Incentives and Market Structure in Intent-Based Exchanges: Evidence from a Solver-Reward Reform

Technical Summary: Incentives and Market Structure in Intent-Based Exchanges

Problem Statement
Intent-based decentralized exchanges (e.g., CoW Protocol, UniswapX) rely on a competitive class of agents called "solvers" to fulfill user orders. While solvers determine execution prices and capture surplus, the causal impact of the protocol-designed reward rules on market structure and competition remains empirically unstudied. Prior work has focused on user-side execution quality or theoretical models of solver entry, but lacks causal identification of how specific reward-rule changes alter who captures value. This paper addresses the gap by analyzing the effect of a specific governance-mandated reform on the competitive structure of the solver market.

Methodology
The study exploits a natural experiment: the implementation of CIP-74 by CoW Protocol on December 8, 2025. This reform replaced a fixed per-auction reward cap with a cap tied to the protocol revenue of the winning solution and introduced an unconditional ad-valorem volume fee.

The empirical strategy employs a multi-faceted approach on 395 days of public on-chain data (March 1, 2025 – March 30, 2026):

1. Interrupted Time Series (ITS) & Changepoint Detection: The authors estimate level breaks in solver concentration metrics (Volume-weighted HHI, top-solver share) at the reform date, utilizing Newey–West HAC standard errors, max-F scan changepoint detection, and placebo-in-time inference to validate the timing.
2. Mechanism Testing (Order-Size Gradient): To distinguish between general market shifts and the specific incentive mechanism, the authors analyze concentration changes across four order-value buckets (0–1k, 1k–10k, 10k–100k, 100k+ USD). They test for a monotone gradient using Spearman correlation and exact permutation tests.
3. Robustness Checks:
   • Leave-One-Solver-Out: Re-running analyses excluding the incumbent top solver to ensure results are not driven by a single firm's idiosyncratic rise.
   • Cross-Venue Placebo: Comparing CoW's results against UniswapX (which had no rule change at the time) and a market-wide DEX-aggregator HHI to rule out synchronized market trends.
   • Triple-Difference: Utilizing a February 2026 fee cut on stable pairs as a secondary event to test if reducing the revenue pool reverses concentration effects (Proposition 2 prediction).
4. Execution Quality Analysis: Synthetic control and difference-in-differences (DiD) methods are applied to price improvement data to detect changes in average user execution quality.

Theoretical Framework
The paper presents a solver-economics model with two key propositions:

• Proposition 1: An ad-valorem fee is competitively neutral; it shifts all bids equally and does not change the identity of the winner.
• Proposition 2: A revenue-linked reward cap raises the marginal payoff to winning high-value orders. Since inventory-rich solvers have costs that grow slowly with order size, the reform disproportionately benefits them on large orders, leading to a concentration of value share in the large-order segment.

Key Results

1. Aggregate Concentration: Following CIP-74, the volume-weighted HHI rose from 0.176 to 0.241, and the top-solver share increased from 0.296 to 0.395. However, the authors note this aggregate movement is substantially driven by the incumbent top solver (Rizzolver).
2. The Robust Signature (Order-Size Gradient): The primary finding is a monotone reallocation of trading value by order size. Concentration fell in small orders (0–1k: -0.086; 1k–10k: -0.052) and rose in large orders (10k–100k: +0.025; 100k+: +0.085). This gradient is perfectly monotone (Spearman $\rho=1.00$, exact permutation $p=0.042$) and survives the exclusion of the largest solver, confirming the result is a structural shift rather than a single-firm artifact.
3. Value vs. Count: The concentration effect is strictly a "value" phenomenon. By trade count, the market actually de-concentrated (count-HHI -0.060). The reform shifted the share of trading value to incumbents without reducing the number of trades or solvers.
4. Timing Specificity: The structural break in CoW occurred 2–4 days after the reform date. In contrast, UniswapX saw a concentration break 23 days later driven by a different actor, and the market-wide HHI break occurred months prior. This isolates the effect to CoW's specific incentive change.
5. Execution Quality: No detectable change in average execution quality was found. The study has the power to rule out changes larger than $\approx$7 basis points, suggesting the reform did not degrade the typical price users received.
6. Triple-Difference Test: The February 2026 fee cut on stable pairs produced a coefficient ($\delta = -0.042$) in the predicted direction (reducing concentration), but the result was statistically indistinguishable from zero (bounded null), likely due to limited power from the partial nature of the fee cut.

Claims of Significance
The paper claims to provide the first causal measurement of how a solver-reward change reshapes the competitive structure of an intent market. Its significance lies in:

• Mechanism Identification: It links the reallocation of value directly to the revenue-linked reward mechanism (specifically the incentive for inventory-rich solvers on large orders) rather than to general market trends or a single firm's growth.
• Policy Implication: It demonstrates that incentive parameters, often viewed as accounting details, are first-order determinants of market structure. The reform successfully shifted value capture toward incumbents in the high-value segment without altering average execution quality.
• Methodological Contribution: It combines interrupted time-series design, changepoint detection, permutation inference, and cross-venue placebo tests on public on-chain data to isolate causal effects in DeFi markets.

The authors explicitly state that while aggregate concentration rose, the core finding is the order-size reallocation, which is robust to the exclusion of the dominant solver. They do not claim the reform improved or degraded user prices, but rather that it measurably altered who captures the surplus within the intent market.