Key takeaways
- Over 1901 to 2020, every stock and bond mix in Vanguard's global data had a losing year. The shallowest worst year was 20% equities at -21.2%.
- Moving from all equities to that 20/80 mix made the worst year 19.8 points shallower and cut the average annual return from 7.3% to 4.8%.
- Stretch the holding period to five years and the shallowest worst case moves to 60/40, at -5.3% a year. At twenty years it moves to 100% equities.
- In the 12 months to 30 September 2022 the ladder inverted. Vanguard's 20% equity fund fell 15.40% and its 100% equity fund fell 4.47%.
- With UK CPI inflation at 10.1% that September, the cautious fund's real loss was about 23.2%, against about 13.2% for the all-equity one.
Every mix lost money somewhere, so a crisis proof portfolio is a choice about which crisis
You want to know whether some allocation holds up in every crash, and what owning it costs. Here's the short answer. There isn't one, and it isn't because nobody has looked. The mix with the shallowest worst case changes depending on which crisis you ask about and how long you hold.
Vanguard's 120 years of returns book runs six global stock and bond mixes through 1901 to 2020. Every one of them had a losing year. The shallowest worst year belonged to the mix holding 20% equities and 80% bonds, at -21.2%. All equities came in at -41.0%. All bonds, the mix most people would nominate as the cautious one, did slightly worse than 20/80 at -21.7%.
Then the holding period changes everything. Over five-year stretches the shallowest worst case wasn't 20/80 at all. It was 60/40, at -5.3% a year. Over twenty-year stretches it was 100% equities, at 2.9% a year. And in the 12 months to 30 September 2022, in real UK funds, the ranking turned upside down.
The 120-year ladder, and the price of each step down it
Vanguard's return spectrum covers 1901 to 2020. The figures are net returns in euros, and in Deutsche Marks before 2002, with no fees or taxes deducted. Bonds are the DMS World Bond Total Return Index to 1999 and the Bloomberg Barclays Global Aggregate after that. Equities are the DMS World Equity Total Return Index to 1969 and the MSCI World Index after that.
| Equities / bonds | Worst 1 year | Worst 5 years, a year | Worst 20 years, a year | Average 1 year |
|---|---|---|---|---|
| 0 / 100 | -21.7% | -7.3% | -0.7% | 4.0% |
| 20 / 80 | -21.2% | -6.6% | 0.0% | 4.8% |
| 40 / 60 | -26.2% | -5.9% | 0.7% | 5.6% |
| 60 / 40 | -31.1% | -5.3% | 1.5% | 6.3% |
| 80 / 20 | -36.1% | -6.4% | 2.2% | 6.8% |
| 100 / 0 | -41.0% | -9.1% | 2.9% | 7.3% |
The first step down the ladder is the one worth pausing on. Going from all bonds to 20% equities made the worst year slightly shallower, from -21.7% to -21.2%, and raised the average annual return from 4.0% to 4.8%. That's the diversification result in its purest form. A small equity holding paid for itself twice, and it's the only step on this ladder that was free.
Every step after it cost money. Moving from 100% equities down to the 20/80 mix made the worst year 19.8 percentage points shallower and cut the average annual return by 2.5 points. Across 120 years that gap compounds into something most people underestimate. Vanguard's figures show 100 euros invested in 1901 growing to 437,413 euros at 100% equities, against 28,778 euros at 20/80. That's roughly 15.2 times more money, given up in exchange for a worst year half as deep.
The chart above plots the worst single year for each of the six mixes. The kink at 20/80 is the free step. Everything to the right of it is the price list. If you want the same arithmetic seen from the recovery side, loss recovery asymmetry explains why a deeper hole needs a disproportionate gain to climb out of.
The shallowest worst case climbs the risk scale as the holding period lengthens
A worst calendar year is the wrong measure for most people, because most people aren't selling in twelve months. Vanguard's book runs the same six mixes over three, five, ten and twenty year holding periods, and the ranking keeps moving up the risk scale.
Over five-year stretches the shallowest worst case was 60/40 at -5.3% a year, a better floor than all bonds managed at -7.3%. The 60/40 mix also averaged 6.7% a year over those windows against 4.2% for all bonds. At that horizon the cautious mix lost on both counts at once.
Over twenty-year stretches the all-bond mix was the only one of the six whose worst window was negative, at -0.7% a year. Every mix holding at least 20% equities had a worst twenty-year window of 0.0% a year or better, and all equities had the firmest floor of the lot at 2.9% a year. The asset that dug the deepest one-year hole produced the steadiest twenty-year floor.
None of that makes shares safe. It makes "worst case" an incomplete phrase. Change the holding period and you change which risk you are measuring, and the asset allocation that minimises it moves with you.
In 2022 the ladder inverted and the most cautious fund fell hardest
The 120-year record stops in December 2020, so it misses the episode that broke the pattern. Vanguard's five UK LifeStrategy funds hold the same underlying Vanguard index funds at fixed equity weights of 20%, 40%, 60%, 80% and 100%, and all five launched on 23 June 2011. Over the 12 months to 30 September 2022 they returned this, in sterling, after their 0.20% ongoing charge:
| Fund | 12 months to 30 Sep 2022 | 12 months to 30 Sep 2020 |
|---|---|---|
| LifeStrategy 20% Equity | -15.40% | 3.13% |
| LifeStrategy 40% Equity | -12.85% | 2.51% |
| LifeStrategy 60% Equity | -10.28% | 1.78% |
| LifeStrategy 80% Equity | -7.46% | 0.87% |
| LifeStrategy 100% Equity | -4.47% | -0.17% |
That 2022 column is a clean inversion across all five rungs. The most cautious fund lost 10.93 percentage points more than the most aggressive one. Nothing in the 120-year table prepares you for it.
