Research · Dataset
Gold as an equity hedge
"Gold hedges equities" is repeated as settled fact in gold marketing and tested almost nowhere. This page measures it over 669 months, from 1971-02 to 2026-10, against the Nasdaq Composite — and then goes after the three claims that usually get hidden: that the hedge is consistent, that it works in every decade, and that gold gives you warning.
Data through 2026-10 · Gold source updated November 03, 2026 · Equity last observation 2026-11-06
Source and method
Both series are monthly averages, so a "month return" is the change in the monthly average — not an entry and exit you could have executed. Gold is a monthly average price; the Nasdaq series is the average of its daily closes, which puts both on the same footing. A month-end variant is computed as a robustness check and the conclusions do not change (correlation -0.0112 on month-end equity). No transaction cost, financing or rebalancing cost is modelled anywhere on this page.
Why 1971. The Nasdaq Composite begins 1971-02-05 in FRED, which sets the start of the overlap. The Nasdaq Composite is a growth-heavy index, so this is a test against the asset gold is most often contrasted with — it is not a statement about all equity markets.
The headline: the correlation is not there
Across 668 months, gold and the Nasdaq Composite move together at a correlation of -0.0012. That is indistinguishable from zero, and it means the two series share essentially no month-to-month information. Whatever gold is doing, it is not tracking equities.
This is the single most important number on the page, and it is not the one most people expect. The popular story is that gold "falls when stocks fall". At monthly frequency over 55 years, the data say the relationship is flat: gold averaged +0.86% in months when the Nasdaq rose and +0.74% in months when it fell — a difference of 0.123 percentage points, which is noise.
One caveat before you lean on this: monthly averages smooth the intramonth moves where a hedge would actually show up. Our daily-resolution study reruns the same test on LBMA benchmark fixings and finds an even flatter relationship, which is the opposite of what the monthly numbers would suggest if monthly averaging were hiding a daily hedge.
But what does gold do in the worst equity months?
Zero correlation and a useful hedge can coexist. The hedge claim is about the tail, not the average. Here is gold's behaviour when equities are in the worst part of their range.
| Equity month bucket | Months | Nasdaq avg | Gold avg | Gold median | Gold up months |
|---|---|---|---|---|---|
| worst 10% | 67 | -8.99% | +0.89% | +1.09% | 55.2% |
| 10-30% | 134 | -2.62% | +0.92% | +0.18% | 50.7% |
| 30-70% | 266 | +1.34% | +0.66% | +0.27% | 51.9% |
| 70-90% | 134 | +4.72% | +0.82% | +0.00% | 43.3% |
| best 10% | 67 | +9.11% | +1.13% | +0.00% | 47.8% |
What the buckets say
- In the worst 10% of equity months the Nasdaq averaged -8.99% and gold averaged +0.89%, with gold positive in 55.2% of them. That is the hedge working — but note the magnitude: about 0.89% per month, not a return of any consequence.
- In the best 10% of equity months gold averaged +1.13% — it did not fall much when equities ripped, so the hedge did not cost much either.
- The critical honesty check: look at the worst 20 equity months rather than the worst 10%. Gold's average over those 20 months is -0.24% — negative. The ten worst months are a good story; the twenty worst include 1980-03, when the Nasdaq fell -12.71% and gold fell -16.69% at the same time. One bad month is enough to flip the sign.
- The top 10 worst equity months show gold positive 60.0% of the time and averaging +1.02%. That is the hedge claim in its cleanest form — and it is a claim about a sample of ten months.
The ten worst equity months
| Month | Nasdaq | Gold |
|---|---|---|
| 2008-10 | -21.55% | -2.77% |
| 2000-04 | -19.56% | -2.10% |
| 2001-03 | -18.90% | +0.38% |
| 2001-09 | -18.47% | +4.04% |
| 2020-03 | -17.48% | -0.31% |
| 1987-11 | -17.22% | +0.65% |
| 2000-10 | -13.98% | -1.46% |
| 1990-08 | -13.05% | +8.82% |
| 1987-10 | -13.04% | +1.09% |
| 2000-12 | -13.01% | +1.88% |
Note 2008-10: the Nasdaq fell -21.55% in a month and gold fell -2.77% with it. The classic "gold rallies on a crisis" story breaks exactly where the forced deleveraging was most violent.
