Do Bitcoin ETF flows predict the price? 684 days of daily flows

Flows move with the price, not ahead of it

We matched 684 US trading days of total net flows into US spot Bitcoin ETFs, from their launch on 11 January 2024 to 2 October 2026, with BTC prices. Net inflows over the period totalled about US$57.9 billion. The rank correlation between a day’s flow and Bitcoin’s move over the same 24 hours to the US close is 0.44; with the move over the previous three days it is 0.42.

The table shows the pattern plainly. On the fifth of days with the largest outflows, BTC had fallen an average of 1.63% over the same day; on the days with the largest inflows it had risen 1.89%. Flows are reported after the US close, so by the time a reader sees a large inflow, the price move associated with it has generally already happened.

Little information about the next few days

Once the figures are public, the link nearly disappears. The rank correlation between a day’s flow and BTC’s return over the next 24 hours, starting at 14:00 UTC the following day, is 0.01; over 72 hours it is also 0.01. By calendar year the next-day correlation was −0.02 in 2024, +0.02 in 2025 and −0.01 in 2026, while the same-day correlation stayed between 0.33 and 0.64.

Across the five groups, next-day returns do not rise steadily with flows. The largest-outflow days were followed by +12.8 bp over 24 hours, close to the all-day average of +13.5 bp. The difference between largest-inflow and largest-outflow days was −3.6 bp in 2024, −7.1 bp in 2025 and +33.6 bp in 2026: it does not keep a sign. Holding constant the same-day and prior three-day price moves, one standard deviation more flow was associated with +10.7 bp over the next 24 hours (t 1.03), which is not distinguishable from zero.

Why the headline numbers mislead

Big inflow headlines tend to arrive on strong days, which makes the flow look like the cause of the rise. The same-day relationship is consistent with ETF buyers responding to price, with authorised participants creating shares as demand arrives, and with common news moving both. These data cannot separate those channels.

Daily ETF flow is also a partial measure of demand. It does not include direct spot buying, derivatives positioning or flows in other countries’ products, and a creation can be hedged elsewhere by the market maker that provides it.

How the numbers were calculated

Flows are Farside Investors’ published daily totals for US spot Bitcoin ETFs, in US$ millions, with market holidays removed. BTC prices are Binance BTC/USDT hourly opens. Same-day moves run from 16:00 to 16:00 New York time. Forward returns start at 14:00 UTC on the next calendar day, by which time the issuers’ daily figures are normally published, and run 24 or 72 hours. Correlations are Spearman rank correlations; the regression uses ordinary least squares.

Days overlap in the 72-hour returns, so those observations are not independent. Third-party flow tables can be revised after first publication; the fingerprint identifies the snapshot used.

Do Bitcoin ETF inflows push the price up?

Inflows and price rises occur together: large-inflow days had already seen BTC up 1.89% on average over the same day. That does not show which causes which, and the day’s published flow told us almost nothing about the next one to three days.

Can you trade Bitcoin on daily ETF flow data?

Not on this evidence. The correlation between a published day’s flow and BTC’s next 24 or 72 hours was about 0.01, and the gap between big-inflow and big-outflow days changed sign from year to year.

Limitations

One third-party flow series, one spot venue and about two and three-quarter years of data. Publication times vary by issuer, so the entry clock is an approximation. No causal identification, fees, spreads or account-level backtest are included; this is not trading advice.

Sources and data access

• Farside Investors: Bitcoin ETF flow, all data
• SEC: approval of spot Bitcoin ETP shares
• Binance public data and archive formats

Cited sources document external facts or data formats; they do not endorse our findings. The public snapshot contains the table and methods. Source-file fingerprints are included for recalculated datasets. Full archived raw files are not distributed here; fingerprints identify the files used, rather than independently proving the results.

Aggregate data snapshot (JSON) · Markdown · Methods and limitations

Live prior 30-day hourly baseline — separate from the historical study

Only the current watch list is shown, up to 12 assets. Missing or stale estimates remain unavailable. Venue and quote currency are kept separate; beta needs matching BTC data. Hourly statistics are not minute-move probabilities.

Observation cutoff (UTC): 24/10/2026, 14:43

Related research