Abstract
In many two-sided markets, the parties to be matched have incomplete information about their characteristics. We consider the settings where the parties engaged are extremely patient and are interested in long-term partnerships. Hence, once the final matches are determined, they persist for a long time. Each side has an opportunity to learn (some) relevant information about the other before final matches are made. For instance, clients seeking workers to perform tasks often conduct interviews that require the workers to perform some tasks and thereby provide information to both sides. The performance of a worker in such an interview—and hence the information revealed—depends both on the inherent characteristics of the worker and the task and also on the actions taken by the worker (e.g., the effort expended), which are not observed by the client. Thus there is moral hazard. Our goal is to derive a dynamic matching mechanism that facilitates learning on both sides before final matches are achieved and ensures that the worker side does not have incentive to obscure learning of their characteristics through their actions. We derive such a mechanism that leads to final matching that achieve optimal performance (revenue) in equilibrium. We show that the equilibrium strategy is long-run coalitionally stable, which means there is no subset of workers and clients that can gain by deviating from the equilibrium strategy. We derive all the results under the modeling assumption that the utilities of the agents are defined as limit of means of the utility obtained in each interaction.
Introduction
Matching workers to tasks in digital marketplaces is a fundamental economic challenge, particularly when two types of information asymmetry coexist: Adverse Selection (hidden worker skills) and Moral Hazard (unobservable effort). In environments like medical residency matching or high-skill freelance platforms, these factors are deeply entangled. A client observing high output cannot easily distinguish between a high-skill worker and a low-skill worker exerting maximum effort. Furthermore, workers may strategically manipulate their performance during trial periods to influence future matches or compensation.
This paper addresses these challenges by proposing a mechanism designed for "extremely patient" agents—those who prioritize long-term partnership stability over short-term gains. By modeling utility as the limit of means, the authors develop a framework where the long-term consequences of a match outweigh the costs of an initial learning phase.
The FILI Mechanism
The proposed solution is the FILI (First Impression is the Last Impression) mechanism. It structures the matching process into three distinct phases to decouple the learning period from the earning period:
- Assessment Phase: Every worker is matched with every client or task exactly once. The planner observes the resulting outputs. Because agents are modeled with limit-of-means utility, the finite costs incurred during this phase do not diminish their long-run utility.
- Reporting Phase: Workers submit their preferred rankings of tasks. Simultaneously, the planner ranks workers for each client based on the objective outputs observed during the assessment phase.
- Operational Phase: The planner applies the Gale-Shapley (GS) algorithm, with workers as proposers, to establish a permanent, stable matching.
Equilibrium and Strategy
The core of the FILI mechanism's success lies in the MTBB Strategy (Maximum effort, Truthful reporting, Bang-Bang effort). The authors prove that under FILI, it is a weakly dominant strategy for workers to exert maximum effort during the assessment phase to "prove" their productivity and subsequently report their preferences truthfully.
In the long-run operational phase, rational workers adopt a "bang-bang" effort profile: they either exert maximum effort if the task's rewards justify the cost or zero effort if they do not. This equilibrium ensures that the "first impression" created during assessment serves as a reliable proxy for true productivity, as workers have no incentive to obscure their skills when the prize is a superior permanent match.
Key Results and Stability
The paper introduces the concept of Long-run Coalitional Stability. It demonstrates that once the Gale-Shapley matching is finalized, no subset of workers and clients can improve their long-term average utility by deviating from the assigned matches, even when moral hazard is present.
Numerical simulations validate the theoretical findings:
- Revenue Gains: The FILI mechanism, paired with optimal payment rules, achieved revenue gains exceeding 75% compared to standard matching methods based on noisy initial beliefs or resumes.
- Optimality: For homogeneous task types (e.g., all software engineering tasks), the authors prove that FILI achieves the optimal total long-run revenue among all possible matching rules.
- Robustness: Simulations with up to 100 workers and 10,000 iterations confirmed that the mechanism remains effective at scale.
Limitations and Future Directions
While the FILI mechanism provides a robust theoretical foundation for long-term matching, it relies on the assumption of infinite patience (limit-of-means utility). In real-world scenarios, agents often have discount factors that make immediate costs more significant. Additionally, the model assumes static productivity. Future research could explore "learning by doing," where worker skills evolve over time, and the categorization of diverse task types into clusters to maintain the mechanism's optimality in heterogeneous markets.
Summary updated