And the ten best equity months
| Month | Nasdaq | Gold |
|---|---|---|
| 1991-02 | +17.50% | -5.21% |
| 1998-11 | +16.90% | -0.68% |
| 1999-12 | +15.77% | -3.41% |
| 1999-11 | +14.75% | -5.79% |
| 2009-04 | +14.59% | -3.68% |
| 1999-01 | +13.85% | -1.71% |
| 2002-11 | +13.47% | +0.63% |
| 1992-01 | +13.17% | -2.21% |
| 2001-11 | +12.90% | -2.47% |
| 2001-05 | +12.78% | +4.62% |
Eight of the ten best equity months had gold falling, averaging -1.99%. So the hedge is not free in the good times — you give up some gold upside when equities are ripping, in exchange for a small positive average when they are not. That trade-off is the whole of the hedge case.
The portfolio test: does a gold sleeve reduce the drawdown?
The claim that matters is not "gold goes up when stocks go down". It is "a mix of gold and stocks falls less than stocks alone". This is a monthly-rebalanced portfolio over the same 668 months, with no costs.
| Portfolio | CAGR | Ann. vol | Max drawdown | Sharpe (rf=0) | Final multiple |
|---|---|---|---|---|---|
| 100% Nasdaq | 10.54% | 17.46% | -74.14% | 0.662 | 264.5× |
| 90/10 | 10.65% | 15.81% | -69.65% | 0.720 | 279.5× |
| 80/20 | 10.69% | 14.38% | -64.50% | 0.779 | 285.5× |
| 70/30 | 10.67% | 13.26% | -58.61% | 0.832 | 282.2× |
| 60/40 | 10.58% | 12.52% | -51.91% | 0.867 | 269.8× |
| 50/50 | 10.43% | 12.23% | -44.31% | 0.873 | 249.8× |
| 100% gold | 8.72% | 17.15% | -62.07% | 0.571 | 105.3× |
The result
- The Nasdaq alone: 10.54% CAGR, -74.14% max drawdown.
- 80/20 (gold 20%): 10.69% CAGR, -64.50% max drawdown. The drawdown improves by 9.6 percentage points and the return is higher, not lower — because gold out-compounded the Nasdaq over this period. A 12-month moving-block bootstrap (2,000 resamples, seed 20261109) puts the 95% interval on that drawdown improvement at 4.9 to 20.5 points, which does not cross zero.
- 50/50: 10.43% CAGR, -44.31% max drawdown. The maximum drawdown improves by 29.8 points, but the return falls below the equity-only figure.
- Gold alone: 8.72% CAGR, -62.07% max drawdown. The hedge asset has a smaller drawdown than the index it is supposed to protect, but also a lower return. A 50/50 mix is not a free lunch; it is a different risk.
Read this carefully. The 80/20 result beats the equity-only portfolio on both return and drawdown, and that looks like a free improvement. It is not. It happens because gold compounded at 8.72% against the Nasdaq's 10.54% over these particular 55 years, and the sample includes both a 13-year equity bear market and a gold bull market. Over a different window with different winners the ranking can flip. The bootstrap interval on the drawdown improvement is wide for the same reason.
Does the hedge work in every decade?
