What actually protects you when stocks fall
Government bonds and trend following are the only two things here with a record worth having, and each works against one kind of crisis rather than against all of them.
The number every fund leads with, how closely it moves with the stock market on average, will not tell you which is which.
Low number, good diversifier, or so the fact sheet implies. That number is close to useless, and one example shows why.
A broad commodity fund moves with US stocks at 0.44 over its whole history, on a scale where zero means the two move independently and one means they move in lockstep. Respectably low. Now split the record in two. In the months when stocks fell, the fund fell 139% as far as they did. In the months when stocks rose, it rose 20% as far. Seven to one, in the wrong direction. The average conceals it completely, because averaging the two halves together is exactly what destroys the information.
The academic version is older and starker. Comparing US and international stocks from 1979: in months when both were up by more than one standard deviation, the relationship between them was minus 17%. In months when both were down by more than two standard deviations, it was plus 93%. An ordinary statistical model fed the same data says the second number should have been 14%. Diversification thins out at precisely the moment you were holding it for. Chua, Kritzman and Page published that result; a 2018 paper in the Financial Analysts Journal by Page and Panariello found the same shape in nearly every risky asset they examined, including hedge funds that describe themselves as market neutral.
A crisis is not one thing
This is the part almost nobody says out loud. Protections are specific to the kind of trouble, and the three big episodes of this century were three different kinds.
2008 was a credit and growth crisis. Treasury bonds worked, up 17.9%. Trend following worked, up 20.9%. Gold fell 29.5% from its peak, including 21% in eleven days that October.
2020 was a liquidity crisis. Treasuries worked. Funds that own volatility worked spectacularly. Trend following barely moved, up 6.3% for the year, because the fall was far too fast for a rule that looks back twelve months. Gold fell 12.4% in nine days.
2022 was an inflation crisis. Trend following worked, up 27.4%. Commodities worked, up 19.3%. Treasuries lost between 15% and 31%. And crash insurance lost 13.1% in the very year a balanced portfolio fell 16%, because fear never spiked: the main volatility index did not reach 36 at any point all year.
Anything sold to you as protection in all weather is being oversold.
The record, plainly
Take the worst months for US stocks since 2000 and count the share of them in which each thing went up. Read the denominator carefully, because it is not the same down the column. Each row is counted over the bad months that fall inside its own history, which runs to 40 months for the holdings reaching back to 2000 and to fewer for the ones that started later, bitcoin most of all. A 3% share is one month out of that row’s own count rather than one month out of forty.
| Positive in bad months | |
|---|---|
| Trend following | 64% |
| Gold | 59% |
| Medium-term Treasuries | 58% |
| Long Treasuries | 55% |
| Commodities | 28% |
| Bitcoin | 19% |
| International stocks | 6% |
| Property funds | 3% |
| High-yield bonds | 3% |
On the 20 worst single days, property funds, commodities and high-yield bonds were positive zero times out of twenty.
Read the top of that table carefully, because it is not the good news it looks like. Government bonds are the best single protection an ordinary investor can buy, and they are still close to a coin flip on any given bad month.
What each one is actually good for
Government bonds work when the trouble is growth. They do not work when the trouble is inflation, and 2022 is the whole lesson in one year.
Worse, the relationship has flipped. Stocks and bonds moved against each other, at around minus 0.4, through the 2010s. Since 2022 they have moved together, and the relationship right now is the most positive in the sample. A balanced portfolio is not protected the way it was for the past generation, and anyone whose plan rests on the 2010s should know that.
Gold across a whole crisis often does work. It gained 25.5% over the 2007 to 2009 bear market. Inside the worst individual weeks it falls, because it is easy to sell and frightened people sell whatever they can. Through the 2020s its average return in bad months for stocks has been negative.
Trend following is the only thing here that becomes more negatively related to stocks when stocks fall, rather than less. That is the property you actually want, and it is rare. It is also the thing that lost 18.6% over the twelve months to April 2025, the worst run in its record.
Bitcoin moved with stocks at about zero before 2020 and about 0.4 since. In the March 2020 crash it was 0.6. It behaves like a stock position with roughly three times the jumpiness, rising and falling further than shares do in both directions.
Property funds are not diversification. They are stocks with more borrowing attached. They fell 41.5% in the 2020 crash, worse than the stock market did.
One thing about the sources
Nearly every good crisis story is published by a firm that sells the product the story makes look good. Read all of it that way, this site included. It is worth noting that some of the most useful warnings came from those same firms. The 18.6% loss above was published by a manager who sells trend-following funds, and the gold industry’s own trade body conceded that the 2026 rise in the gold price was mostly people chasing the move rather than people seeking safety.
What would change our mind
A protection that holds up across a growth crisis, a liquidity crisis and an inflation crisis, measured in all three rather than chosen after the fact. Or evidence that stocks and bonds have gone back to moving against each other for reasons other than the last two years of data.
Where these numbers come from
- Daily and monthly index and fund prices for each asset class, from 2000 or from that series’ own start, whichever is later, for the count of bad months and bad days.
- Chua, Kritzman and Page, "The Myth of Diversification", Journal of Portfolio Management, 2009.
- Page and Panariello, "When Diversification Fails", Financial Analysts Journal, 2018.
- Published calendar-year returns for Treasury, commodity, trend-following and protection funds, read August 2026.
- The World Gold Council’s own commentary on the 2026 gold price move.