Key takeaways
- Bitcoin's annualised volatility fell from 58.3% in the 936 days before US spot ETFs to 47.5% in the 937 days after.
- Weekly correlation with the S&P 500 was 0.279 before and 0.281 after, so institutional access left the equity relationship essentially where it started.
- Scaled against S&P 500 volatility, bitcoin's ratio moved only from 3.24 to 3.08, so about three-quarters of the volatility drop was market-wide.
- The calmest twelve months in the whole sample ended on 10 January 2024, the day before the ETFs listed, at 44.6% annualised.
- Bitcoin still fell 53.1% from its October 2025 high to 30 June 2026, with the spot ETFs open for the entire episode.
Say it's the second week of January 2024 and you've been waiting years for this. The SEC has approved spot bitcoin ETFs, and from tomorrow you can buy bitcoin in the same brokerage account that holds your index funds, with no exchange sign-up and no private keys to lose. The pitch that came with it was specific: institutional money would make the asset steadier, and it would sit more comfortably alongside everything else you own.
Two and a half years of prices later, did any of that happen?
Partly, and less than the headline suggests. Bitcoin's annualised volatility over the two and a half years since US spot ETFs began trading was 47.5%. Over the matching stretch before, it was 58.3%. The deepest peak-to-trough fall shrank from 76.7% to 53.1%. Days with a move of 5% or more dropped from 87 to 48.
That reads like a structural change in the thing you own. Most of it isn't.
Bitcoin's weekly correlation with the S&P 500 went from 0.279 to 0.281. Equity volatility fell over the same stretch. Scale bitcoin's volatility by the S&P's and about three-quarters of the improvement disappears. And the calmest twelve months anywhere in the sample ended on 10 January 2024 — the day before the first ETF traded.
What the SEC actually approved
On 10 January 2024 the Commission approved eleven proposed rule changes in a single order, covering listings on NYSE Arca, Nasdaq and Cboe BZX. Trading started the next day. The order sits at 89 FR 3008.
The reasoning matters here, because it's not the reasoning the marketing used. The Commission leaned on how tightly regulated CME bitcoin futures already tracked spot prices. It found the correlation between CME futures and a subset of spot platforms was "no less than 98.4 percent using data at an hourly interval, 94.2 percent using data at a five-minute interval, and 76.9 percent using data at a one-minute interval". The argument was that manipulation in spot would show up in futures, so a surveillance-sharing agreement with the CME could catch it.
Notice what that implies about the promise you were sold. The regulator's own case rested on spot and futures being fused already. Nothing in the order forecast calmer prices, deeper liquidity or a different relationship with shares. Those were market predictions, not regulatory ones, which means prices rather than press releases can settle them for you. The rest of this piece does that.
Volatility fell, and mostly not because of ETFs
Here are your two windows side by side — the two and a half years before the first ETF traded, and the two and a half years since — using daily log returns.
| Measure | 19 Jun 2021 - 10 Jan 2024 | 11 Jan 2024 - 4 Aug 2026 |
|---|---|---|
| Annualised volatility | 58.3% | 47.5% |
| Deepest drawdown | -76.7% | -53.1% |
| Days moving 5% or more | 87 of 935 (9.3%) | 48 of 936 (5.1%) |
| Worst single day | -16.7% (13 Jun 2022) | -15.1% (5 Feb 2026) |
| Price change over window | +31.5% | +38.2% |
So was it the ETFs? Here's the control that answers it. Restrict bitcoin to days when US equities traded, so the two series are measured on identical calendars. Bitcoin's annualised volatility runs 58.8% before and 47.9% after. The S&P 500's runs 18.1% before and 15.5% after.
The ratio is the interesting number. Bitcoin was 3.24 times as volatile as the index in the pre-window and 3.08 times as volatile in the post-window. Bitcoin's own volatility fell by 18.5%. The ratio fell by 4.9%. Roughly three-quarters of the calmer behaviour you've been enjoying is explained by a calmer market for everything.
The residual quarter is real but small, and it sits inside a sample where a single quiet quarter moves the estimate. Anyone quoting you the 58.3% to 47.5% drop without the equity comparison is reporting a market-wide effect and calling it a bitcoin one.
Drawdowns are shallower, and the current one isn't over
The pre-window contains the 2022 collapse: 67,554.84 on 8 November 2021 down to 15,760.14 on 21 November 2022, a fall of 76.7%. The post-window contains a peak of 124,720.09 on 6 October 2025 and a trough of 58,523.93 on 30 June 2026, a fall of 53.1%.
So the worst episode with ETFs open was 23 percentage points shallower than the worst episode without them. That's the strongest single result in the whole exercise, and it still needs three caveats.
