Three portfolios, one crash, and the answers don't line up the way the brochures imply. Cutting equity from 100% to 60% roughly halved the depth of every big drawdown since 1928. It didn't reliably halve the time spent underwater. Twice it made the wait longer.
Here's the sharpest case. Start at the year-end 1972 peak. Across 1973 and 1974 an all-equity portfolio lost 48.0% of its real value. A 60/40 lost 35.2%, a far gentler ride. Both regained their real starting point at almost the same moment: 1984 for all-equity, 1985 for 60/40. Twelve years against thirteen. The extra bonds bought a shallower hole and not one year off the climb.
The series, and what it can't tell you
Everything below comes from one dataset. Aswath Damodaran's annual returns file at NYU Stern runs from 1928 to 2025 and carries S&P 500 total returns, a ten-year US Treasury return and CPI inflation in the same table. Three portfolios are built from it — 100/0, 80/20 and 60/40 — each rebalanced back to target at every year end.
Four caveats, and some of them matter more than the results.
The data is annual. A peak-to-trough measured on calendar years is always shallower than the real one, because the actual bottom rarely falls on 31 December. Dimson, Marsh and Staunton's monthly work in the UBS Yearbook dates the 1929 trough to July 1932 and full recovery to February 1945, fifteen and a half years. My annual figure for the same episode is eight years. Both are honest; they measure different things. Treat every depth and duration here as a floor, not a maximum.
The bond leg is synthetic. Damodaran doesn't use a traded bond index. He takes the constant-maturity ten-year Treasury yield, reprices a par bond at the new yield, and adds the coupon. That's a reasonable proxy and it isn't a fund you could have bought. It carries no credit risk, no fees and no spread.
The equity leg is back-filled. The S&P 500 was created in 1957. Damodaran says plainly that he splices in other large-cap indices for the earlier years, so the 1928-1956 numbers are a reconstruction rather than a live index record.
And it's one country. United States, in dollars, before costs and taxes, with free annual rebalancing. The US was the best-performing major market of the twentieth century, which makes this sample flattering to equities. Across the twenty-one markets with continuous 125-year histories, the Yearbook puts the average annualised real bond return at just 0.9%. The equity numbers below would look different almost anywhere else.
Depth scales with equity weight, and almost linearly
This part holds up exactly as you'd expect. Here are the peak-to-trough falls, measured from the year-end before each crash began, in nominal terms.
| Crash | 100% equity | 80/20 | 60/40 |
|---|---|---|---|
| 1929-32 | -64.8% | -53.5% | -40.1% |
| 1973-74 | -36.5% | -28.9% | -20.8% |
| 2000-02 | -37.4% | -24.7% | -10.6% |
| 2008 | -36.6% | -25.2% | -13.9% |
| 2022 | -18.0% | -18.0% | -18.0% |
Ignore the last row for a second. In the other four, 60/40 took between a third and a quarter of the pain that all-equity took. The 80/20 mix landed roughly two-thirds of the way toward all-equity, which is what a linear blend should do.
Worst single years tell the same story. Over the 98 years, the worst calendar year for all-equity was 1931 at -43.8%. The 80/20 lost 35.6% and the 60/40 lost 27.3%. Down years ran 26 out of 98 for all-equity, 25 for 80/20 and 21 for 60/40. Cutting equity buys fewer bad years and much smaller ones. That's the whole point of it, and it works.
Strip out inflation and the ranking survives but the gaps narrow. Real falls from the end-1928 peak were 54.8%, 42.6% and 28.6%. From the end-1972 peak they were 48.0%, 41.7% and 35.2% — a 12.8-point spread, less than half the nominal spread over the same crash. Inflation is the reason. Bonds don't defend against it, and 1973-74 was the test case.