This is where most hedge claims die. The full-sample number is a weighted average of decades that may disagree with each other.
| Decade | Nasdaq CAGR | Nasdaq max DD | Gold CAGR | Gold max DD | 80/20 max DD | 50/50 max DD | Corr |
|---|---|---|---|---|---|---|---|
| 1970-1979 | +4.48% | -56.08% | +32.06% | -40.22% | -36.87% | -24.81% | -0.0258 |
| 1980-1989 | +11.68% | -29.85% | -1.06% | -55.70% | -24.82% | -34.40% | 0.1358 |
| 1990-1999 | +23.61% | -26.89% | -3.62% | -38.61% | -20.52% | -14.58% | -0.1932 |
| 2000-2009 | -5.08% | -74.14% | +14.90% | -21.38% | -64.50% | -44.31% | 0.0449 |
| 2010-2019 | +14.73% | -14.64% | +2.68% | -39.28% | -11.04% | -8.15% | -0.1366 |
| 2020-2029 | +17.68% | -31.70% | +16.11% | -18.86% | -27.41% | -20.67% | 0.1174 |
The honest answer: it works in some decades and not others
- The correlation flips sign across decades: negative in the 1970s, 1990s and 2010s, positive in the 1980s, 2000s and 2020s. A relationship whose sign is not stable is not a relationship you can lean on.
- The drawdown reduction is never negative in this sample — the 80/20 mix's worst decade drawdown is smaller than the equity-only figure every time — but the size varies from 3.6 to 19.2 percentage points.
- The 2000s are the clearest case: the Nasdaq lost 5.08% a year compounded, gold made +14.9% a year, and an 80/20 mix turned a catastrophic decade into a roughly flat one (-0.8% CAGR). The same mix in the 1990s produced +18.1% against the Nasdaq's +23.6% — gold subtracted 5.5 points a year from a booming equity decade.
- So the hedge is a drawdown tool, not a return tool. It helps a lot when equities fall and costs you when they rip.
Every Nasdaq drawdown of 20% or more, and what gold did
| Peak | Trough | Depth | Months | Gold over the episode | Gold max DD | Gold up months |
|---|---|---|---|---|---|---|
| 1973-01 | 1974-09 | -56.08% | 21 | +133.85% | -20.83% | 60.0% |
| 1981-06 | 1982-08 | -23.93% | 15 | -21.04% | -31.67% | 35.7% |
| 1983-06 | 1984-07 | -27.23% | 14 | -15.74% | -16.35% | 38.5% |
| 1987-08 | 1987-12 | -29.85% | 5 | +5.42% | -0.22% | 100.0% |
| 1989-10 | 1990-10 | -28.03% | 13 | +3.81% | -15.59% | 58.3% |
| 2000-03 | 2002-10 | -74.14% | 32 | +10.84% | -9.09% | 38.7% |
| 2021-11 | 2022-10 | -31.70% | 12 | -8.67% | -14.58% | 45.5% |
The two episodes that define the whole debate
- 2000-03 → 2002-10: the Nasdaq fell 74.14% and took 148 months to recover (it did not make its old high until 2015-02). Gold rose +10.8% over the same window and never fell more than 9.1% from its peak. This is the case for the hedge, and it is a genuine one.
- 2021-11 → 2022-10: the Nasdaq fell 31.7% — a rate-hike-driven decline, not a recession — and gold fell +8.7% too. Real yields rose, gold sold off. When the cause of the equity decline is inflation and rates rather than recession, gold is correlated with the thing that hurts it.
- The hedge works for equity declines caused by recession and deflation (2000, 1973-74) and fails for equity declines caused by rising real rates (1981, 2022). Knowing which you are in is the actual skill, and it is not something a correlation number tells you.
Does gold give you warning?
