What actually protects you in a crash?

Foreign stocks, property and high-yield bonds are stocks by another name when it matters. Commodities and bitcoin help less than a coin flip. The only things that rose more often than not were trend following, gold and Treasuries.

A second way to score it: take the worst tenth of all stock months since 1926 and ask what moving 10% of a stock portfolio into each asset would have added in the average such month. Treasury bills: 0.92%. Long Treasuries: 0.94%, in exchange for swings of 8% a year and a worst fall of 59% of their own. Gold: 0.92%. Trend following: 1.08%, and it was positive in 69% of those months. Cash does the job usually credited to bonds, for free, without the risk that bonds fall alongside stocks.

Insurance that did not pay

TAIL is the best-known crash fund. It holds about 91% Treasuries and 5% put options, contracts that pay out when the market falls, and charges 0.59%. Between 2017 and the end of 2024 it lost 8.4% a year while US stocks made 14.5%. It made 7% in 2020. In 2022, the bear market it was bought for, it lost 13% while the S&P 500 lost 18%, because the Treasuries it holds fell too. Holding 10% of a portfolio in it would have cost about 2.2 points a year of growth. CAOS, a gentler version, made 3% a year over ten years against 15% for the S&P 500. CYA, a fund that promised to profit from crashes, lost 99% of its value between September 2021 and March 2024 and was closed.

BTAL bets against high-flying stocks and is often sold as protection. It is the only fund I measured whose shape could be read at all, and the shape is wrong: it lost 0.26% for every 1% the market rose and also lost 0.12% for every 1% it fell. It has lost 3.6% a year since launch against 15.4% for the S&P 500, at a 1.4% fee.

Buffer funds promise to absorb the first 10 or 15% of a fall in exchange for a cap on gains. Of 102 that I could compare with a plain mix of stocks and cash carrying the same risk, 86 returned less, 70 fell further at their worst, and none beat the mix on both counts. A fund that loses money during a bear market is a bet on the shape of the fall, not protection against falling. Holding fewer stocks does the same job for no fee.

Dividend and property funds get sold as the steadier kind of stock fund, and they are still stock funds. SCHD, the best-known dividend fund, tracks the whole US market month to month with a link of 0.82 out of 1, and its payouts cost a taxable holder about half a point a year in tax, so in a taxable account it is the market with a screen and a tax bill. VNQ, the largest property fund, fell 41.5% in 2020, rose in 3% of the worst stock months since 2000, and over its history caught 112% of the market's falls for 80% of its rises.

Bonds and TIPS

Too close to call A choice about how much you can lose, not a way to earn more.

Bonds are different. Nobody should buy them for a return edge, and they are not insurance either. They are how you choose to lose less.

Inside the portfolio on this site, over 96 years, moving 10% from stocks to cash cost 0.77 points a year of growth, and moving it to Treasury bonds cost 0.55. Those were the only changes I tested that cut the depth of a 1929-scale or 2008-scale fall, by 4 to 5 points. That is the trade: give up about half a point a year, lose a bit less at the bottom. Whether it is worth it depends on the fall you could sit through, which is why the portfolio pages ask that question and not this one.

Adding bonds on top of stocks through a fund that borrows, the way RSSB does, is a different bet, and it is a bet on bonds rising when stocks fall. On 96 years it came out 0.34 points a year ahead of the same portfolio without the bonds, with a range that includes zero. All of that gain came between 1981 and 2020. For 576 months in a row, 1933 to 1981, the version with bonds trailed the version without, and across 1977 to 1981 it cost 22 points. Since 2020 stocks and bonds have mostly fallen together: their three-year link stood at plus 0.51 at the end of 2025, against about minus 0.4 through the 2010s. I do not hold it.

Which bond

TIPS are Treasury bonds whose payments rise with inflation. Today a ten-year TIPS pays 2.44% a year above inflation, the highest since October 2008, and a thirty-year pays 2.99%. That is the best contractual real return on offer anywhere, and it is the reason the version of portfolio three priced for today holds some.

TIPS and ordinary Treasury bonds are one idea, not two. Across eighteen bond funds' own monthly returns since 2019 the two kinds moved together at +0.76 to +0.85 out of 1, so holding both does not spread your bets. They differ in one way that matters. TIPS moved slightly with stocks and ordinary Treasuries slightly against them, so TIPS are the marginally worse crash diversifier and the better match for spending that rises with prices. Hold TIPS because your future spending is real, and take any crash protection from cash.