Recovery time doesn't follow the same rule
Here's where the intuition breaks. These are the years from each pre-crash peak until the portfolio regained that level in real terms.
| Peak | 100% equity | 80/20 | 60/40 |
|---|---|---|---|
| End-1928 | 8 years | 7 years | 5 years |
| End-1972 | 12 years | 13 years | 13 years |
| End-1999 | 14 years | 8 years | 6 years |
| End-2007 | 6 years | 5 years | 3 years |
| End-2021 | 3 years | 3 years | 4 years |
Two of these five rows go the wrong way. After 1972, the 60/40 was underwater a year longer than the all-equity portfolio despite falling barely two-thirds as far. After 2021 it was underwater a year longer again. In a third row, 2007, the difference is three years against six on a crash where 60/40 lost less than 14% and all-equity lost more than 36%. Depth ratios of two or three to one collapse into duration ratios near one.
The 1973-74 case is the cleanest illustration of why. Bonds cushioned the fall and then couldn't climb out. Inflation ran hot for most of the following decade, so the 40% sitting in Treasuries dragged. All-equity fell further and then compounded faster off a lower base. The two paths crossed and finished within a year of each other. If you want the sequence in monthly detail, the piece on how long bear markets have lasted since 1968 traces it.
The 2000-2002 row is the one that goes decisively the other way. All-equity took 14 years to recover its end-1999 real value, because the tech bust and the financial crisis arrived close enough together that the second one caught the portfolio before it had healed. Its real low over that span came in 2008, not 2002. The 60/40 took six years. That's a genuine and large win for bonds, on the single worst equity sequence in modern US data.
2022 broke the mechanism entirely
In 2022 the S&P 500 returned -18.0% and the ten-year Treasury returned -17.8%. Inflation ran at 6.4%. All three portfolios lost about 18% nominal and about 23% real. The bond weight did nothing at all, because the thing bonds are supposed to hedge is a growth shock, and this was a rate shock that hit both legs together.
That isn't a fluke of one year. The Yearbook measures the long-run stock-bond correlation at 0.33 across countries and 0.19 in the US. Negative correlation, the version most investors have internalised, was a feature of roughly the late 1990s to 2021, and it ended. A 60/40 built on the assumption of negative correlation is a different instrument from one built on the assumption of low positive correlation. Related evidence on which asset classes actually diversified during crises points the same way.
What the long horizon paid
Compounding is where the three separate for good. $100 invested at the end of 1927, held to the end of 2025 with annual rebalancing, finished at $61,445 in real terms at 100% equity, $31,542 at 80/20 and $13,652 at 60/40. Annualised real returns: 6.77%, 6.05% and 5.14%.
A 1.63-point gap over 98 years is a 4.5-fold difference in terminal wealth. That's the price of the shallower drawdowns in the first table. Whether it's worth paying isn't a question the data answers, and it depends far more on the gap between risk tolerance and risk capacity than on any historical average.
Independent modelling reaches a harsher verdict on bonds. Anarkulova, Cederburg and O'Doherty, in a March 2025 working paper, run a lifecycle model on 39 developed countries from 1890 to 2023, over 2,600 years of country-month returns, using a block bootstrap that preserves long return sequences. Their optimal fixed-weight portfolio is 33% domestic stocks and 67% international stocks, with nothing in bonds or bills. Relative to that, a couple has to save 20% more to hold 20% bonds, 35% more to hold 30% and 54% more to hold 40%. A 60/40 saver needs 19.3% of income to match what the all-equity saver achieves on 10.0%.
That paper is contested, and for good reason. It's a working paper, not yet a published journal article. Its block bootstrap treats a decade drawn from French history as exactly as likely as one drawn from US history, which is a modelling choice with a large effect on the answer. And its own results show all-domestic-equity carrying a 17.1% chance of financial ruin in retirement against 16.9% for the balanced strategy — so even in the authors' framework, the equity case rests on international diversification rather than on equity alone.
The counter-argument: recovery time is the wrong metric
The strongest objection to everything above is that recovery time only matters if you have a fixed pot and add nothing to it. Critics of drawdown-first thinking argue that anyone still contributing is buying through the fall, and a long drawdown is precisely when the contributions are cheapest. On that view a portfolio that stays down for a decade isn't a punishment. It's a discount.
