← Research · Replication · Market microstructure

Does bitcoin lead ether’s price?

Not in a way that can be used. A 2026 study found bitcoin’s lead over ether strengthening in volatile markets on one-minute data; we reproduced its numbers almost exactly, but the same “lead” appears in simulated markets where the two coins share shocks and neither leads, and ether “leads” bitcoin just as much. Beyond ether’s own moves, bitcoin’s recent returns added a rank correlation of at most 0.008 to ether’s next 5–60 minutes.

Did not replicateBTC and ETH spot and perpetual futures, one-minute, 2020 – 2026Last reviewed

The claim

Source we re-testChernoff, A. R. (2026). Bitcoin, Ethereum, and the Ambiguity of Price Discovery: A Multi-Measure High-Frequency Analysis. Journal of Risk and Financial Management, 19(9), 678. doi:10.3390/jrfm19090678

Chernoff studies four years of one-minute bitcoin and ether trades (2022–2025) with rolling price-discovery, spillover and predictability measures. The answer depends on the metric: Granger causality from bitcoin to ether strengthens significantly in high-volatility regimes, the Hasbrouck information share moves the same way without significance, and the Diebold–Yilmaz net spillover moves the other way. The everyday version is simpler: bitcoin moves first and ether follows.

What we tested

One-minute BTC and ETH prices from spot and perpetual futures markets, 2020 to 2026, in rolling five-day windows with the paper’s measures and volatility regimes. The deciding comparison is a set of simulated markets that keep both coins’ volatility, its regimes and their same-minute co-movement but contain no lagged link at all, so any lead measured there is an artefact. We also measured the reverse direction, ether to bitcoin, which the paper did not report.

Result

The paper’s figures reproduce closely: 22.1% of windows cointegrated against its 22.4%, and Granger causality rising in volatile markets. But the simulated markets without any lead show the same rise — on the paper’s own period the measured increase was +2.78 against +2.79 in the simulations — and the reverse statistic, ether to bitcoin, rose by +2.81. The classic Granger test on minute data is inflated by volatility clustering and fat tails, more so in volatile windows. Measured directly, bitcoin’s recent moves added at most 0.008 of rank correlation to ether’s next 5–60 minutes.

Rise of the Granger statistic in high- versus low-volatility windows

DirectionReal dataSimulated, no lead
Bitcoin → ether (paper’s period)+2.78+2.79
Ether → bitcoin+2.81+2.59

Windows cointegrated: 22.1% in our data, 22.4% in the paper.

What it means

Bitcoin and ether move together within the same minute. A lead worth trading would have to show up after that, and it does not; strong lead statistics in volatile markets mostly measure the volatility.

What would change the verdict

Second-level data with own execution could reveal a sub-minute lead that one-minute bars cannot. A lead that survives simulated markets with shared shocks would change the verdict.

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.