Jerry Anunrojwong

Battery Operations in Electricity Markets: Strategic Behavior and Distortions

Jerry Anunrojwong, Santiago R. Balseiro, Omar Besbes and Bolun Xu

Proceedings of the ACM Conference on Equity and Access in Algorithms, Mechanisms, and Optimization (EAAMO), 2025.
Proceedings of the ACM Conference on Economics and Computation (EC), 2025.

arXiv SSRN slides media 60-min video

Battery storage can lower generation costs by shifting energy across time, but privately owned batteries may exercise market power. We develop a tractable two-settlement electricity market model and compare centralized cost-minimizing operations with decentralized profit-maximizing operations. Strategic batteries distort storage through quantity withholding, shifting participation from day-ahead to real-time markets, and reduced real-time responsiveness. We quantify the resulting efficiency losses by the Price of Anarchy, defined as the ratio of cost reductions from centralized versus decentralized batteries. We prove that the Price of Anarchy lies between 9/8 and 4/3 for a single strategic battery and is at most 1 + 1/[n(n+2)] for n competing batteries. Calibrations to California and Texas show nontrivial but moderate losses for a single battery, while just a small amount of competition brings outcomes close to efficient.

The Best of Many Robustness Criteria in Decision Making: Formulation and Application to Robust Pricing

Jerry Anunrojwong, Santiago R. Balseiro and Omar Besbes

arXiv SSRN poster slides

This paper initiates a systematic study of overfitting to robustness criteria. How good is a prescription derived from one criterion when evaluated against another criterion? Does there exist a prescription that performs well against all criteria of interest? We formalize and study these questions through the prototypical problem of robust pricing under various information structures, and the three robust criteria: maximin revenue, minimax regret, and maximin ratio. We find that mechanisms optimized for one criterion often perform poorly against other criteria, highlighting the risk of overfitting to a particular robustness criterion. Remarkably, we show it is possible to design mechanisms that achieve good performance across all three criteria simultaneously, suggesting that decision-makers need not compromise among criteria.

Robust Auction Design with Support Information

Jerry Anunrojwong, Santiago R. Balseiro and Omar Besbes

arXiv SSRN poster slides

Management Science, 2025.
Proceedings of the ACM Conference on Economics and Computation (EC), 2023.

The seller wants to sell an item to n i.i.d. buyers and only the support [a,b] is known; a/b quantifies relative support information (RSI). The seller either minimizes worst-case regret or maximizes worst-case approximation ratio. We show that i) with low RSI, second-price auctions (SPA) is optimal; ii) with high RSI, SPA is not optimal, and we introduce a new mechanism, the "pooling auction" (POOL), which is optimal; iii) with moderate RSI, a combination of SPA and POOL is optimal. Under POOL, whenever the highest value is above a threshold, the mechanism still allocates to the highest bidder (just like SPA), but otherwise the mechanism allocates to a uniformly random buyer, i.e., pools low types.

On the Robustness of Second-Price Auctions in Prior-Independent Mechanism Design

Jerry Anunrojwong, Santiago R. Balseiro and Omar Besbes

Operations Research, 2024.

journal arXiv SSRN poster EC video slides

Proceedings of the ACM Conference on Economics and Computation (EC), 2022.

Finalist, George Nicholson Student Paper Competition, 2022.
Spotlight Presentation, INFORMS Revenue Management and Pricing (RMP) Conference, 2022.

The seller wants to sell an item to n buyers such that the buyers' valuation distribution is from a given class (i.i.d., mixtures of i.i.d., affiliated and exchangeable, exchangeable, and all distributions) and the seller minimizes worst-case regret. We derive in quasi closed form the minimax values and the associated optimal mechanism. We show that the first three classes admit the same minimax regret, decreasing in n, while the last two have the same minimax regret equal to that of the case n = 1. Across all settings, the optimal mechanisms are second-price auctions with random reserve.

Information Design for Congested Social Services: Optimal Need-Based Persuasion

Jerry Anunrojwong, Krishnamurthy Iyer and Vahideh Manshadi

journal arXiv SSRN 20 min video 10 min video 20 min slides 10 min slides

Management Science, 2022.
Proceedings of the ACM Conference on Economics and Computation (EC), 2020.

Oral Presentation, Workshop on Mechanism Design for Social Good (MD4SG), 2020.
Oral Presentation, MSOM Service SIG, 2021.

We study the effectiveness of information design in reducing congestion in social services catering to users with varied levels of need. Each arriving user decides either to wait for the service by joining an unobservable FCFS queue, or to leave and get her outside option. To reduce congestion, the service provider seeks to share partial information about the current queue length to persuade more low-need users to leave, thereby better serving high-need users. We show that with enough heterogeneity in need, information design not only Pareto dominates full-info and no-info mechanisms, in some regimes it achieves the same welfare as the "first-best".

Persuading Risk-Conscious Agents: A Geometric Approach

Jerry Anunrojwong, Krishnamurthy Iyer and David Lingenbrink

journal arXiv SSRN slides

Operations Research, 2023.
Proceedings of the Conference on Web and Internet Economics (WINE), 2019.

We consider a persuasion problem between a sender and a receiver where the receiver may not be an expected-utility maximizer. We prove that the sender's problem can be reduced to a convex optimization program and bound the number of signals needed. When the receiver has two possible actions, the problem is reduced to a linear program, which yields a canonical set of signals in the optimal mechanism.

Social Learning Under Platform Influence: Consensus and Persistent Disagreement

Ozan Candogan, Nicole Immorlica, Bar Light and Jerry Anunrojwong

arXiv SSRN

Major Revision at Operations Research.

We introduce an opinion dynamics model where agents are connected in a social network, and repeatedly update their opinions based on the content shown to them by the platform's personalized recommendation and their neighbors' opinions. We prove that agents always converge to some limiting opinion, which can be categorized into two groups: extreme consensus where all agents agree on an extreme opinion, and persistent disagreement where agents disagree. We analyze the relationship between extremism and polarization in terms of the strength of the platform's influence and initial opinions.

Computing Equilibria of Prediction Markets via Persuasion

Jerry Anunrojwong, Yiling Chen, Bo Waggoner and Haifeng Xu

conference arXiv

Proceedings of the Conference on Web and Internet Economics (WINE), 2019.

We study the computation of equilibria in prediction markets with two players and three trading opportunities (the "ABA game"). To do so, we show equivalence of prediction market equilibria with those of a simpler signaling game with commitment introduced by Kong and Schoenebeck (2018). We then extend their results by giving computationally efficient algorithms for additional parameter regimes. Our approach leverages a new connection between prediction markets and Bayesian persuasion.

Naive Bayesian Learning in Social Networks

Jerry Anunrojwong, Nat Sothanaphan

conference arXiv slides 20 min video

Proceedings of the ACM Conference on Economics and Computation (EC), 2018.

Suppose people talk to their neighbors in a social network and update their beliefs using Bayes' rule locally, but they naively assume that their neighbors are always independent sources of information. We characterize the consensus belief. Agents that are more centrally located and are more confident have more influence over the consensus, and we know precisely how the two factors interact.