Kai Hao Yang


I am a postdoctoral associate at the Cowles Foundation, Yale University. I will join the economics group at Yale School of Management as an assistant professor in July 2021. 

My field of interest is microeconomic theory, in particular, mechanism design, information design and game theory






Working Papers

(Last Update: 03/23/2021) Abstract

The structure of a market describes the nature of competition between participating firms. This paper introduces a framework for designing market structures and characterizes the socially efficient ones. We define a market structure as the set of firms’ strategies and mappings from those strategies to market entry rules, an allocation of products to consumers, and firm revenues. We show that efficient market structures are equivalent to price competition with transfers and yardstick price caps that depend on the published prices of competing firms. As a result, efficient market structures can be implemented without prior knowledge of individual consumer preferences, firms' realized costs, or firm conduct.

(Last Update: 04/15/2021; An earlier version and its supplemental material with additional results) Abstract

A data broker sells market segmentations created by consumer data to a producer with private production cost who sells a product to a unit mass of consumers with heterogeneous values. In this setting, I completely characterize the revenue-maximizing mechanisms for the data broker. In particular, every optimal mechanism induces “quasi-perfect price discrimination” . That is, the data broker sells the producer a market segmentation described by a cost-dependent cutoff, such that all the consumers with values above the cutoff end up buying and paying their values while the rest of consumers do not buy. The characterization of optimal mechanisms leads to additional economically relevant implications. I show that the induced market outcomes remain unchanged even if the data broker becomes more active in the product market by gaining the ability to contract on prices; or by becoming an exclusive retailer, who purchases both the product and the exclusive right to sell the product from the producer, and then sells to the consumers directly. Moreover, vertical integration between the data broker and the producer is Pareto-improving, since consumer surplus is zero under any optimal mechanism. 

(Last Update: 11/21/2019) Abstract

An intermediary has the technology to provide information about a product to consumers and serves as a platform through which transactions between a monopoly and consumers take place. This paper explores the intermediary's revenue maximization problem across all possible business models. By examining the revenue maximizing solutions under three critical business models, I discover that the market outcomes---consumers' expected surplus, producer's expected profit and the intermediary's expected revenue---are equivalent across all business models if and only if the gains from trade are large enough, which provides some insights into, and implications for online selling platforms.

(Last Update: 07/10/2019; R&R at JET) Abstract

This paper explores the buyer-optimal information structures in a monopolistic screening context with nonlinear production technology. It shows that the buyer’s optimal surplus may increase even when the production cost becomes more uncertain or when the efficient surplus decreases. Under a binary prior, this paper further shows that the buyer-optimal information structures must lie in a family described by truncated Pareto distributions. Such characterization effectively reduces the surplus maximization problem to a monopsony’s pricing problem, which further implies that the buyer-optimal surplus is quasi-convex in technologies that are ranked by the rotational order.

(Last Update: 11/14/2019) Abstract

In an environment that features second-degree price discrimination, this paper fully characterizes the set of surplus divisions that can arise from all possible information consumers have about their valuation. By extending the techniques developed in a companion paper (Yang, 2019a), I show that the set of feasible surplus divisions can be characterized by a family of information structures that induce Pareto-distributed interim expected values. Unlike the linear model as in Roesler & Szentes (2017) where posted price is always optimal, the efficient frontier is generically not attainable under any information structures and there are environments in which a (nontrival) subset of the feasible surplus divisions collapses to a one-dimensional set. Nevertheless, the sets of feasible surplus divisions are stable around the linear environments.

(Last Update: 04/20/2017) Abstract

In this paper, without fully specifying the underlying game form, we showed that the probability of an inefficient breakdown in any bilateral crisis bargaining model is smaller when the more informed party has more bargaining power. Moreover, introduction of additional private information does not necessarily lead to extra efficiency loss. Several implications can be drawn from these results. Specifically, regarding international security, reducing incomplete information is not the only way to reduce the probability of war. Instead, reallocating bargaining power properly would also be effective in terms of preventing conflicts. Furthermore, these results also provide a formal justification for the power transition theory as the status-quo power can be interpreted as the party with more bargaining power when the information structure shifts due to power transition.

(Last Update: 04/20/2017) Abstract

This paper investigates the strategic interactions between the counter proliferator and the proliferator in a nuclear proliferation crisis, as well as their impacts on international security and stability. A baseline model of contest with interdependent values is established and its implications are discussed. Furthermore, we characterize the equilibria in a class of models in a "detail-free" fashion and analyze equilibrium outcomes, with particular attentions to international stability and likelihood of a successful development. It thus yields some results and implications that are robust to game forms and model details and provides several generalizations and insights to the effects of various counter-proliferation measures as well as the consequences of nuclear proliferation.

Short Notes

  • Hyperstable Sequential Equilibria. (with Philip Reny, draft coming soon) Abstract
    A set of sequential equilibria is hyperstable if it is a smallest closed set of assessments such that any nearby game obtained by perturbing the payoffs at the end of the tree has a sequential equilibrium near the set. We show that every finite game with perfect recall has a hyperstable set contained a single connected component of sequential equilibrium assessments.
  • A Note on Topological Properties of Outcomes in a Monopoly Market. (Last Update: 10/27/2020) Abstract
    A monopolist with a nonnegative constant marginal cost faces an arbitrary nondecreasing and upper-semicontinuous demand function on \mathbb{R_+} that takes a value in {0,1} outside of a fixed compact interval. This note derives topological properties of outcomes induced by this monopolist's optimal pricing problem. Specifically, the monopolist's optimal profit is continuous in both the marginal cost and the demand (under the weak-* topology); the induced output is lower (upper)-semicontinuous in both the marginal cost and the demand when the monopolist always charges the highest (lowest) optimal price; the optimal price correspondence is upper-hemicontinuous in both the marginal cost and the demand, which in turn implies that the consumer surplus is upper (lower)-semicontinuous in both the marginal cost and the demand when the monopolist always charges the lowest (highest) optimal price. These results further imply similar topological properties of outcomes in settings that feature either second-degree price discrimination or third-degree price discrimination.
  • A Note on Generating Arbitrary Joint Distributions using Partitions. (Last Update: 10/27/2020) Abstract
    Consider a probability space (\Theta,\mathcal{F},\mathbb{P}), two standard Borel spaces (V,\mathcal{V}), (S,\mathcal{S}), and a random variable \mathbf{V}:\Theta \to V. This note shows that for any probability measure \mu \in \Delta(V \times S, \mathcal{V}\otimes \mathcal{S}) with \mathrm{marg}_V \mu=\mathbb{P}\circ\mathbf{V}^{-1}, there exists a random variable \mathbf{S}:\Theta \to S such that (\mathbf{V},\mathbf{S}) has law \mu, provided that (\Theta,\mathcal{F}) is rich relative to \mathbf{V}. This result has applications in generating market segmentations using consumer characteristics; segmenting the residual demand by only partitioning the consumers according their values in a multi-firm, multi-product setting; and connects back to well known results in information economics.


In Progress

  • Revenue Maximization with Rich Allocation Space: Irrelevance of Contractibility
  • Price Discrimination in Oligopoly: An Information Design Approach
  • Profiting from the Money Burnt: Optimal Advertisement Design and its Implications
  • Information Design in Bilateral Trade and (Unintended) Welfare Loss (with Wenji Xu)