Optimal Mediation Mechanism in Bilateral Trade

This paper investigates revenue-maximizing mechanisms for a selfish mediator in bilateral trade with interdependent valuations, establishing an impossibility trilemma between incentive compatibility, obedience, and informativeness, and subsequently characterizing optimal threshold-based mechanisms under conditions of constant seller costs or mediator veto power.

The Gist

Imagine you are trying to sell a used car, and I am trying to buy one. You know exactly how good the car is (maybe you've driven it for years), but I don't. I only know what I like in a car. We both want a deal, but we don't trust each other enough to just shake hands.

Enter the Mediator (like a real estate agent or a car broker). This person doesn't know the car's true condition or my budget, but they are smart enough to design a system to get us to trade. Their goal? To make as much money as possible from the fees they charge us.

This paper asks a big question: How should this mediator design the rules of the game to make the most money, given that neither of them can be forced to trade?

Here is the story of their findings, broken down into simple concepts.

1. The "Menu" Trick

The authors first realized that mediators don't need to create complicated, multi-round games with secret messages and confusing fees. It's too messy.

Instead, they found that the mediator can just offer a two-dimensional menu.

• You (the Buyer) pick a "ticket" from a list based on how much you like cars.
• I (the Seller) pick a "ticket" from a list based on how good my car is.
• When we both pick our tickets, the mediator looks at the combination and says, "Okay, based on these two choices, I recommend you trade at this price."

It's like a vending machine where you pick a flavor and a size, and the machine decides if it will dispense a soda. The mediator just needs to design the menu and the prices.

2. The "Impossible Trilemma" (The Three-Way Standoff)

Here is the big shocker the authors discovered. They proved that a mediator cannot have all three of these things at the same time:

1. Honesty: You and I both tell the truth about our types.
2. Obedience: You and I actually listen to the mediator's advice to trade (or not trade).
3. Useful Information: The mediator actually tells us something new about the car's quality.

The Catch: If the mediator wants us to be honest and obedient, they have to give us zero information. They have to either say "Trade everything!" or "Don't trade anything!" They can't give us a nuanced recommendation like "This car is good, but only if you really like red cars."

Why? Because if the mediator gives us information, we can game the system. We might lie about our preferences to get a better deal, or we might ignore the advice if it doesn't suit us. To keep us in line, the mediator has to stop giving us useful hints.

3. Solution A: When the Seller's Cost is Fixed

Since the "perfect" system is impossible, the authors looked at a specific, realistic scenario: What if the seller's cost to make the item is the same, no matter how good the item is? (Imagine a factory making widgets; the machine costs the same whether the widget is perfect or slightly scratched).

In this case, the mediator can design a clever system:

• The Threshold Rule: The mediator says, "We will only trade if the item's quality is above a certain line."
• The Twist: This "line" changes depending on who the buyer is.
  • If you are a picky buyer (high type), the line is low. You get to trade almost anything.
  • If you are a casual buyer (low type), the line is very high. You only get to trade if the item is perfect.

The Result: The casual buyers get more information (they know exactly what they are getting because the bar is so high), but they pay a higher price. The picky buyers get less information but pay less. The mediator squeezes the seller for all their profit and charges the buyers based on how much "information" they need.

4. Solution B: When the Mediator Has a "Veto"

Now, imagine a different scenario: The mediator is the boss. If they say "No," the deal cannot happen, even if we both want it. (Think of an investment bank blocking a merger).

In this case, the optimal strategy flips completely:

• The Reverse Threshold: The mediator recommends trading only if the item's quality is below a certain line.
• The "Lemons" Effect: High-quality items are actually less likely to be traded!
  • Why? Because high-quality sellers are smart. They know the mediator will only let them trade low-quality items. So, they refuse to participate.
  • This leaves mostly "lemons" (bad items) in the market.
  • The mediator ends up facilitating trades for low-quality items, and the sellers of those items make money, while the high-quality sellers walk away.

The Big Picture

This paper teaches us that in markets where trust is low and information is hidden:

1. Simplicity wins: You don't need complex contracts; a simple menu works best.
2. Information is a double-edged sword: Giving people too much information can break the deal. Sometimes, the mediator has to hide the truth to keep everyone playing by the rules.
3. Power dynamics matter: Whether the mediator can force a deal or just suggest it changes everything. If they can't force a deal, they use information to charge different prices. If they can force a deal, they might accidentally create a market full of bad products (the "Lemons" market).

In short, the mediator is like a magician. To make the most money, they sometimes have to pull a rabbit out of a hat, and sometimes they have to make the hat disappear entirely, depending on how much control they have over the audience.

Technical Summary

Here is a detailed technical summary of the paper "Optimal Mediation Mechanism in Bilateral Trade" by Fan, Shen, Tang, and Wang.

1. Problem Statement

The paper addresses the design of a revenue-maximizing mechanism for a selfish, uninformed mediator facilitating trade between a seller and a buyer.

• Setting: Bilateral trade with interdependent valuations.
  • Seller: Privately observes the item quality $q$. The seller's cost is $c(q)$.
  • Buyer: Has a private type $t$ (preference). The buyer's valuation is $v(q, t) = \alpha(q)t$, depending on both their type and the item's quality.
  • Mediator: Initially uninformed about $q$ and $t$. Designs a mechanism ex-ante to maximize expected revenue (the spread between the buyer's payment and the seller's receipt).
• Constraints:
  • Dynamic Interaction: The mechanism can involve multi-round private communication.
  • Autonomy: Both parties can exit the mechanism at any stage. The transaction only occurs if both parties agree to the mediator's recommendation.
  • Solution Concept: Perfect Bayesian Equilibrium (PBE).
• Core Challenge: The mediator must simultaneously incentivize truthful reporting (Incentive Compatibility, IC) and ensure parties follow the recommendation (Obedience), while the buyer's valuation depends on the seller's private information.

