Every major peak-to-trough decline for seven stock/bond mixes — 100/0 to 0/100 — measured on one consistent monthly dataset from 1871 to June 2026: how deep, how long down, and how long back, before and after inflation.
The 2008 answer first: on this dataset — Shiller S&P composite total returns plus constant-maturity 10-year Treasuries, rebalanced monthly — a 60/40 portfolio fell 27.1% peak-to-trough in the financial crisis: 17 months down from October 2007 to March 2009, then 19 months back, regaining the prior peak by October 2010. Adjusted for inflation: 28.4%, recovered July 2011. (Daily-price measurements with a blended bond benchmark run deeper — roughly 34%, the figure used in our 2008 stress-test article; the methodology section explains exactly why the conventions differ.)
The context most pages skip: 2007–09 was only the second-deepest nominal 60/40 drawdown since 1871 — 1929–32 reached 61.0% — and in purchasing-power terms it ranks sixth. The 1973–74 episode was 25.1% nominal but 37.1% real, and took until February 1985 to recover its purchasing power. Every number on this page comes from the published dataset.
Peak-to-trough declines in dollar terms for a 60% stock / 40% bond portfolio, monthly rebalanced, dividends and coupons reinvested. "Months back" counts from the trough to the month the prior peak was regained, with no withdrawals along the way.
| Peak | Trough | Depth | Months down | Recovered | Months back | Peak to peak |
|---|---|---|---|---|---|---|
| Sep 1929 | Jun 1932 | -61.0% | 33 | Jul 1936 | 49 | 82 |
| Oct 2007 | Mar 2009 | -27.1% | 17 | Oct 2010 | 19 | 36 |
| Mar 1937 | Apr 1938 | -25.7% | 13 | Mar 1943 | 59 | 72 |
| Jan 1973 | Sep 1974 | -25.1% | 20 | Jan 1976 | 16 | 36 |
| Sep 1906 | Nov 1907 | -20.9% | 14 | Nov 1908 | 12 | 26 |
| Mar 1876 | Jun 1877 | -19.0% | 15 | Sep 1878 | 15 | 30 |
| Dec 2021 | Oct 2022 | -18.6% | 10 | Feb 2024 | 16 | 26 |
| Aug 2000 | Feb 2003 | -18.2% | 30 | Feb 2004 | 12 | 42 |
| Nov 1968 | May 1970 | -17.0% | 18 | Jan 1971 | 8 | 26 |
| Nov 1916 | Dec 1917 | -17.0% | 13 | Mar 1919 | 15 | 28 |
| Aug 1987 | Dec 1987 | -15.6% | 4 | Feb 1989 | 14 | 18 |
| Sep 1902 | Oct 1903 | -15.3% | 13 | Oct 1904 | 12 | 25 |
The 2007–09 episode sits second — well behind 1929–32 — and the 2021–22 decline (18.6%, recovered February 2024) ranks close to the long-run median for depth despite being driven by the bond side as much as the stock side, the only entry in the table for which that is true.
The identical 60/40 series deflated by CPI. Inflation reshuffles the ranking: deflationary episodes (1929–32) get milder, inflationary ones (1915–20, 1973–74, 1946–48, 2021–22) get deeper and much longer. This is the version that matters for spending power in retirement.
| Peak | Trough | Depth | Months down | Recovered | Months back | Peak to peak |
|---|---|---|---|---|---|---|
| Sep 1929 | Jun 1932 | -50.4% | 33 | Aug 1935 | 38 | 71 |
| Dec 1915 | Jun 1920 | -43.7% | 54 | Aug 1924 | 50 | 104 |
| Jan 1973 | Dec 1974 | -37.1% | 23 | Feb 1985 | 122 | 145 |
| Feb 1937 | May 1942 | -30.9% | 63 | Feb 1945 | 33 | 96 |
| Apr 1946 | Feb 1948 | -29.2% | 22 | Aug 1951 | 42 | 64 |
| Oct 2007 | Mar 2009 | -28.4% | 17 | Jul 2011 | 28 | 45 |
| Sep 1906 | Oct 1907 | -25.0% | 13 | Jan 1909 | 15 | 28 |
| Nov 1968 | Jun 1970 | -24.1% | 19 | Mar 1972 | 21 | 40 |
| Dec 2021 | Oct 2022 | -23.8% | 10 | Dec 2024 | 26 | 36 |
| Aug 2000 | Feb 2003 | -22.8% | 30 | Oct 2006 | 44 | 74 |
| Jun 1876 | May 1877 | -17.7% | 11 | Nov 1877 | 6 | 17 |
| Jun 1901 | Oct 1903 | -16.9% | 28 | Jan 1905 | 15 | 43 |
| Aug 1987 | Dec 1987 | -16.3% | 4 | Jun 1989 | 18 | 22 |
Two episodes deserve special attention. The 1973–74 decline took 145 months peak-to-peak in real terms — a retiree who hit that window spent over a decade underwater. And 2007–09, famous as it is, ranks sixth here: inflation was low, which made the real recovery only 28 months from the trough.
