Regulating Matching Markets with Constraints: Data-driven Taxation

Akira Matsushita, Kei Ikegami, Kyohei Okumura, Yoji Tomita, Atsushi IwasakiarXiv 2022 · arXiv · OpenReview

TL;DR

This paper introduces a transferable utility matching model that incorporates regional constraints (like caps and floors on matches) to evaluate the efficiency of different regulatory policies. They demonstrate that a data-driven taxation policy can achieve better social welfare outcomes than traditional cap-based regulations, using data from Japan's Residency Matching Program.

What Problem Does This Paper Address?

Real-world matching markets, like those for medical residents, often have distributional goals, such as ensuring a minimum number of doctors in rural areas. Policymakers try to achieve these goals using regulations like caps on the number of matches in certain regions. However, these "blunt instruments" can lead to significant efficiency losses by preventing potentially high-surplus matches. This paper investigates whether alternative policies, specifically monetary interventions like taxes and subsidies, can achieve the same distributional goals more efficiently and with higher social welfare. They aim to provide a framework for evaluating and designing such policies.

Methods

The core of the paper is a transferable utility (TU) matching model that explicitly incorporates regional constraints (caps and floors). They extend the work of Galichon and Salanié (2021a) to allow for empirical estimation using aggregate-level matching data and observed transfers (like salaries). This framework enables them to:

  • Estimate preference structures: They use a discrete choice model and aggregate matching data to infer the underlying utilities of doctors and hospitals.
  • Design optimal taxation policies: They formulate a social surplus maximization problem subject to regional constraints, showing that an optimal taxation policy can implement the welfare-maximizing matching. This taxation policy is shown to strictly outperform any cap-based policy.
  • Conduct counterfactual simulations: They compare the welfare outcomes of different policies: the status quo (Artificial Caps), an unregulated market (No Caps), and an Optimal Subsidy policy.

Datasets

The primary dataset used is from the Japan Residency Matching Program (JRMP), covering matching results from 2016 to 2019. This data includes:

  • Matching patterns between medical schools and hospitals.
  • Characteristics of medical schools and hospitals (e.g., size, location, public/private status, T-scores).
  • Salaries paid to residents.

Key Insights and Ideas

  • Caps are inefficient: The paper strongly suggests that quantity restrictions (caps) are inherently inefficient for achieving distributional goals in matching markets. They lead to substantial welfare losses compared to more flexible monetary interventions.
  • Subsidies are powerful: Targeted subsidies can achieve distributional goals (like ensuring rural placements) with minimal welfare loss, often approaching the welfare levels of an unconstrained market.
  • Data-driven policy design: The framework allows policymakers to design optimal tax/subsidy schemes based on observed market data, rather than relying on potentially suboptimal fixed caps.
  • Horizontal differentiation matters: Preferences are not just about ranking but also about specific characteristics. Both doctors and hospitals have preferences influenced by factors like distance and past matches, which the model captures.

Key Results

  • The current cap-based policy in the Japan Residency Matching Program generates substantial efficiency losses.
  • A small, targeted subsidy policy can achieve the same distributional goals as the current caps but with significantly higher social welfare.
  • Even sophisticated cap-based algorithms (like the Flexible Deferred Acceptance algorithm) still result in considerable welfare losses compared to an optimal subsidy policy.
  • The welfare loss from the JRMP's cap-based policy is largely attributable to the choice of instrument (caps) rather than the distributional goal itself.

Limitations and Future Work

  • Unobserved heterogeneity: While the model accounts for unobserved heterogeneity in preferences, the empirical estimation relies on aggregate data, which can be a limitation.
  • Data availability: The study uses salary as a proxy for monetary transfers, acknowledging that other non-monetary aspects also contribute to transfers.
  • Future research: The authors suggest extending the approach to analyze more complex distributional constraints and to explore the long-run effects of policies on physician career paths and settlement patterns.

Summary updated

Abstract

This paper develops a framework to conduct a counterfactual analysis to regulate matching markets with regional constraints that impose lower and upper bounds on the number of matches in each region. Our work is motivated by the Japan Residency Matching Program, in which the policymaker wants to guarantee the least number of doctors working in rural regions to achieve the minimum standard of service. Among the multiple possible policies that satisfy such constraints, a policymaker wants to choose the best. To this end, we develop a discrete choice model approach that estimates the utility functions of agents from observed data and predicts agents' behavior under different counterfactual policies. Our framework also allows the policymaker to design the welfare-maximizing tax scheme, which outperforms the policy currently used in practice. Furthermore, a numerical experiment illustrates how our method works.
BibTeX
@misc{matsushita2022regulating,
  author = {Akira Matsushita and Kei Ikegami and Kyohei Okumura and Yoji Tomita and Atsushi Iwasaki},
  title = {{Regulating Matching Markets with Constraints: Data-driven Taxation}},
  year = {2022},
  eprint = {2205.14387},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2205.14387},
}
Figure