Key takeaways
- Bitcoin's weekly correlation with the S&P 500 was 0.04 from August 2016 to 2019, then 0.24, 0.34 and 0.28 in the periods since.
- Correlation with gold was 0.11 across 522 weeks, and only one calendar year since 2017 exceeded 0.25, which was 2020 at 0.39.
- From 8 November 2021 to 12 October 2022 bitcoin fell 71.6% while the S&P 500 fell 23.9% and gold fell 8.3%.
- In the worst 10% of equity weeks bitcoin averaged -4.47% against the index at -4.24%, while gold averaged plus 0.01%.
- Annualised weekly volatility over the decade was 66.5% for bitcoin, 17.2% for the S&P 500 and 14.5% for gold.
It's the middle of March 2020 and every screen is red. Say you own three things: shares, a little gold, and some bitcoin you bought partly because everybody called it digital gold. On 12 March the S&P falls 9.5%. Gold falls 5.0%. Bitcoin goes from $7,957.02 to $4,980.00 — down 37.4% between one close and the next.
Your insurance has lost four times as much as the thing it was meant to be insuring. So is bitcoin digital gold, or isn't it?
The phrase carries a testable promise: when shares fall, bitcoin holds its value the way gold is supposed to. A decade of weekly prices can settle that, and the answer isn't the one the phrase implies.
Start with the two numbers that do most of the work. Bitcoin's weekly return correlation with the S&P 500 was 0.04 from August 2016 to the end of 2019. It hasn't gone back there since. The reading ran 0.24 through 2020-21, 0.34 through 2022-23 and 0.28 from 2024 to now. Over the trailing 52 weeks it sits at 0.52, the highest figure anywhere in the sample.
Bitcoin's correlation with gold across the same decade was 0.11.
That pair is the whole test. If bitcoin were digital gold in any behavioural sense, you'd expect the second number to be the large one. It isn't, and it never has been for long.
What the digital gold claim actually says
Two things get bundled together under that phrase, and only one of them is what you're buying. The first is a supply claim: bitcoin's issuance is capped and known in advance, which is a design fact and isn't in dispute. The second is a behavioural claim, and that's the one people trade on. It says bitcoin should hold its value when other risk assets are falling.
Only the second claim is testable against prices. Gold's reputation rests on it. So the question is narrow: over the episodes we have, has bitcoin behaved like a haven, or like a leveraged bet on the same thing equities are betting on?
The correlation record, period by period
| Period | Weeks | Bitcoin vs S&P 500 | Bitcoin vs gold | Beta to S&P 500 |
|---|---|---|---|---|
| Aug 2016 - Dec 2019 | 177 | 0.04 | 0.09 | 0.22 |
| 2020-2021 | 105 | 0.24 | 0.21 | 0.77 |
| 2022-2023 | 104 | 0.34 | 0.12 | 1.07 |
| 2024 - Aug 2026 | 136 | 0.28 | 0.07 | 0.84 |
| Full sample | 522 | 0.20 | 0.11 | 0.76 |
Weekly log returns. Correlations computed from FRED and LBMA daily series.
If you bought before 2020, you were holding a different animal from the one in your account now. The step change lands in 2020 and it's large. Beta to the S&P went from 0.22 to above one. The 2022-2023 window is the sharpest: bitcoin moved with equities at 0.34 weekly and with gold at 0.12.
The pre-2020 result isn't an artefact of a short window either. FRED's S&P 500 series only reaches back to August 2016, so I ran the same test against the Nasdaq 100, which goes back decades. Over the full 260 weeks of 2015-2019 the bitcoin-Nasdaq correlation was 0.07.
None of this is a finding you have to take on trust. A Federal Reserve Board staff paper published in 2022 charted 30-day and 90-day rolling correlations between bitcoin and the S&P 500 and described them as starting close to zero at the beginning of 2017 and running much higher afterwards, with spikes around COVID. The New York Fed repeated the chart in its 2024 review of digital-asset stability risks.
