Cost of Decentralization: Governance-Free Design and User Adoption in a DeFi Stablecoin Bank—An Empirical Investigation

Abstract

Abstract Crypto-asset services without governance mechanisms maximize transparency and censorship resistance through automation but may sacrifice adaptability to changing market conditions, depending on their institutional design. This study examines the consequences of user adoption for a fully automated stablecoin bank that offers zero-interest loans: Liquity Protocol. Using 1586 daily observations from April 2021 to August 2025, this paper investigates whether user decline stems from portfolio allocation rationale or internal design constraints, under heightened competitive pressure and a tight monetary policy environment. We employ probit specifications to analyze the relationship between stablecoin (LUSD) peg deviations and three behavioral outcomes: collateralization adjustments, loan position closures, and capital withdrawals. Results provide strong evidence that negative peg deviations predict defensive position management, with marginal effects that are 4–6 times larger during post-competitive shock periods. The closure of loan positions exhibits the greatest sensitivity, with 8.7 percentage points across the pre-shock period versus 51.8 percentage points post-shock. In comparison, collateralization ratios increased significantly by 6.0 percentage points, versus 38.8 percentage points in the same periods, indicating a systematic deterioration in capital efficiency. By contrast, the directional probability of capital flight during the post-shock period remains comparatively insignificant. An extension analysis incorporating yield differentials from major competing services is implemented using both probit and OLS specifications. The OLS results show that yield differentials predict larger capital outflows in the pre-shock period ( $$p = 0.023$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mrow> <mml:mi>p</mml:mi> <mml:mo>=</mml:mo> <mml:mn>0.023</mml:mn> </mml:mrow> </mml:math> ), while full-sample and post-shock specifications are not significant. Concurrently, the probit results reveal significant links with the direction of capital withdrawal in the pre-shock period ( $$p = 0.007$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mrow> <mml:mi>p</mml:mi> <mml:mo>=</mml:mo> <mml:mn>0.007</mml:mn> </mml:mrow> </mml:math> ), with no further significant associations in the post-shock period. However, yield differentials show no significant predictive power for the magnitude or direction of position management or collateralization behavior in any specification. The evidence points to a coexistence of mechanisms throughout different temporal periods: yield competition acts as a magnitude amplifier for capital flows prior to 2024, when competitive pressure had not reached its peak. In contrast, as competition reaches a high point for stablecoin saving instruments by early 2024, the systematic day-to-day behavioral dynamics of position management (loan positions and collateral) becomes more consistent with protocol-internal design frictions. Regime-based robustness checks examining Federal Reserve tightening and major crypto market shock periods reveal distinct temporal patterns, with macro stress periods leading to capital flight, whereas active position management in the subsequent period of increasing competitive stress does not. These findings provide insight into the critical design trade-offs between deterministic automation and adaptive governance in the decentralized finance industry, particularly for decentralized banks, with implications for protocol developers and researchers studying the viability of governance-free design subject to alternating external market conditions.