Abstract
In the rapidly evolving landscape of Decentralized Finance (DeFi), lending and borrowing are governed by algorithmic interest rate models rather than traditional banking intermediaries. This paper provides a rigorous analysis of Protocols for Loanable Funds (PLFs)—specifically Compound, Aave, and dYdX. By formalizing a taxonomy of interest rate models and applying econometric frameworks, the study identifies significant market inefficiencies, the emergence of Compound as a market leader, and critical liquidity risks associated with concentrated supply and 100% utilization events.
Introduction
Protocols for Loanable Funds (PLFs) represent a cornerstone of the DeFi ecosystem, enabling permissionless lending and borrowing through immutable smart contracts. Unlike traditional finance, these protocols lack a "lender of last resort," making the design of their algorithmic interest rate curves the primary defense against liquidity crises. This research examines how these models behave under stress, the efficiency of these nascent markets, and the degree of interconnectedness between major protocols.
Taxonomy of Interest Rate Models
The researchers categorize PLF interest rate models into three primary types based on how they respond to capital utilization (the ratio of borrowed funds to total supply):
- Linear: Rates increase at a constant rate relative to utilization.
- Non-linear: Rates follow a curved path, often increasing more rapidly as utilization rises.
- Kinked: The most common model, where the interest rate jumps sharply after a specific "optimal" utilization threshold (e.g., 80% or 90%).
Market Efficiency and Uncovered Interest Parity
To assess market efficiency, the authors apply the Uncovered Interest Parity (UIP) framework, typically used in foreign exchange markets. UIP suggests that the difference in interest rates between two tokens should be offset by expected changes in their exchange rate.
The study rejected both "strict" and "weak" forms of UIP for daily data across 20 market pairs at the 1% significance level. This indicates that DeFi markets are currently capital inefficient, with persistent "carry trade" opportunities—borrowing low-interest tokens to hold high-interest ones—that arbitrageurs have yet to close. Efficiency only begins to emerge slightly when the time horizon is extended to weekly data.
Interconnectedness and Market Leadership
Using a Vector Error Correction Model (VECM), the researchers analyzed how interest rate "shocks" in one protocol affect others. The results identify Compound as the clear "market leader" or "central bank" of the PLF space. When Compound adjusts its rates, competing protocols like Aave and dYdX move to match them. For instance, in the USDC market, Aave was found to adjust its rates to match Compound's with a coefficient of 0.607, significantly faster than dYdX's 0.115. Conversely, rate changes in Aave or dYdX have negligible impact on Compound.
Liquidity Risks and Concentration
A critical finding of the research is the vulnerability of PLFs to illiquidity. Because these protocols allow for 100% utilization, there are periods where depositors are unable to withdraw their funds because all capital is currently lent out.
This risk is exacerbated by extreme liquidity concentration. For DAI on Compound, the study found that just three accounts controlled over 50% of total liquidity. A withdrawal by a single "whale" could instantly trigger a state of total illiquidity, effectively locking other depositors into the protocol until borrowers repay or new supply enters.
Conclusion
While PLFs offer a revolutionary approach to financial intermediation, they remain in an early, inefficient stage. The clustering of interest rates around "kinks" in the algorithmic models and the dominance of Compound suggest a market still finding its footing. Furthermore, the concentration of liquidity and the reality of 100% utilization events highlight significant systemic risks that must be addressed as the ecosystem matures. Future developments in on-chain identity and credit systems may eventually allow for more efficient, under-collateralized lending models.
Summary updated