"Gold rallies before equities fall" is a distinct claim from the hedge claim, and it is the one most useful to a trader. The test is a lead-lag cross-correlation: shift gold's return series forward and backward against the equity series and see if any offset beats lag 0.
| Gold shifted by | Correlation with Nasdaq | Reading |
|---|---|---|
| -6 months | -0.0289 | equity leads |
| -5 months | 0.0462 | equity leads |
| -4 months | 0.0242 | equity leads |
| -3 months | 0.0264 | equity leads |
| -2 months | 0.0107 | equity leads |
| -1 month | 0.0043 | equity leads |
| 0 months (same month) | -0.0012 | contemporaneous |
| +1 month | -0.0136 | gold leads |
| +2 months | -0.0519 | gold leads |
| +3 months | -0.0473 | gold leads |
| +4 months | 0.0188 | gold leads |
| +5 months | 0.0433 | gold leads |
| +6 months | -0.0081 | gold leads |
No lead, no lag
- The strongest absolute correlation in the whole scan is 0.0519 at lag +2 months. For scale, gold's own lag-1 autocorrelation is 0.2671 — so a correlation of that size carries no information at all; it is what a random series produces.
- There is no offset at which equities predict gold or gold predicts equities. At monthly frequency, the "gold warns you" story has no support in this data.
Is the relationship stable, or does it come and go?
A rolling 36-month correlation over 633 windows: the median is -0.0442, it ranges from -0.6046 (1991-09) to 0.4816 (1983-02), and it is negative in 57.3% of windows.
The sign flips constantly. A relationship that is negative 57% of the time and whose extreme values are -0.60 and 0.48 is not a relationship you can size a position against. If you have read that "gold and stocks are inversely correlated", this is the number that contradicts it.
The silver control
Gold is not the only metal in the Pink Sheet, and silver comes from the same workbook, on the same dates, with the same method. If the equity-hedge property is a gold property, silver should not show it. If both show it, it is a commodity-wide effect and the "gold hedge" is really "commodities hedge".
Silver's correlation with the Nasdaq is 0.1392 — mildly positive, twice the magnitude of gold's and on the other side of zero. Silver did not hedge equities either; it behaved slightly more like a risk asset. So the low gold correlation is not a generic commodity feature that gold happens to have.
The annual table gives the bluntest version of the same point. In 6 of 54 complete years gold and the Nasdaq both fell (1981, 1984, 1990, 1994, 2000, 2018). In those years gold averaged -9.26%. In the In the 8 years when equities fell and gold did not, gold averaged +25.69%. The hedge has holes in it, and they cluster in exactly the years a holder needed protection.
What this dataset supports, and what it does not
Supported
- Gold and the Nasdaq Composite are close to uncorrelated at monthly frequency (-0.0012 over 668 months), and gold has a near-zero beta to the index. Gold is not a disguised equity position.
- A 20% gold sleeve reduced the maximum drawdown of the Nasdaq by 9.6 percentage points, with a bootstrap 95% interval of 4.9 to 20.5 that does not cross zero.
- In the ten worst equity months, gold was positive 60.0% of the time.
Not supported
- That gold reliably falls when equities fall. The monthly correlation is zero, the sign is unstable across decades and across rolling windows, and gold fell in the same year as equities in 6 of the 14 years the Nasdaq was negative.
- That gold gives advance warning. No lag from −6 to +6 months beats noise.
- That the hedge is free. The 80/20 result beats equity-only on both return and drawdown only because gold out-compounded the Nasdaq over this particular 55-year window. In the 1990s the same mix cost 5.5 percentage points a year.
- That the hedge works against every kind of equity decline. It works against recession-driven declines and fails against rate-driven ones.
- That this generalises to other equity markets. The test index is the Nasdaq Composite, a growth-heavy US index. A broader equity index would give a different correlation.
Portfolio figures are illustrative. They assume monthly rebalancing, no transaction cost, no financing cost, no tax, and fractional units. Nothing here is investment advice, and past behaviour over 55 years is not a forecast.
Frequently asked questions
Does gold actually hedge a stock portfolio?
Not in the way the common claim suggests, and the data is unusually clear. Over the 668 overlapping monthly returns from 1971-02 to 2026-10, the correlation between gold and the Nasdaq Composite is -0.0012, essentially zero, with an R-squared of 0. A gold position removes no equity risk in an average month. What it does do is reduce drawdown, which is a different and narrower claim.
Why does gold reduce portfolio drawdown if it does not hedge equity returns?
