MODELING CREDIT RISK IN DECENTRALIZED DIGITAL ASSET LENDING WITH LIQUIDITY ADJUSTMENT

Abstract

Introduction. In DeFi lending, a position becoming eligible for liquidation does not by itself imply a loss for the protocol. Loss arises when a fall in collateral prices, limited short-horizon liquidity, the liquidation bonus and delayed governance response jointly reduce executable collateral value below outstanding debt.

Purpose. To build a model for probability of loss, loss given default and expected loss in DeFi lending where the core unit of analysis is a position observed under a specific market regime rather than a generic wallet score.

Materials and methods. The paper applies the SLG-EL model, which combines mark health factor, executable position safety factor, a liquidity-adjusted recovery function, a logit specification for loss-PD and a governance-sensitive decision layer. The empirical basis relies on open protocol and market inputs for Aave v3 Ethereum, while stress regimes and representative positions are treated as scenario assumptions. The research design is model-based and scenario-based rather than a full panel back-test.

Results. A position may remain formally safe under the standard health factor while already losing its buffer for loss-free liquidation under the executable metric. For representative WETH and WBTC positions, expected loss rises non-linearly; in the stylised mini-portfolio it increases from zero in the normal regime to 0.27% of EAD in the fragile regime and 2.56% in the stress regime. The strongest sensitivity is observed for large positions relative to available market depth.

Prospects. Further research should validate the framework on full position-level Aave and Compound panels, refine the executable-liquidity function for alternative DEX routes and extend the model to multi-collateral portfolios.