That argument is right, and it's slower than people think. Here's the same series run as a saver instead of a lump sum. One inflation-adjusted unit is contributed at the start of every year from 2000, straight into the worst equity sequence in the record.
| Balance at | 100% equity | 80/20 | 60/40 |
|---|---|---|---|
| End-2002 (3 paid in) | 2.00 | 2.28 | 2.58 |
| End-2008 (9 paid in) | 6.50 | 7.67 | 8.81 |
| End-2012 (13 paid in) | 15.28 | 16.19 | 16.83 |
| End-2013 (14 paid in) | 21.20 | 20.98 | 20.31 |
| End-2025 (26 paid in) | 94.25 | 74.84 | 58.15 |
The saver in 60/40 was ahead for thirteen straight years. The crossover came in 2013, and after that the all-equity balance never looked back — it ended 62% larger. So the objection wins on a 26-year view and loses on a 13-year one. Both facts are true and they answer different questions.
Length of horizon is doing all the work. After the 1973-74 crash the same exercise flips much faster: the 60/40 saver led at the end of 1974 by 1.41 units to 1.18, and the all-equity saver was permanently ahead from 1978, six years in. Same mechanism, less than half the wait, because equities recovered sooner that time. A saver can't know in advance which version they've drawn.
There's supporting evidence for the contributions argument from a different angle. Vanguard's 2023 study of cost averaging found that lump-sum investing beat a three-month cost-averaging split 66.4% of the time in the US over 1979-2022, and that the margin widened with equity weight: 2.2% at the median for a 100% equity portfolio against 1.2% for a 40/60. Time in the market is worth more when the market is riskier. The mirror image is that a saver dripping money in through a drawdown is capturing the risk premium the whole way down.
Where the objection genuinely fails is at the other end of life. Once contributions stop and withdrawals start, a long drawdown stops being a discount and becomes an amplifier. That's the case covered by how sequence risk turns identical returns into opposite outcomes. Javier Estrada's 2015 study of 19 countries over 110 years, running 81 rolling 30-year retirements at a 4% withdrawal rate, found a static 60/40 failed 4.9% of the time in US data, tying the best result on record. On the world market it failed 16.0% of the time against 14.8% for all-equity. His sample stops in 2009, so it contains neither 2022 nor anything after it.
What would change the conclusion
Monthly data would change the depth figures and could change some of the durations. Annual measurement understates every trough here, and it understates deep fast crashes more than slow shallow ones, which flatters the all-equity column. If the ordering of the recovery table survives monthly measurement, the finding is much stronger than what's shown.
A different bond proxy would matter too. A real fund holds credit, pays fees and rolls at market prices rather than at a modelled par. Each of those makes the 40% leg worse than the synthetic version used here, though not by enough to reverse the 1973-74 result.
A non-US series would matter most. Run this on Japan from 1989 or on any market that closed for a war, and the terminal-wealth column probably inverts. The Yearbook's own figures put twenty-first-century global equity real returns at 3.5% a year, well below the 8.5% annual average the US delivered since 1900. If forward equity returns look like the former rather than the latter, the compounding advantage that justifies the deeper drawdowns shrinks a lot.
And a persistent positive stock-bond correlation would change what 60/40 even is. If 2022 was a preview rather than an anomaly, the mix stops being a hedge and becomes a return-dilution decision with the same crash risk. Vanguard's own life-cycle researchers concede that the debate over glide-path shape remains unsettled, which is the honest state of it.
What wouldn't change: the depth ranking. More equity means deeper falls, in every episode and every measurement basis in this record. It's the leap from that to a claim about how long you'll wait, or what you'll end up with, that the evidence refuses to support.
Key takeaways
- From the end-1928 peak, a 100% equity portfolio fell 64.8% in nominal terms, 80/20 fell 53.5% and 60/40 fell 40.1%, measured on annual data.
- After the end-1972 peak, real recovery took 12 years at 100% equity and 13 years at both 80/20 and 60/40, so extra bonds bought no time back.
- In 2022 equities lost 18.0% and ten-year Treasuries 17.8%, so all three portfolios fell about 18% nominal and about 23% after 6.4% inflation.
- $100 at the end of 1927 became $61,445 in real terms at 100% equity, $31,542 at 80/20 and $13,652 at 60/40 by the end of 2025.
- A saver contributing yearly from 2000 stayed behind at 100% equity until 2013, then finished end-2025 with a balance 62% above the 60/40 saver.