The mechanism was interest rates rather than company earnings. The Bank of England raised Bank Rate from 0.25% on 16 December 2021 to 2.25% on 22 September 2022. Bond prices fall when yields rise, and the long end of the curve moved further than Bank Rate did. The Bank's December 2022 Financial Stability Report records that between 1 August and their peak on 14 October, 30-year gilt yields briefly exceeded 5%, which it calls "a more than 270 basis point increase" and describes as more than twice the size of the move during the March 2020 dash for cash.
This wasn't one fund family behaving oddly. The Investment Association sector averages printed alongside those fund returns are industry-wide benchmarks rather than Vanguard's own, and they show the same inversion. Over the same 12 months the IA Mixed Investment 0-35% Shares sector returned -11.34%, IA Mixed Investment 20-60% Shares returned -10.71%, IA Mixed Investment 40-85% Shares returned -10.27% and IA Global returned -9.16%. More shares, smaller loss, all the way up the scale.
Two years earlier the same ladder behaved exactly as the textbook says it should. Over the 12 months to 30 September 2020, a window containing the February and March 2020 crash, the 20% equity fund returned 3.13% and the 100% equity fund returned -0.17%. Same funds, same weights, opposite ordering, one crisis apart. That is the pattern diversification in a crisis is built on, and 2022 is the year it reversed.
Inflation made the cautious loss far worse than the headline number
Nominal returns understate 2022, because the force pushing yields up was also eating the money. UK CPI inflation ran at 10.1% in the 12 months to September 2022, on the Office for National Statistics series D7G7.
Deflate the fund returns by that and the 20% equity fund's -15.40% becomes a real loss of about 23.2%. The 100% equity fund's -4.47% becomes about 13.2%. A portfolio almost anyone would describe as defensive lost close to a quarter of its purchasing power in a year, and it did so without an equity bear market being the cause. The all-equity fund, in real terms, lost a little over half as much.
The strongest objection is that 2022 was a bond repricing, not a crash
There's a serious case against reading too much into one inverted year, and it deserves its best shot.
2022 was not an equity crisis. Global shares fell, but the fall came from discount rates rather than from collapsing earnings or a banking failure. A bond-heavy portfolio isn't built for that. It's built for the case where shares halve, and across 1901 to 2020 it did that job: the worst year at 20/80 was -21.2% against -41.0% for all equities, and the worst three-year stretch was -11.2% a year against -25.7%.
The 12 months to 30 September 2022 is also a single window, in funds with no Global Financial Crisis in their record, no dot-com bust and no 1974. One observation that cuts against a 120-year pattern is a warning, not a refutation. Both datasets can be true at once. Bonds have usually cushioned equity crashes, and in 2022 they did the opposite, because the shock arrived from the rate side rather than the growth side. The same tension shows up in fixed-weight designs like the permanent portfolio, which spreads the bet across regimes rather than picking one.
What the inversion does rule out is the idea that a crisis proof portfolio can be written down in advance as a fixed set of weights. The weights that would have minimised your loss depend on which shock turned up, and nobody gets to choose that.
What this data cannot tell you
Start with the currency. Vanguard's 120-year series is measured in euros, and in Deutsche Marks before 2002. A UK investor's experience of the same assets differs by whatever sterling did, and sterling did a great deal over that span. The -21.7% worst year for global bonds is a Deutsche Mark number, not a pound one.
The series is an index rather than a fund. Vanguard states that fees and taxes are not taken into account and that an index is not an accurate representation of a particular investment. The holding-period figures also come from overlapping windows, so the worst five-year stretch and the worst three-year stretch sitting inside it are not independent observations.
Most of the sources here are Vanguard's own, and Vanguard sells the funds this ladder is built from. The 120-year table is theirs and so are the fund returns, and the firm has an interest in the conclusion that broad index exposure is enough. The Investment Association sector averages and the Bank of England report are here partly for that reason.
And a backtest is not a forecast. The period from 1901 to 2020 is one path. It contains two world wars and the Great Depression, and it ends before the rate rises that made 2022 what it was. Nothing in it says the next crisis has to resemble any of them.
What would change the conclusion
If shares and bonds return to moving in opposite directions, the 120-year ladder reasserts itself and 2022 reads as an anomaly. The sign of that relationship is the single variable the whole argument rests on, and it has flipped before.
If inflation settles near target, the real-terms damage shrinks and the nominal numbers become the right ones to read again. UK CPI inflation was 3.5% in September 2026, roughly a third of the September 2022 rate.
If your horizon is one year rather than twenty, the table flips back. The shallowest worst case climbs the risk scale with time, so somebody spending the money next year and somebody spending it in thirty years are not reading the same row.
The figure worth watching isn't any one mix's worst year. It's whether your own stock and bond mix still sits where you put it, because portfolio drift quietly moves you to a different row of that table. LedgerTouch tracks that drift continuously and a spreadsheet does it once a year. The table doesn't care which you use.