2. Methodology and Framework

A. Reduction to Joint Menu-Selection Mechanisms

Despite the potential complexity of multi-round communication, the authors prove a strong Revelation Principle (Proposition 1). They show that without loss of generality, the mediator can restrict attention to Joint Menu-Selection Mechanisms:

• A two-dimensional menu where the buyer selects an option from a set based on type $t$, and the seller selects an option based on quality $q$.
• The joint selection determines a probability distribution over signals ("trade" or "no trade").
• Based on the signal, the mediator recommends trade and sets prices.
• This simplifies the infinite-dimensional dynamic game into a static optimization problem over functional variables: the signaling rule $\pi(q, t)$ and payment functions $P_b(t)$ and $P_s(q)$.

B. Optimization Formulation

The problem is formulated as maximizing the mediator's revenue subject to:

1. Individual Rationality (IR): Participation yields non-negative utility.
2. Obedience: Given truthful reporting, following the recommendation is optimal.
3. Incentive Compatibility (IC): Truthful reporting is optimal. Crucially, the authors account for "double deviations": a player may misreport their type and choose to disobey the recommendation if it yields higher utility.

C. The Impossibility Trilemma

The authors establish a fundamental impossibility result (Theorem 1):

• Result: No mechanism can simultaneously satisfy IC, Obedience, and Informativeness.
• Implication: Any mechanism that is IC and Obedient must be uninformative (i.e., it always recommends "trade" or always recommends "no trade").
• Proof Logic:
  1. Seller IC and Obedience imply the seller's payment must be constant in profitable regions.
  2. If the mechanism is informative, the buyer's IC constraints force the buyer's payment to be constant.
  3. A constant buyer payment cannot simultaneously satisfy the buyer's IC and Obedience constraints for an informative mechanism.

3. Key Contributions and Results

Due to the impossibility result, the authors characterize optimal mechanisms under two specific relaxation conditions.

Case 1: Constant Seller Cost ($c(q) = c$)

• Assumption: The seller's cost is independent of quality (e.g., fixed production cost, quality is purely a signal of value).
• Optimal Structure: The mechanism exhibits a threshold structure.
  • Trade is recommended if quality $q \geq \lambda(t)$.
  • The threshold $\lambda(t)$ is decreasing in the buyer's type $t$.
• Key Findings:
  • Information Discrimination: Low-type buyers face higher quality thresholds, making the "trade" signal more informative for them.
  • Pricing Anomaly: The buyer's payment function $P_b(t)$ is decreasing in type. Low-type buyers pay more than high-type buyers.
  • Intuition: Because low-type buyers receive more informative signals (higher uncertainty resolved), the mediator has more power to extract surplus from them.
  • Surplus Extraction: The mediator fully extracts the seller's surplus (pays exactly $c$), while the buyer retains positive surplus that increases with type.

Case 2: Mediator with Veto Power

• Assumption: The mediator can unilaterally block a transaction (trade occurs only if the mediator approves). This relaxes the "Obedience" constraint for the "no trade" signal.
• Optimal Structure: The mechanism exhibits a reverse threshold structure.
  • Trade is recommended if quality $q \leq \lambda(t)$.
  • The threshold $\lambda(t)$ is increasing in the buyer's type $t$.
• Key Findings:
  • Lemons Market Effect: High-quality items are less likely to be traded. Sellers with high qualities earn less surplus, discouraging their participation.
  • Market Outcome: The market becomes dominated by lower-quality items, creating a "lemons market" (Akerlof, 1970) dynamic.
  • Virtual Values: The solution utilizes an interdependent virtual value function $\alpha(q)\phi_b^-(t)$ compared against the seller's virtual cost.

Technical Innovations

1. Handling Double Deviations: The authors develop a technique to relax the complex IC constraints (which include double deviations) by first solving a relaxed problem and then proving the solution remains feasible for the original problem.
2. Modified Ironing Procedure: For general distributions where the regularity condition (monotone virtual value) fails, standard Myersonian ironing cannot be directly applied due to the interdependent valuation term $\alpha(q)$. The authors propose a modified ironing procedure that performs ironing on a weighted measure space defined by $\alpha(q)dG(q)$ rather than the standard probability space.

4. Significance

• Theoretical Advancement: The paper bridges the gap between mechanism design and information design in a three-player setting (Mediator, Buyer, Seller) with interdependent values. It extends the revelation principle to dynamic settings with PBE and double deviations.
• Resolution of Impossibility: By identifying the "Impossibility Trilemma," the paper clarifies why standard efficient or informative mechanisms fail in revenue-maximizing mediation with autonomous agents.
• Practical Insights:
  • For Platforms: In markets with interdependent values (e.g., real estate, M&A, used cars), optimal revenue extraction may require counter-intuitive pricing (charging low-types more) or accepting "lemons market" outcomes if the mediator holds veto power.
  • Market Design: The results highlight the trade-off between information revelation and revenue extraction. A mediator cannot simply be a perfect information provider; they must strategically distort information to maximize profit.
• Methodological Contribution: The development of the joint menu-selection framework and the modified ironing technique provides new tools for analyzing mechanism design problems with interdependent valuations and multiple strategic agents.

In summary, the paper demonstrates that in bilateral trade with interdependent valuations, a revenue-maximizing mediator faces a fundamental trade-off between providing information and maintaining incentives. The optimal solution often involves strategic information distortion and non-monotonic pricing structures that differ significantly from standard auction theory.