DeFi Liquidations: Volatility and Liquidity

TLDR

This paper analyzes the impact of DeFi lending liquidation volume on short-term volatility. The impact of a 1% increase in USD millions liquidated debt is associated with 0.017% – 0.024% increase in volatility.

78% of all liquidations are between 5 assets type pairs - WBTC, USDC, UNI, LINK, and ETH/WETH.
The leverage effect in crypto is roughly 2x - volatility is twice as impacted by negative vs positive events.
Borrowing interest rates for a lending pool pair are uncorrelated to exchange rates for the pair.
The impact of a 1% increase in USD millions liquidated debt is associated with 0.017% – 0.024% increase in volatility.

This paper doesn't propose anything but rather articulates/analyzes the relationship between liquidation volume and volatility.
The paper quantifies known market phenomena well and the liquidation analysis methodology is robust. This procedure can be applied directly to our future liquidity research.

The analysis in this paper uses historical transaction-level data on Aave, Compound and Maker from December 2020 and December 2022. Data is sourced directly on-chain and from Kaiko.
Swap price data is sampled at 1 minute intervals and volatility is calculated over a 2 hour window. The time series are logarithmically transformed to analyze the marginal effect of liquidation amounts on volatility.
Expected volatility is blocked for via a lag on the dependent variable as an explanatory variable. An additional lagging variable for realized volatility is used for greater persistence.
The leverage effect in crypto is also quantified and accounted for for the analyzed assets.
The key, and most important, analytical method is the 2SLS method. The 2SLS (two-stage least-square) method is proposed to account for endogeniety concerns. This accounts for the reverse causality bias potential when estimating regression coefficients.