Gold has fallen 28% from its January 2026 record. What did it protect?

On 12 November 2026 gold closed at $3,870.20 an ounce, 28.4% below its record of $5,405.00 on 29 January and its lowest close of this fall; in euros it closed at €3,495.56, 23.0% below its 2 March record. It lost 3.6% on 12 November alone, and its dollar price was the lowest since 30 September 2025. World equities rose over the same months: MSCI World with net dividends gained 16.7% in euros from gold's record to 12 November.

For a euro investor who holds gold next to an equity fund, the useful question is what the sleeve did for the whole portfolio. Did it rise when equities fell, including in the October-November AI sell-off, and did it raise or lower compounded growth after costs? We measured both with daily prices since 2001, and the answer changes with the decade you look at.

Key takeaways

1. Gold's fall is large, but not unusual for gold. Since 1970 its dollar price has fallen 20% or more eight times; this 28.4% fall ranks fifth, and the 1980 and 2011 records took 28.0 and 8.9 years to regain.
2. In the AI sell-off gold held roughly level instead of rising. From the Nasdaq's 6 October record to its 5 November low, MSCI World lost 4.2% in euros and gold 0.7%; by 12 November gold was 5.4% lower than on 6 October, while equities had almost recovered.
3. Its diversification is real; its crash protection is not reliable. The weekly correlation with world equities in euros was 0.01 over 1,349 weeks since 2001, and gold rose in five of eight selected equity falls of 10% or more, but it fell in 2001-03 and March 2020.
4. A 10% sleeve cut the worst fall in all 191 ten-year windows, but raised growth after costs in only 109. Every window starting between July 2008 and March 2015 ended with less growth; the sleeve helped whenever gold compounded within 0.9 points a year of equities and hurt whenever it trailed by more than 2.4 points.
5. A household can set the weight, the cost and the rebalancing rule, not gold's return. Gold pays no income, so its contribution to growth rests on price alone, and a January reset buys gold only if its portfolio weight is below the target at that date.

How far has gold fallen, and how does this fall compare with earlier ones?

At 28.4% below its record, gold's dollar price is in its fifth-deepest fall since 1970, and only the falls from 1975, 1980 and 2008 had gone further 287 days after their record. The fall followed a steep rise: from 12 November 2024 to the January record the dollar price rose 107.3%. Two years on it is still 48.5% higher in dollars and 42.3% higher in euros than on 12 November 2024; over the last twelve months it lost 6.4% and 2.2%.

In dollars: −70.3% (1980-99, 28.0 years to a new record), −44.6% (2011-15, 8.9 years), −44.1% (1975-76, 3.0 years), −28.8% (2008, 1.5 years) and −28.4% now; in euros −36.8% (2012-13, 6.9 years) and a −23.2% trough on 16 July 2026, versus −23.0% on 12 November. Source: World Gold Council gold prices; drawdown table (CSV).

The 1980 and 2011 falls took much longer to recover than the current sell-off has lasted. After a 70.3% fall, the January 1980 record of $850 was regained only in January 2008. The 2008 fall is the closest match to this one: 28.8% in seven months, recovered within 1.5 years of the record. From the 12 November close, a return to the January record needs a 39.7% gain, the arithmetic set out in our guide to recovery time and loss maths.

These are nominal benchmark prices, before fund fees or storage costs. Regaining a quoted high does not mean regaining the same purchasing power or covering the cost of holding gold. Our earlier study of gold's crisis drawdowns examines those differences and the time spent below the old peak.

Why did a euro investor lose less than a dollar investor?

Because the dollar rose against the euro: the ECB reference rate went from $1.1968 per euro on 29 January to $1.1081 on 12 November, so each dollar was worth 8.0% more in euros. Gold is priced in dollars, so a stronger dollar lifts its euro price. Measured from the dollar record day, gold lost 22.7% in euros against 28.4% in dollars.

The euro record came later, at €4,537.54 on 2 March 2026, two days after the 28 February strikes on Iran that our oil shock study uses as its starting point. The lowest euro close of the fall so far was €3,486.77 on 16 July, 23.2% below the record; the 12 November close was only €8.79 above it.

The currency effect also works the other way. From May 2001 to March 2003 gold gained 21.0% in dollars, but a stronger euro turned that into a 5.6% loss in euros. For an unhedged product, quoting a trading price in euros does not remove the currency effect.

Did gold offset the October-November AI sell-off?

