Xinghao Yao

School of Economics
The University of Queensland

Brisbane, Queensland, Australia

PhD in Economics
(expected early 2027)

xinghao.yao@uq.edu.au

Research Interests: Behavioural Economics; AI-related; Experimental Economics; Information Economics; Auctions and Market Design.

Job Market Paper

How Robust Is LLM Economic Decision-Making? An Auction Benchmark

with Kenan Kalayci

Abstract — How Robust Is LLM Economic Decision-Making? An Auction Benchmark

Economic robustness requires both appropriate adjustment to changing fundamentals and resistance to information that theory predicts should not affect decisions. Using a benchmark of 78 auction environments, we evaluate six large language models (LLMs) in nine configurations against equilibrium predictions for risk-neutral bidders. We find that strong performance is often specific to particular conditions and decisions. In first-price auctions, high bidding accuracy under uniformly distributed private values does not consistently carry over to normally distributed private values. These limits also emerge across connected decisions: entry or information-purchase frequencies close to theoretical predictions can coexist with inaccurate subsequent bids. Conversely, accurate bids among entrants can coexist with entry frequencies far from those predictions. In first-price auctions with costly entry, historical-price information is associated with bidding changes and generally higher entry-error rates than without this information, despite leaving theoretical predictions unchanged. Without an entry decision, most configurations’ bids change little across historical-price conditions. These findings support evaluating LLMs under the economic conditions and across the connected decisions relevant to their intended use, since isolated successes and aggregate scores can conceal substantial gaps in performance.

Working Papers

Information Disclosure in Private-Value Auctions

with Kenan Kalayci

What You See Is All There Is? A Registered Direct Replication of Enke (2020)

R&R · Journal of the Economic Science Association

Abstract — What You See Is All There Is? A Registered Direct Replication of Enke (2020)

Enke (Q J Econ 135:1363–1398, 2020) shows that when an information source reveals information consistent with a person’s prior, people average what they see and ignore what the selection rule implies about the hidden draws. We reproduce the main finding in a different language, laboratory, and software implementation. Compared with participants who receive an explicit summary of the hidden signals, those who must infer their implications place excessive weight on the mean of the visible signals when estimating the mean of the full set. Evidence on the proposed mechanism is mixed and provides only limited support for the hypothesis that people neglect unobserved information because that information fails to come to mind. We also find that greater computational complexity increases such neglect.

Keywords: Bounded Rationality; Replication; Decision Making; Information

Draft dated 22 July 2026.

Original study: Enke (2020)

Sponsorship Disclosure as a Strategic Informational Cue: A Two-Sided Labelling Model and Meta-Analysis of Influencer Marketing

with Tianyuan Feng Major revision · Journal of the Academy of Marketing Science (FT50)

Abstract — Sponsorship Disclosure as a Strategic Informational Cue: A Two-Sided Labelling Model and Meta-Analysis of Influencer Marketing

Consumers increasingly rely on social media influencers for product advice, yet paid endorsements can make that advice commercially motivated in ways followers cannot fully observe. We integrate a two-sided labelling model with a three-level meta-analysis of 63 articles, 89 studies, 258 effect sizes. Relative to no disclosure, sponsorship disclosure has a negative average effect on consumer responses, but this effect varies substantially across contexts. The evidence points to persuasion knowledge as the central mechanism: disclosures make consumers more aware of persuasive intent, which dampens downstream responses. Effects are more negative for early disclosures, hedonic products, macro-influencers, and some cultural contexts. The findings recast disclosure as a strategic informational cue and help managers and regulators calibrate disclosure design to product, influencer, and audience conditions.

Contribution. My contribution was to develop the paper’s two-sided labelling model.

From Analysis to Affect: Modeling and Measuring Finfluencer Communication with Generative LLMs

with Tianyuan Feng Under review · Journal of the Academy of Marketing Science (FT50)

Abstract — From Analysis to Affect: Modeling and Measuring Finfluencer Communication with Generative LLMs

The same investment view can be presented through analytical reasoning or emotional expression. We examine how the former shapes finfluencers’ perceived competence and why analysis matters when emotion draws more engagement. Using a codebook-guided large language model, we classify 47,707 posts from 396 finfluencers on X. Emotional expression is associated with more likes and replies than analytical content, with a stronger advantage for emotion-dominant posts by widely followed creators. Our model explains why analysis can remain valuable despite this disadvantage: the reputation it builds supports engagement with both formats. Two preregistered experiments reproduce emotion’s advantage in liking intentions, strengthened by a higher displayed follower count, while showing that analysis increases perceived competence. Mediation analyses link this competence benefit to following and subscription intentions. Together, the model and evidence suggest that analysis can contribute to an engagement-oriented strategy by signaling competence and supporting interest in future content.

Contribution. My contribution comprises the economic modelling and generative LLM-based textual analysis, including the classification of semantic content and discrete emotion frames.

Work in Progress

Accuracy or Affirmation? Competition in AI Advice Markets

Underquoting in Endogenous Entry Auctions

Do you know how attractive you are? A Two-Sided Search Model with Self-Uncertainty

Collaborative Projects

Project Silicon

Contribution — Project Silicon

Our team contributes experimental data and replication materials and reviews simulation results based on our experiments.

Conferences

2026

Asia-Pacific ESA Meeting, University of Melbourne · Presenter

Australasian Economic Theory Workshop (AETW), Auckland · Presenter

2025

Australian PhD Conference in Economics and Business, UNSW · Presenter

Summer School in Experimental and Behavioral Economics, University of Crete · Participant

2024
Foundations of Utility and Risk (FUR), University of Queensland · Presenter