Blockchain price oracles: Accuracy and violation recovery

Reliable asset price data are critical for the functioning of decentralized finance (DeFi) protocols, particularly those involving collateralized lending. The accuracy of blockchain-based price oracles directly affects key processes such as collateral valuation, liquidation, and risk management. This paper presents a comprehensive empirical analysis of Chainlink Price Feeds (CPFs), the dominant oracle infrastructure in DeFi. We compile a novel dataset of over 150 million observations from 40 CPFs on Ethereum over an 18-month period, matched to benchmark prices from a centralized exchange. To identify the determinants of oracle inaccuracy, we estimate pooled OLS and fixed effects regressions, relating price deviations to design parameters, reporter dynamics, and market conditions. We then introduce a Markov-like state transition framework to model the resolution of target corridor violations, using multinomial logistic regression to estimate transition probabilities. Finally, we exploit position-level data from one of the largest decentralized lending markets and apply entity fixed effects regressions to examine how users adjust collateralization in response to oracle design. Our findings highlight economically significant deviations that are systematically related to oracle accuracy configurations and market stress, and show that users internalize these risks in their financial decisions. The results offer new insights for the design of resilient oracle systems and the management of risk in decentralized financial markets.