Only on the sharpest days: from the Nasdaq Composite's 6 October record to its 5 November low, MSCI World lost 4.21% in euros and gold 0.68%, and gold then fell 4.71% while equities rebounded. The week to 6 November, the core of the AI sell-off, is where gold did help. On 5 November, when MSCI World lost 3.21% in euros, gold rose 2.17%.

Source: MSCI World index 990100, World Gold Council gold prices; daily levels (CSV). In dollars gold fell 1.05% from 6 October to 5 November and 6.89% to 12 November. The 90/10 column weights the two euro returns without costs or rebalancing, as an illustration.

Over the whole episode the gold sleeve softened the fall by 0.35 points and then cost more than that in the rebound. By 12 November a 90/10 mix was 1.02% below its 6 October level, against 0.54% for equities alone. The gold price is a London benchmark set before US markets close, while MSCI's level reflects the later close, so single-day comparisons can mismatch.

Gold also moved more with equities this year than usual. The weekly correlation between the two in euros was 0.36 in 2026 to 6 November, the second-highest calendar-year figure since 2001 after 0.50 in 2010. Yet their returns since gold's January record differ by 39.4 points, a reminder that correlation measures co-movement week by week, not the direction of the trend.

European investors kept buying. In its October 2026 ETF flows report, the World Gold Council wrote that "Europe has now seen inflows for four months in a row, adding US$1.8bn in October", and that global holdings reached 4,275 tonnes, "another month-end record."

Has gold fallen with equities in earlier crashes?

In euros, gold rose in five of the eight selected MSCI World sell-offs of 10% or more since 2001 and fell in three: 2001-03, 2015-16 and February-March 2020. Its best case was the 2007-09 bear market, when gold gained 49.3% in euros while world equities lost 53.4%.

MSCI World versus gold in euros: 2001-03 −50.9% and −5.6%; 2007-09 −53.4% and +49.3%; 2020 −33.8% and −5.1%; 2022 −16.9% and +8.8%; gold's 2026 fall +16.7% and −22.7%. Source: MSCI World index 990100, World Gold Council gold prices; episode table (CSV).

March 2020 is the warning case: the fastest equity crash in the sample, with gold 5.1% lower at the end and 9.9% lower at its worst point inside the window. The World Gold Council's October report suggests one mechanism, writing of October's selling in US technology shares: "The resulting market stress may have contributed to the region’s gold ETF outflows as some investors turned to their gold holdings as a ready source of liquidity." An asset that is easy to sell is also sold in a scramble for cash.

Gold's weekly returns had little linear co-movement with equities over the full sample. Over 1,349 weeks from January 2001 to 6 November 2026 the correlation of weekly returns in euros was 0.01. In the 134 worst weeks for MSCI World, losses of 2.69% or more, gold fell in 41.0% of them with a median gain of 0.40%; in the other 1,215 weeks it fell in 45.8%, with a median gain of 0.17%.

That is diversification in copio's sense: a return source with little average co-movement, despite losses in some of the same weeks. Our diversification study found a monthly correlation of −0.06 between gold in euros and global equities over 2006-2026, and the crash-haven comparison found gold rising in five of seven US crashes since 1929. In this sample gold still fell in roughly two of every five of equities' worst weeks. That observed frequency is a separate result; near-zero correlation alone does not imply it.

What did a gold sleeve do to a euro portfolio's compounded growth?

Over all 25.9 years of data, a 10% gold sleeve reset each January raised annual growth after costs from 6.51% to 7.09% and cut the worst fall from 54.3% to 47.1%; started at gold's October 2012 euro peak, it lowered growth from 12.59% to 12.20%. Both scenarios combine MSCI World with net dividends in euros and gold in euros, with fund costs deducted, before spreads and tax.

Annualised after costs, world only versus 10% gold: 6.51% and 7.09% from December 2000; 12.59% and 12.20% from October 2012; −1.30% and +0.65% in the best ten years for the sleeve; 14.36% and 13.23% in the worst. Source: MSCI World index 990100, World Gold Council gold prices; scenario results (CSV), ten-year windows (CSV).

The full-sample gain came from gold itself. From December 2000 gold compounded 10.0% a year in euros against 6.7% for world equities, starting 17 months after the low of its 1980-1999 fall. From its October 2012 peak it compounded 6.8% against 12.8%, and the sleeve became a drag. At 20% the same pattern is stronger: 7.61% a year over the full sample, 11.76% from October 2012.

The portfolio comparisons below keep the January rebalancing rule and the same cost assumptions throughout. They show both sides of the allocation: shallower losses during several equity falls and a lower gain in 2026, when gold's decline offset part of the equity recovery. The weights are scenarios, not proposed allocations.