SCHP and TIP hold the same bonds; SCHP charges 0.03% and TIP 0.18%, so the fee is the whole decision. Short TIPS funds such as VTIP and STIP follow inflation with little swing from interest rates. A long one such as LTPZ locks in today's thirty-year real yield and swings far more. The cautious version on this site uses a ten-year fund.

Maturity matters, in the opposite direction from the usual advice. Long Treasury bonds are sold as the crash hedge, and over a century they added 0.94% back in the average worst stock month for a 10% swap out of stocks, against 0.92% for Treasury bills. Two hundredths of a point a month is the whole case for long bonds, and it is bought with swings of 8.4% a year and a worst fall of −59.1% of their own, including a 40% loss through the late 1970s that cash never had. Cash and short Treasuries do the crash job. Long bonds are a bet that interest rates fall.

Whichever you hold, hold it in a traditional IRA or 401(k) first. Bond interest is taxed as ordinary income, so bonds belong in the sheltered account ahead of stocks by a factor of about four at every tax rate. The fees and taxes page covers municipal bonds, which only make sense once your bonds overflow the sheltered accounts.

Corporate and high-yield bonds

Corporate bonds are Treasuries plus a slice of stock risk. An investment-grade corporate bond fund such as LQD or VCIT tracks Treasuries at +0.83 out of 1, so it is the same bet with a small extra payment for lending to companies. High-yield bonds, the loans to shakier companies in HYG and JNK, rose in 3% of the worst stock months since 2000. In a fall they are stocks by another name.

There is one version of the idea worth knowing about. Strip the interest-rate risk out of corporate bonds and what remains, the extra payment for lending to companies, moved with long Treasuries at only +0.02 out of 1 between 1926 and 2014. It earned the same return as long Treasuries at half the swings and a third of the worst fall, and it lost 13% in 2008 while Treasuries gained 13%. That last number is the point: it is a way to earn a little more inside a bond allocation, never protection. The funds that do this, LQDH and IGBH, or the AAA-rated loan funds such as JAAA at 0.20%, are cheap, and none has lived through a March 2020. With investment-grade bonds paying only 0.82% over Treasuries in August 2026, near the bottom of their range, I hold none.

Foreign and emerging-market bonds

I have not measured foreign bond funds, so this is reasoning rather than a result. A hedged fund such as BNDX or IAGG removes the currency and leaves foreign government and corporate bonds, which move with US bonds for the same reasons US bonds move. Until it is tested, treat it as BND with a currency hedge attached and a slightly higher fee. Unhedged foreign bond funds carry the currency itself, which adds swings and no expected return, the same point the international page makes about foreign stocks. Emerging-market bond funds such as EMB and VWOB pay more because they carry the currencies that fall hardest in a crisis. That is the insurance this page says not to sell, in bond form.

Catastrophe bonds

Too close to call The one asset whose losses come from hurricanes. Priced too thin to bother with today.

A catastrophe bond is a loan to an insurer that is written off if a named disaster, usually a hurricane or an earthquake, is bad enough. Its return has nothing to do with stock markets, which makes it the only true black-swan asset I looked at. The funds that hold them returned 3.31% a year between 2017 and 2025 after fees, about one point over Treasury bills, across nine years that included the three best the market has had. They can lose 10% in a week, as they did when Hurricane Ian landed in September 2022.

The price is the problem. These bonds pay a spread above their expected loss, and in mid-2026 that spread was 2.2 times the expected loss, against 4.9 times at the start of 2023. At today's multiple, after the 1.6% to 2.4% the retail funds charge, about one point a year is left for a risk that can remove a tenth of the position at once. The fund that made this reachable, ILS, launched in April 2025 at a 1.58% fee, though its own report shows 2.00% actually paid. If the multiple returns to about 3.5 times, a number published weekly, the arithmetic changes and a small position, held only in a retirement account, becomes reasonable.

Cash

When this page says cash it means a Treasury-bill fund such as SGOV or BIL, which paid 3.88% in August 2026 with no swings. In a taxable account BOXX does the same job differently. It holds option contracts that pay the Treasury rate, has returned 4.70% a year since December 2022 at a 0.19% fee, and pays it as a rise in price rather than as interest, so you owe tax when you sell rather than every year. Of everything on this page, a 10% move from stocks into cash is the one that added protection in the worst months, cost no fee, and never had a bad year.