Forecasting Crypto Volatility and Cross-Asset Correlations Using Aave v2 and v3 On-Chain Signals

Abstract

This paper investigates whether variables associated with the leading decentralized finance (DeFi) lending protocol Aave exert measurable effects on the volatilities of Ethereum (ETH) and Bitcoin (BTC), as well as on their dynamic correlations with the S&P 500. We construct a dataset spanning January 2021 to December 2025, covering both Aave V2 and V3, from which we derive explanatory variables including utilization ratios, supply and borrowing rates and measures of pool-level lending risk. Principal component analysis (PCA) is applied to address dimensionality and multicollinearity, yielding two economically interpretable signals: a crypto inflow component capturing risk appetite, and a stablecoin inflow component consistent with flight-to-safety behavior. Cryptocurrency volatilities are computed using the Garman–Klass (GK) range-based estimator while dynamic cross-asset correlations are recovered via the Dynamic Conditional Correlation (DCC) framework. Granger causality tests and ARDL regression models reveal that the Aave risk appetite signal predicts changes in ETH GK volatility at lags 1 and 5 and dynamic correlations at all lags, with weaker effects on BTC volatility. Secondly, we observe a robust negative cumulative effect on all three dynamic correlations, indicating that expansions in DeFi activity systematically dampen crypto–equity co-movements and also lead to a decoupling of BTC and ETH. These results are robust to different volatility specifications and suggest that on-chain DeFi data contains forward-looking information and can help reduce parameter estimation risk, a critical issue in deriving optimal portfolios involving digital assets. We expect that DeFi protocol-level signals will become a standard component of empirical work on digital-asset pricing and risk.