Extreme Perpetual Futures Funding and Subsequent Bitcoin Returns: Intraday Evidence from Binance

Abstract

Perpetual futures funding rates are central to cryptocurrency derivatives markets because they help anchor perpetual-contract prices to spot prices while simultaneously reflecting the state of the derivatives market. This study examines whether historically extreme Bitcoin perpetual-futures funding rates contain information about subsequent Bitcoin spot returns. Binance BTCUSDT perpetual funding observations from 10 September 2019 to 10 August 2026 are aligned with Binance BTCUSDT spot prices sampled at 15-minute frequency. Extreme funding is classified ex ante using the current funding rate's mid-rank percentile within the preceding 180-day distribution, with the lower and upper deciles defining extreme negative and extreme positive states. Bitcoin spot returns are measured over 15-minute, one-hour, four-hour, eight-hour and 24-hour post-settlement horizons. Parametric, non-parametric and directional tests are supplemented by heteroskedasticity and autocorrelation consistent regressions controlling for the four-hour price movement preceding settlement. The main result is concentrated in the negative tail: extreme negative funding is followed by an average 24-hour Bitcoin return of 0.506%, significant under t, Wilcoxon and directional-frequency tests and robust to false-discovery-rate adjustment. Extreme positive funding does not exhibit a comparable 24-hour effect. In categorical HAC regressions, extreme negative funding is associated with approximately 0.491 percentage points of additional 24-hour return after controlling for the pre-settlement return. Non-overlapping 24-hour subsamples retain positive and statistically significant mean returns after extreme negative funding. The evidence suggests that funding-rate information is nonlinear, asymmetric and horizon dependent rather than adequately represented by a simple linear relation between funding and future returns.