The stress episodes, one at a time
Averages hide the thing that matters to you. A haven earns its name in the weeks nobody wants to look at the screen. Here are the four biggest equity drawdowns of the sample, measured peak to trough on the S&P.
| Episode | S&P 500 | Nasdaq 100 | Bitcoin | Gold |
|---|---|---|---|---|
| 19 Feb - 23 Mar 2020 | -33.9% | -27.9% | -31.4% | -4.9% |
| 8 Nov 2021 - 12 Oct 2022 | -23.9% | -34.0% | -71.6% | -8.3% |
| 31 Jul - 5 Aug 2024 | -6.1% | -7.6% | -16.4% | -1.3% |
| 19 Feb - 8 Apr 2025 | -18.9% | -22.9% | -16.0% | +2.7% |
Price indices, not total return. Gold is the LBMA afternoon fix in dollars.
Three of the four rows go the wrong way for the haven story. Sit with the second one for a moment. Equities lost roughly a quarter over those eleven months, gold lost 8.3%, and bitcoin lost 71.6%. Which of the two would you rather have been holding alongside the shares?
The fourth row is the honest exception, and it deserves its due. Through the tariff shock of early 2025 bitcoin fell 16.0% while the S&P fell 18.9% and the Nasdaq 100 fell 22.9%. That's one episode where bitcoin held up better than the index. It lasted about seven weeks. Gold rose 2.7% over the same stretch, which is what a haven is supposed to do.
What happens in the worst weeks for shares
A cleaner way to ask your question is to sort every week by the S&P's return and look only at the tail. Take the worst 10% of weeks, 52 of the 522. The S&P averaged -4.24% in them. Bitcoin averaged -4.47%. Gold averaged +0.01%.
Put another way: in the weeks you'd most want a cushion, bitcoin gave you one 35% of the time. Gold gave you one 56% of the time.
Now the other tail. In the best 10% of weeks, when the S&P averaged +4.20%, bitcoin averaged +1.74%. So in this sample bitcoin matched the index on the way down and lagged it badly on the way up. Its measured beta across all 217 down weeks was 0.87.
That asymmetry is the least reliable number here, because bitcoin's own four-year cycle drives its drift independently of equities. I wouldn't lean on it. The downside figure is the robust one, and it's the one that matters for the claim being tested. This is the same distinction that separates volatility from the drawdown you actually feel: a haven is judged on the left tail, not the average.
Scale matters too. Annualised from weekly returns over the decade, bitcoin's volatility was 66.5%. The S&P's was 17.2% and gold's was 14.5%. Bitcoin is roughly four and a half times as volatile as gold, which tells you most of what a like-for-like swap between the two would do to your portfolio. The New York Fed's 2024 review charted the same ordering across normal and stressed periods.
Bitcoin and gold don't move together
The 0.11 full-sample correlation with gold is statistically distinguishable from zero, at t = 2.6 over 522 weeks. It's also economically trivial. By calendar year the readings were -0.01, 0.17, 0.16, 0.39, -0.03, 0.01, 0.20, -0.03, 0.12 and 0.12 from 2017 to 2026. The only year above 0.25 was 2020. About 22% of rolling 52-week windows were negative.
Recent years make the divergence vivid. In 2025 gold rose 67.4% and bitcoin fell 7.1%. Through 4 August 2026 gold is down 6.5% and bitcoin is down 27.0%. Two assets that supposedly serve the same function in your portfolio have spent two years going opposite directions and then falling together.
The academic work lines up. Ladislav Kristoufek studied quantile correlations of bitcoin against the S&P 500 and the VIX through the COVID crash, comparing them to gold, and concluded the safe-haven story was "unsubstantiated and far-fetched" with gold "a clear winner in this contest". That's a verdict on the single crash where the claim mattered most.
A second line of evidence gets there from the buyers' side rather than the prices'. A BIS working paper testing what actually drives crypto adoption across 95 countries found that bitcoin's own price explains entry far better than global uncertainty does, and wrote that safe-haven explanations "fall short". People show up when the price is rising, not when they're frightened — which is the opposite of how a haven attracts money. The World Gold Council reached the same view from the other side.
So has bitcoin simply not had a fair test yet?
Three objections deserve a real answer, and this is the first of them.
The sample is short and lopsided. 522 weeks sounds like a lot. It isn't. There are four genuinely independent equity stress events in it, and one of them, the 2022 drawdown, coincided with the collapse of leveraged crypto lenders. Fitting a haven claim to four episodes is thin evidence in either direction. Anyone who argues that bitcoin hasn't yet been tested properly has a point the data can't refute.