First, the drawdown is recent. Suppose you had bought a single coin at that October 2025 peak, paying 124,720.09 for it, in a regulated fund, inside your ordinary brokerage account. By 30 June 2026 it was marked at 58,523.93. On 4 August 2026 it closed at 64,050.51 — off the bottom, and still 48.6% below what you paid. The trough was five weeks before the last observation, and a drawdown you're standing in is not a completed drawdown. CoinShares, writing on 29 July 2026, put the same episode at "over 50%" from the October 2025 high and counted ten falls of more than 50% in bitcoin's eighteen years.
Second, 53% is only shallow by bitcoin's standards. If the gap between volatility and drawdown as measures of risk matters to how you size a position, halving is not a regime change from three-quartering.
Third, the post-window contains three separate drawdowns past 20%: 26.2% into September 2024, 28.2% into April 2025 and the 53.1% episode. The pre-window contains two. The number of times you'd have had to sit through one didn't fall.
The correlation with equities didn't move
This is where the institutional-access story should show up in your portfolio most clearly, and it doesn't.
| Measure | Pre-ETF window | Post-ETF window |
|---|---|---|
| Weekly correlation with S&P 500 | 0.279 | 0.281 |
| Weekly beta to S&P 500 | 0.95 | 0.85 |
| Daily correlation with S&P 500 | 0.415 | 0.385 |
| Daily beta to S&P 500 | 1.24 | 1.09 |
Weekly correlation moved by 0.002, which is nothing you would ever notice on a statement. Daily correlation and both betas drifted slightly lower, which is the opposite of what tighter integration with traditional finance is meant to produce.
Rolling windows tell you why single-number comparisons are fragile here. The trailing 180-day daily correlation was 0.551 in January 2023, 0.145 in January 2024, 0.515 in January 2025 and 0.464 at the start of August 2026. The highest and the lowest readings in that list both come from before the ETFs existed.
One sub-result is genuinely new. Take the worst 10% of weeks for the S&P 500 in each window. Before the ETFs, those thirteen weeks averaged -4.23% for the index and -6.07% for bitcoin. After, they averaged -3.15% for the index and -2.86% for bitcoin. Correlation across all down-weeks for equities fell from 0.383 to 0.081.
Does that count as diversification finally working? Read it carefully before you decide. Bitcoin stopped amplifying equity selloffs mainly because its own worst episode was idiosyncratic. Between 31 March and 30 June 2026 bitcoin fell from 68,221.85 to 58,523.93 while the S&P rose from 6,528.52 to 7,499.36. A low correlation earned that way isn't a hedge doing its job. It's a separate problem arriving on its own schedule, which is a distinction that also shows up in how asset classes actually behaved through past crises.
The case against an ETF effect
The strongest objection is that two years can't separate an ETF effect from a market-cycle effect. It's a good objection, and the data supports it more than it supports the ETF story.
Run the same equal-length window backwards through bitcoin's history. Annualised volatility was 81.0% from July 2016 to November 2018, 79.9% from November 2018 to June 2021, 58.3% from June 2021 to January 2024 and 47.5% since. Volatility has fallen in every window for a decade. The post-ETF reading continues a trend that started six years before anyone filed a prospectus you could buy.
Calendar years make the same point more sharply. Bitcoin's annualised volatility was 64.5% in 2022, 43.6% in 2023, 52.9% in 2024, 41.9% in 2025 and 46.7% in 2026 to 4 August. The first full ETF year was more volatile than the last full pre-ETF year, by nine percentage points.
The rolling measure is the cleanest version. Trailing one-year volatility to 10 January 2024, the last pre-ETF close, was 44.6%. One year later it was 52.7%. So volatility rose during exactly the period when ETF assets were building fastest. If the ETFs suppressed volatility, they did it with a delay long enough to be indistinguishable from the cycle turning.
There's also a confound nobody can strip out. The pre-window covers the fastest tightening cycle in forty years and the post-window covers its unwind, and every risk asset's volatility path is entangled with that. The same discount-rate machinery that shows up in how real yields repriced long-duration assets was operating on bitcoin throughout. No sample of this length can hold it constant.
What the published research says
Hong, Feng, Wang and Li, in a December 2025 arXiv paper from Olin Business School, ran a rolling-correlation and DCC-GARCH study around the approval. They report that bitcoin's correlation with the S&P 500 "increased significantly post-ETF approval", with a Chow test rejecting no-break at p=0.0000. Their pre-window is 1 October 2023 to 9 January 2024 and their post-window is 11 January 2024 to 30 April 2024, so roughly three and a half months each side.
Two things follow. Their finding is the reverse of the one here, and their window is a tenth the length. When I extend the same comparison to two and a half years each side, the correlation shift washes out entirely.
