Panic methodology, definitions, caveats and reproducibility notes.
- Historical savings/deposit-rate methodology — why 3% is an assumption here.