Bitcoin doesn't respond to macro news. This is the strongest counter and it comes from the Federal Reserve Bank of New York. Gianluca Benigno and Carlo Rosa used intraday data from 2017 to 2022 and found bitcoin was orthogonal to every macroeconomic and monetary surprise they tested except CPI, while gold, silver, the S&P 500 and major exchange rates all responded with large, significant coefficients. If bitcoin were simply high-beta equity exposure, you'd expect it to jump on an FOMC surprise. It doesn't. So the correlation probably reflects a shared risk-appetite channel rather than a shared discount-rate channel, and risk appetite is more regime-dependent than fundamentals.
The regime was unusual. Rates went to zero in 2020, then rose faster than at any point in forty years. Critics of the correlation reading argue that any long-duration speculative asset would have looked equity-like through that. Fair. So what happened afterwards, once spot bitcoin ETFs arrived in 2024 and the marginal buyer changed? The annual correlations were 0.15 in 2024, 0.37 in 2025 and 0.42 so far in 2026. It went up, not down.
One more caveat cuts against the headline. Correlation measures co-movement of weekly changes, not shared direction over time. Right now the 52-week bitcoin-S&P correlation is at its sample high of 0.52, while bitcoin is 48.7% below its record close of $124,720.09 on 5 October 2025 and the S&P is up 15.2% since that day. High correlation, opposite outcomes. Both statements are true and they measure different things.
Gold's own record isn't spotless
It would be unfair to hold bitcoin to a standard gold doesn't meet. Gold fell 11.9% between 9 and 19 March 2020, in the same dollar squeeze that hit everything. What separated it was the recovery. By 30 April 2020 gold was 6.1% above its 19 February level while the S&P was still 14.0% below its own.
And gold has its own ugly stretches, which you'd have had to sit through. It fell 26.1% from a record $5,405.00 fix on 29 January 2026 to $3,993.55 on 16 July 2026. In nominal terms it lost 70.3% between January 1980 and July 1999. Imagine holding through that one. The history is covered in more depth in the piece on the 5-10% gold weight and its 45-year drawdown problem. Neither asset is a free lunch, and broader work on which asset classes actually diversified through past crises makes the same point about most of them.
What would change the conclusion
Four things, each measurable, none of which has happened yet.
First, a full equity bear market of 20% or more in which your bitcoin's peak-to-trough loss is clearly shallower than the index, and lasts longer than the seven weeks of the 2025 tariff episode. One shallow drawdown in a short shock is a hint, not a pattern.
Second, a rolling 52-week bitcoin-equity correlation that stays below 0.10 for two years and survives a stress event inside that window. The reading has been below zero in only 16% of rolling windows since 2016, and the recent trend runs the other way.
Third, a bitcoin-gold correlation that holds above roughly 0.40 for several years. The highest annual reading in a decade was 0.39, in 2020, and it didn't persist into 2021.
And fourth, volatility convergence. At 66.5% against gold's 14.5%, bitcoin would have to become a substantially different asset before the comparison is even fair. Flow data helps here as a leading indicator, and the piece on what CoinShares fund flow figures really measure covers what those numbers can and can't tell you.
If you want to rebuild these numbers
Here's exactly what they're made of. Everything above is computed from daily closes: Coinbase bitcoin and the S&P 500 price index from FRED, and the London afternoon gold auction price from the LBMA. Weekly log returns, Friday to Friday, 522 overlapping weeks from August 2016 to 7 August 2026.
Three caveats on method. All three series are price indices, so dividends are excluded from the S&P figures, which understates equity returns but barely touches correlations. The gold fix is set once a day in London while bitcoin trades continuously, so daily correlations between the two suffer a timing mismatch. Weekly sampling reduces that, which is why the numbers you've been reading are weekly.
And correlation is a linear, unstable statistic. It moves with the window you pick, which is why the table above splits the record into periods rather than quoting one figure. The rolling series ranged from -0.23 in December 2019 to 0.52 in August 2026. So if someone quotes you a single bitcoin correlation number with no date attached, they're quoting an average across regimes that don't resemble each other.
What the record supports is narrow and fairly durable. Bitcoin has behaved as a high-volatility risk asset that falls with equities in stress and has no meaningful relationship with gold. Whether that's a permanent property or a feature of its first decade, the data can't yet say — which leaves you deciding what to call the sleeve, and sizing it accordingly.