Microeconomic Theory | Industrial Organization | Business Economics

Mengqi Zhang

View Research CV Recent Research

I am a Postdoctoral Fellow at the Watt Family Innovation Center (Media Forensics Hub) and the Wilbur O. and Ann Powers College of Business at Clemson University. I received my Ph.D. in Economics from the University of Colorado Boulder in 2025. My research focuses on information economic theory and its application in different fields.

[email protected] | Google Scholar | LinkedIn | Media Forensics Hub

Recent Research

Persuasion Through Reviewers: Implementation Observability and Commitment

Persuasion can occur when the receiver meets a randomly drawn informed reviewer whose private opinion represents a fixed mapping rule from states to signals. Reviewers have different opinions. By managing their distribution, the sender implements an information structure. When the receiver observes only coarse statistics of the distribution, the sender can commit to an information structure if and only if it is represented by her favorite distribution among those with the same statistics. Committable information structures therefore form the admissible set, which is endogenous to the persuasion problem. We provide a geometric characterization of this set for any given persuasion problem. As an endogenous constraint, partial commitment can force the sender to choose a suboptimal experiment or make the unconstrained optimum admissible only in expectation and at a discounted value. For policy design, the framework identifies which reviewer-type shares should be observable for given objectives without requiring full transparency.

PDF Paper page

Featured research figure 1 — Panel (a) displays the sender’s value function in an example persuasion problem. Panels (b) and (c) show the purely admissible regions, namely, the information structures to which the sender can commit under partial observability. Panel (b) corresponds to the region where v₁ > v₂, represented by the positive-slope segments in Panel (a), while Panel (c) corresponds to the region where v₂ >v₁. In this example, full-commitment benchmark (red) is not purely admissible. The sender must instead move to a constrained optimal admissible strategy (green).

Featured research figure 2 — Panel (a) displays the sender’s value function in a different example persuasion problem. Panels (b) and (c) show the purely admissible regions, namely, the information structures to which the sender can commit under partial observability. Panel (b) corresponds to the region where v₁ > v₂, while Panel (c) corresponds to the region where v₂ >v₁. In this example, full-commitment benchmark (red) is not purely admissible. The sender must instead move to a constrained optimal admissible strategy (blue).

Featured research figure 3 — The lower panels display the sender’s value function in this example. Panels (c1) and (c2) provide the pure-admissibility baseline. Panels (a1) and (b1) focus on fibers where no pure-strategy equilibrium exists. The lower endpoints lie in the red regions and the upper endpoints lie in the blue regions. Since pure admissibility fails in these fibers, expected admissibility can be revealed. The green segments identify the information structures on those fibers that are expectedly admissible. On the high-value branch in Panel (a1), the expectedly admissible strategies yield an expected value of 1. On the low-value branch in Panel (b1), they yield an expected value of 0. Panels (a2) and (b2) illustrate the corresponding value equalization along representative fibers, with the green horizontal segments showing how the receiver’s mixture neutralizes the sender’s hidden-deviation incentive.

Background Photo: Dream Lake, Rocky Mountain National Park, Estes Park, CO (8/17/2022)