Their descriptive table also needs flagging, because it's the kind of thing you'll see quoted without the caveat. The standard deviations they report for bitcoin, 7,034.9 before and 9,919.4 after, are standard deviations of closing prices in dollars, not of returns. Bitcoin's price roughly doubled between those windows, so a larger dollar standard deviation is arithmetic, not evidence about volatility. The paper does report separate return statistics in the same table, which is the column that would answer the question.
Claims that ETFs damped volatility tend to trace back to industry commentary rather than a test. One 2025 conference review paper asserts that "increased liquidity has promoted better price discovery and less volatility, as per market data" and that price-discovery efficiency rose "by approximately 75%" after launch. Both figures are cited to third parties rather than computed, and the paper is a literature review published in a proceedings volume, so it carries no independent weight on the volatility question.
The theoretical prior runs the other way in any case. Ben-David, Franzoni and Moussawi found that US stocks with higher ETF ownership were more volatile, not less. In their S&P 500 sample a one-standard-deviation rise in ETF ownership went with a 20 basis point rise in daily volatility, about 16% of a standard deviation, with the effect traced to arbitrage flow between the fund and the underlying. The effect shrank to about 5% of a standard deviation across the broader Russell 3000. I read the April 2014 NBER working paper, which was later published in the Journal of Finance in 2018; the published version may differ in detail.
So how much institutional money actually arrived?
Scale is worth pinning down, because "institutional access" is doing a lot of work in the promise made to you. CoinShares put total assets in digital-asset investment products at US$141 billion on 1 June 2026, down from US$148 billion a week earlier. Bitcoin products took US$1.438 billion of outflows that week, the largest weekly bitcoin outflow of 2026, and year-to-date bitcoin flows were a net US$1.2 billion in.
Before you lean on that figure, know what it is. It covers global digital-asset products, not US spot ETFs alone, and it's a rolling series updated weekly. It's also assets rather than flows, so it rises and falls with price as much as with demand. Using it as a demand signal means reading the wrong variable, which is the trap what CoinShares fund flow data really measures sets out in detail.
Still, the direction is informative. Two and a half years after launch, the flow picture in 2026 has been roughly flat, and the asset base shrank by 5% in a single week. The vehicle exists and is large. It hasn't produced one-way institutional demand.
If you want to rebuild this test
Here's exactly what it's made of. Everything above is computed from daily closing prices. Bitcoin comes from Coinbase's BTC-USD market. Equities are the S&P 500 price index, checked month by month against Yahoo Finance's series across the full window. Returns are logarithmic. The last common observation is 4 August 2026.
The event date is 11 January 2024, the first trading day. From there to 4 August 2026 is 937 daily closes. The pre-window is the equivalent stretch immediately before, 19 June 2021 to 10 January 2024, which gives 936 closes. Equal lengths, no overlap, no gap.
It's a deliberately crude design, and the crudeness is the point. A fancier specification would buy precision this sample can't support. Bitcoin's returns cluster into regimes lasting months, so two and a half years on each side is a handful of independent episodes, not two thousand.
What would change the conclusion
Four things would move this.
A longer post-window is the obvious one. If bitcoin's volatility relative to the S&P 500 stays near 3.0 through a full second cycle, the ratio result stops being a two-year fluke. Another five years would do it. Nothing shorter will.
If you want a cleaner answer, a better control would help more than more data. The right comparison isn't bitcoin before and after. It's bitcoin against a crypto asset with no US spot ETF, over the same calendar. If ether or a broad altcoin basket shows the same volatility decline, the ETF explanation is finished. If bitcoin's decline is larger, there's something to explain. That test is doable and isn't in this piece.
Intraday microstructure would settle the mechanism. The ETF channel works through creation and redemption, so its fingerprint should be visible in bid-ask spreads, depth on regulated venues and the basis between spot and CME futures. Daily closes can't see any of that. A study using order-book data could.
And a stress test with the ETFs at full size hasn't happened yet. The 2026 drawdown ran while equities rose. Picture the episode nobody has observed: bitcoin falling hard on the same morning as shares, with millions of ETF holders able to sell in one click from the app they use for everything else. Whether that channel damps or amplifies is the open question, and the arbitrage evidence from equity ETFs suggests amplification is at least as likely.
The honest summary is narrow. Bitcoin has been less volatile and has fallen less deeply since January 2024. Most of the volatility improvement tracks a calmer market for all risk assets. The drawdown improvement is real but rests on one unfinished episode. The correlation with equities is where it was. And the quietest year in the whole sample was the one that ended the day before you could buy the things, which is exactly the pattern you'd expect if the cycle, not the wrapper, is doing the work.