Title:
Systemic Risk from Financial Leverage in Digital Asset Markets: Evidence From Perpetual Futures

Authors:
Trong Duc Tran*, Thi Kim Cuc Nguyen, Nam Anh Le, Thi Ha Thu Pham, Viet Nam

Abstract:

This study investigates the predictive power of leverage indicators from perpetual futures markets for cryptocurrency price crashes. Using 2.65 million 8-hour observations from Binance covering 10 major cryptocurrencies during 2023-2024, we estimate a series of panel Logit models with coin fixed effects. Results show that leverage-related indicators, particularly realized volatility, open interest changes, and cumulative funding rates, significantly predict extreme price crashes (≥5% decline within 8 hours). The preferred model (selected via BIC) achieves an out-of-sample AUROC of 0.76 and AUPRC of 0.27, nearly doubling random classification performance. Cross-asset indicators, especially Bitcoin open interest changes, emerge as the strongest predictors (OR = 1.48), supporting the Minsky financial instability hypothesis in crypto markets. Robustness checks across four crash thresholds (3%, 5%, 7%, 10%) and comparison with Random Forest confirm the consistency of findings. The results carry implications for exchange risk management and regulatory early warning system design.

DOI: https://doi.org/10.51505/IJEBMR.2026.10315