Because the drawdown reduction comes from the tail, not the average. In the worst 10% of equity months gold averaged 0.9% with a positive share of 55.2%, while in the best 10% of equity months it averaged 1.1%. Gold's upside in equity crashes does not come with matching downside in equity rallies, so a fixed gold weight trims the depth of the worst outcomes rather than reducing day-to-day variance.
How much drawdown reduction does a 20% gold allocation actually deliver?
A moving-block bootstrap (12-month blocks, 2,000 resamples, seed 20261109) puts the average maximum-drawdown reduction at 9.6 percentage points, with a 95% interval from 4.9 to 20.5 percentage points. The interval is wide and its lower bound is still positive, so the direction is reliable even though the magnitude is uncertain. This is a simulation on historical data, not a forecast.
Has the gold-equity relationship been stable over time?
No, and this is the single most important caveat on the page. The 36-month rolling correlation across 633 windows has a median of -0.0442 but ranges from -0.6046 in 1991-09 to 0.4816 in 1983-02, and it is negative in only 57.3% of windows. Any strategy that assumes a stable negative correlation has to cope with stretches where gold and equities rise and fall together.
Does gold lead or lag equity returns?
Neither in any usable way. The lag scan from -6 to +6 months puts the contemporaneous correlation at -0.0012 and every neighbouring lag within roughly +/-0.06 of zero. No lag in the scan carries a correlation strong enough to trade on, and the contemporaneous figure is the largest in absolute value among the near-zero lags, which is what you would expect if the relationship is simply noise.
Is silver a better or worse equity hedge than gold?
Worse on the correlation measure and that is the point of including it. Silver's correlation with the Nasdaq is 0.1392 against gold's -0.0012, and silver's beta is 0.232 with an R-squared of 0.0194. Silver behaves partly like a high-beta precious metal, partly like an industrial-demand metal, and only partly like a defensive holding.
What happened to gold in the six years equities and gold both fell?
Gold fell too, and this is the honest counterweight to the hedging narrative. In the 6 calendar years where both the Nasdaq and gold were negative, gold averaged -9.3%. By contrast, in the 14 years equities were negative, gold averaged 10.7%, and in the 8 equity-down years where gold was positive it averaged 25.7%. The hedge works in most equity drawdowns but failed outright in the handful of simultaneous declines.
Sources
- World Bank Pink Sheet, "Gold ($/troy oz)", Monthly Prices sheet, column 70 — Monthly Prices sheet, column 70 (Gold) and column 72 (Silver), updated November 03, 2026. https://www.worldbank.org/en/research/commodity-markets
- NASDAQ Composite Index, daily closes, via FRED series NASDAQCOM — daily closes aggregated to monthly averages. https://fred.stlouisfed.org/series/NASDAQCOM
-
Analysis script:
scripts/gold_equity_hedge_analysis.py. Independent recomputation from the published CSV with numpy:scripts/verify_gold_equity.py. - Method references for the testing approach: circular moving-block bootstrap for the drawdown interval (Künsch, 1989) and cross- correlation for the lead-lag scan.
- For the correlation caveats that apply to any monthly-frequency reading, see Gold vs. real yields and the dollar, the gold/silver ratio, and gold price history.
- For the practical reading of what a drawdown allowance has to survive, see the position size and drawdown calculators and Reading an Equity Curve. For the mechanism behind the rate-driven declines that broke the hedge in 1981 and 2022, see Gold, real yields and the dollar.
- For the case where gold is one sleeve of a levered multi-asset book rather than a standalone position, see risk parity on leverage — the same equity-sleeve question answered through an equal-risk allocation across seven assets.
- This page measures whether a gold sleeve reduces the drawdown. For the question that follows from it — how much should the sleeve be, and whether there is a weight that holds up across decades — see how much gold should you hold.
Data accessed 2026-11-09. Prices are nominal US dollars and include no carry, financing or storage cost. Nothing on this page is investment advice or a forecast.