Do spot ETF flows predict bitcoin’s volatility?
Not beyond what standard forecasts already capture. From February 2025 to October 2026, adding daily US spot ETF flows to a forecast built from recent realised volatility, options-implied volatility and past returns made next-day bitcoin volatility forecasts slightly worse, not better, and the same held for ether. Flows mostly echo past prices: five-day flows correlate +0.68 with bitcoin’s previous 20-day return.
The claim
Galindo Gil and Lazo-Paz rebuild the price-fragility measure of Greenwood and Thesmar — how volatile the flows of a stock’s owners are — from ETF data, and find that it strongly predicts the stock’s volatility over the next quarter, better than the older mutual-fund version. With US spot bitcoin and ether ETFs moving billions of dollars a week, the obvious question is whether their flows say anything about crypto volatility.
What we tested
Net daily flows into US spot bitcoin and ether ETFs, scaled by market value and aligned to the day they are actually published, January 2024 to October 2026. We added flows — their recent size, direction and variability, including the paper’s fragility measure — to forecasts of one-, five- and ten-day realised volatility built from recent realised volatility, options-implied volatility, past returns, trading volume and the calendar, refitted as data arrive. Forecasts are out of sample from February 2025 for bitcoin and August 2025 for ether.
Result
Flows did not improve the forecasts. For bitcoin’s next-day volatility the rank correlation between forecast and outcome changed by −0.003, and by −0.009 for ether; the fragility measure alone changed it by −0.002 for bitcoin and a negligible +0.002 for ether. Five- and ten-day horizons were too noisy to call. The reason is visible in the data: flows follow prices — five-day flows correlate +0.68 with the previous 20-day bitcoin return and +0.55 for ether — so a model that already knows prices and volatility has little to learn from them.
Effect of adding ETF flows to a next-day volatility forecast (out of sample)
| Measure | BTC | ETH |
|---|---|---|
| Change in forecast accuracy (rank correlation) | −0.003 | −0.009 |
| Fragility measure only | −0.002 | +0.002 |
| 5-day flows vs prior 20-day return (correlation) | +0.68 | +0.55 |
What it means
ETF flows describe who is buying, but as a forecasting input they mostly restate the price. The fragility measure works across many stocks with different ETF owners over quarters; one asset followed day by day gives it nothing to work with.
What would change the verdict
Two full market cycles of ETF history — around 2029 — or quarterly horizons would give the idea a fairer test. ETF premiums and discounts to net asset value, once a clean daily series exists, are a different and more promising input.
This note re-tests a claim made by the cited authors on our own crypto market data; the summary of their method is ours. Tested · published · last reviewed . Research, not investment advice.