Worst-ever drawdown and the two most recent major episodes for each mix. The 2007–09 column shows why bonds earned their reputation as ballast — and the 2021–24 column shows the limit of that reputation: in an inflation shock the bond side stops helping, so the column barely improves as stocks are dialed down.
| Allocation (stocks/bonds) | Worst ever — nominal | Worst ever — real | 2007–09 (nominal) | 2021–24 (nominal) | Declines ≥15% since 1871 |
|---|---|---|---|---|---|
| 100/0 (all stocks) | -81.8% (Jun 1932) | -76.8% (Jun 1932) | -49.0% | -19.3% | 22 |
| 80/20 | -73.1% (Jun 1932) | -65.8% (Jun 1932) | -38.9% | -18.9% | 17 |
| 70/30 | -67.5% (Jun 1932) | -58.7% (Jun 1932) | -33.2% | -18.7% | 14 |
| 60/40 (traditional) | -61.0% (Jun 1932) | -50.4% (Jun 1932) | -27.1% | -18.6% | 12 |
| 50/50 | -53.3% (Jun 1932) | -43.7% (Jun 1920) | -20.7% | -18.4% | 8 |
| 40/60 (conservative) | -44.3% (Jun 1932) | -43.8% (Jun 1920) | -14.6% | -18.3% | 4 |
| 0/100 (all bonds) | -25.0% (Oct 2023) | -59.5% (Sep 1981) | -9.2% | -25.0% | 1 |
Worst-ever columns show depth and trough date. 2007–09 and 2021–24 columns are the deepest drawdown reached inside those windows (2007–2012 and mid-2021–2025), whether or not it exceeded 15%. Episode counts use the nominal series.
An all-bond portfolio shows a single 15%+ nominal decline in 155 years — and it is the current one: 25.0% by Oct 2023, the deepest nominal bond drawdown in the sample, not yet recovered as of June 2026. In real terms bonds are no refuge at all: a 59.5% loss of purchasing power from 1941 to 1981, with the 1941 peak not regained until November 1989.
Percent below the highest value the portfolio had reached up to each month. 0% means at an all-time high.
Same measure as selected above. The 2007–09 episode is a stock event the bonds cushioned; 2021–24 is the only major episode in the sample where the bond side was the event.
"—" in the Recovered column means the prior peak had not been regained by June 2026 (the end of the data).
Depth and duration are separate risks. For a 60/40 mix, 1973–74 was about nine points deeper than 2007–09 in real terms (37.1% vs 28.4%) — but its real recovery took more than four times as long: 122 months back from the trough versus 28. For anyone withdrawing from the portfolio, the long version is far more expensive: every month underwater is a month of selling depressed assets. That mechanism — not the headline percentage — is sequence-of-returns risk.
Before inflation, 1929–32 is the catastrophic outlier and everything since looks moderate. After inflation, the chart redistributes: 1915–20 and 1973–74 join the top tier, 1946–48 appears almost from nowhere, and 2021–22 deepens from 18.6% to 23.8%. Retirement spending is real, so the second table is the one that prices the risk.
The 2007–09 column of the allocation table rewards every step toward bonds, because Treasuries returned +19.1% in calendar 2008 while stocks lost 39.2%. The 2021–24 column barely moves as stocks are dialed down — both assets fell together. Which regime shows up next is not knowable in advance; the regime monitor tracks where the data says we are now.
Related: what 2008 did to a 60/40 retirement plan (Monte Carlo, with withdrawals and crisis correlations), the Withdrawal Strategy Lab (five spending rules on identical simulated paths), and sequence-of-returns risk explained.
On this page's 155-year monthly dataset: 27.1% peak-to-trough, October 2007 to March 2009, recovered by October 2010. On daily prices with a blended bond benchmark that includes corporate bonds, the same episode measures roughly 34% — the conventions, not the history, account for the difference.
1929–32: 61.0% nominal (50.4% after the deflation of the early 1930s). The financial crisis, at 27.1%, ranks second nominal and only sixth in purchasing-power terms.
17 months down, 19 months back — 36 months peak to peak in nominal terms, until July 2011 in real terms, assuming no withdrawals. With withdrawals the recovery stretches further; the 2008 stress-test article works that case through.
Inflationary episodes, much worse: 1973–74 deepens from 25.1% to 37.1% and the recovery stretches to 145 months peak-to-peak. Deflationary ones, milder: 1929–32 softens from 61.0% to 50.4%. For bonds the real picture is dominant — 59.5% of purchasing power lost between 1941 and 1981.
In deflationary declines, strongly (2008: Treasuries +19.1%, stocks 39.2%). In inflationary ones, little or not at all — 2021–24 produced the deepest nominal bond drawdown in the sample (25.0%), and a 40/60 mix fell nearly as far as an 80/20 in that window. The protection is regime-dependent, which is why the stress tester models crisis correlations explicitly.
Price sampling (monthly averages vs daily closes), bond benchmark (Treasuries vs broad indexes with corporates), and rebalancing frequency. This page holds one convention constant across 155 years so that eras are comparable with each other; daily-convention numbers for 2007–09 are deeper. Both are correct under their stated assumptions.
Occasional deep dives — drawdown history, withdrawal strategies, regime models, simulation methodology. No more than monthly. Free.