Locked In, Levered Up: Risk, Return, and Ruin in DeFi Lending

Abstract

Using comprehensive transaction level loan data for the MakerDAO protocol (2019–2023), this study investigates decentralized finance (DeFi) lending dynamics, focusing on the deter- minants of loan demand and the interplay between leverage, skill, and user performance. We document a counterintuitive positive relationship between the cost of borrowing and loan demand, consistent with yield seeking behavior. Moreover, blockchain- and protocol-specific frictions, such as gas fees shape borrowing activity. At the vault level, leverage universally reduces returns and amplifies liquidation risk, with unskilled users incurring significantly greater losses than skilled counterparts under extreme leverage. While skilled users mitigate moderate leverage risks through active management, excessive leverage erodes performance across all skill levels, with forced liquidations accounting for a significant proportion of this decline. The findings reveal critical trade-offs in DeFi permissionless architecture. While skilled participants exploit leverage strategically, systemic design features disproportionately penalize less sophisticated users.