Source: moves in equity falls (CSV), portfolios reset each January, after fund costs. In 2020 the 10% sleeve took 2.9 points off a 33.8% fall, less than the 3.4 points that 10% held in cash at a zero return would have taken off. In 2026 it has cost 2.4 points so far; from gold's 29 January record, €10,000 grew to €11,650 in world equities and €11,257 with a 10% sleeve.

When did the sleeve add growth, and when did it only reduce swings?

In all 191 ten-year windows starting each month from January 2001, the 10% sleeve reduced the worst fall, by 2.1 to 7.5 points, but it raised compounded growth in only 109. Every window starting between July 2008 and March 2015 ended with less growth than world equities alone, by up to 1.13 points a year for the decade from October 2011.

What decided the outcome was gold's own return. The sleeve raised growth in every window where gold compounded less than 0.93 points a year below equities, and lowered it in every window where gold trailed by more than 2.37 points. Between those gaps the path decided it, because resetting each January sells the asset that has risen and buys the one that has fallen.

That rebalancing effect was small. Over the full sample, the 10% sleeve reset each January compounded 7.09% a year, against 6.99% when it was never rebalanced and grew to 19.9% of the portfolio. In this sample the two rebalancing rules differed by about 0.1 percentage points a year; it did not rescue the decades in which gold lagged by several points.

Gold's contribution to growth changed across decades despite its low full-sample correlation. From 2001 to 2011 the sleeve added growth as gold out-compounded equities. From 2011 to 2021 it reduced both growth and losses: growth fell from 14.36% to 13.23% a year, while the worst fall was 2.9 percentage points smaller.

The evidence does not support two common claims. Gold did not reliably protect in the fastest crash, in 2020, and its 10.0% a year since 2000 is one outcome that followed a 19-year fall, not an expected return. The 191 windows overlap, so they contain fewer than three independent decades.

What do holding costs and rebalancing change?

Fund costs are small next to gold's own swings, but they are certain: in our scenario the two fund fees took the 10% portfolio's growth from 7.29% to 7.09% a year over the full sample. We deducted 0.20% a year on equities and 0.12% on gold, the published ongoing charges of iShares' Core MSCI World UCITS ETF and Physical Gold ETC, used as cost assumptions rather than product suggestions.

Realised costs can exceed the stated fee. The iShares Physical Gold ETC's net asset value grew 6.44% a year from its launch in April 2011 to 11 November 2026, against 6.66% for its LBMA Gold Price benchmark, a gap of 0.21 percentage points a year. From the 29 January record to 11 November, the fund lost 25.81% in dollars and the benchmark 25.74%. These are declines from the peak, rather than calendar-year returns. Bars and coins carry dealing spreads and storage costs that we did not measure.

The rebalancing rule matters most now. A January reset buys gold if its weight is below target and sells it if it is above target. The outcome depends on subsequent returns and costs; a rebound alone does not establish that rebalancing improved long-run growth. Selling to rebalance in a taxable account can also realise gains, which these pre-tax figures leave out.

What can you check in your own portfolio?

The historical results make the assumptions behind a gold allocation worth checking. The relevant inputs are its current weight, holding costs, currency treatment and rebalancing rule, alongside the spending the portfolio must fund. Your statements and the downloadable data can help compare those inputs without assuming that gold will repeat its strongest decade.

1. Find gold's weight today, not at purchase. Add every gold fund, certificate, bar and coin and divide by your total investments; after the rise to January and the fall since, the weight may be far from what you set.
2. Name the job you expect it to do. In 2007-09 a 10% sleeve took 7.0 points off the fall; in 2020 it took 2.9. Treat the second figure as possible, not the first as guaranteed.
3. Check the rebalancing rule. Identify its dates, thresholds and likely trading costs. Compare scenarios in which gold rebounds, stays flat or continues falling; a past peak does not establish a future fair value.
4. Check the currency you measure it in. From its 29 January record, gold lost 28.4% in dollars and 22.7% in euros; in 2001-03 a dollar gain became a euro loss.
5. Compare the cost with the cushion. Check your product's fee and its record against the benchmark; the iShares fund ended 3.0% below its benchmark over 15.6 years, with an annualised-return gap of 0.21 percentage points.
6. Compare growth alongside losses. The sleeve paid off when gold compounded close to equities and cost growth when it trailed by more than about two points; ask what return you are assuming.