Research

A list of papers is provided below, each accompanied by a brief summary of its main results, central message, and technical contributions. You can also click the links to access the paper’s PDF, presentation slides, and online appendix.

Working Papers

1. Procrastination and Commitment   [Slides   Online Appendix]   Revise & Resubmit, American Economic Journal: Microeconomics

I propose a tractable model of procrastination where an agent, who exhibits present bias, must complete a certain workload by a fixed deadline. I characterize the agent's work schedule and the total effort costs of completing the task under various task environments. I show that (i) behavioral frictions—present bias and naivete—amplify the agent’s sensitivity to task features, and (ii) given the agent's flow effort cost function, the exponential discounting rate and present bias can be jointly identified from the agent's work trajectory, whereas present bias and naivete are observationally equivalent.

Takeaway: For a present-biased agent, although full commitment is welfare-enhancing, limited commitment—whereby the agent cannot always commit her actions when she gets to make a move—can backfire. I identify a negative effect, Tunnel-Vision Effect, which captures an exacerbation of the self-control problem during unregulated periods when the agent focuses on meeting imminent short-term goals. This effect arises from a combination of present bias and limited commitment, and it gets stronger as  actions become more frequent relative to the commitments. In the limit, when choices are made in continuous time while commitments remain countable, the Tunnel-Vision Effect fully  neutralizes the positive Keeping-on-Track Effect, and may strictly dominate it when short-term goals are not chosen optimally. Therefore, even absent (i) naivete, (ii) taste shocks that generate a demand for flexibility, or (iii) direct costs of commitment devices, an individual may still be better off without committing herself. This result provides a cautionary counterpoint to the bulk of literature on time inconsistency, where commitment can strictly enhance welfare for present-biased agents.

Technical difficulty: To analytically solve dynamic optimization under dynamic inconsistency in finite horizon, I assume a CRRA flow cost of effort and adopt the continuous-time quasi-hyperbolic discounting proposed by Harris and Laibson (2013). I further establish a monotonicity result in discrete time showing that the net value of short-term goals declines as the frequency of actions increases relative to the committed goals, converging to zero in the limit.

2. Signaling Design with Matteo Camboni, Mallesh M. Pai, and Rakesh Vohra  [Slides]   

We revisit the classic job-market signaling model of Spence (1973) and endogenize the signaling opportunities available on the market by introducing profit-seeking schools as intermediaries. Schools design the mapping from candidates’ efforts to job-market signals for fees. We show that, in equilibrium, a monopolist school captures the entire social surplus by committing to an uninformative monitoring policy and charging fees that extract students’ surplus from being hired. In contrast, competition shifts surplus to students but inevitably causes inefficiency by inducing wasteful effort.

Takeaway: First,  assuming signaling opportunities available to students are exogenously given is with loss. Under monopoly, the separating equilibrium envisaged in Spence (1973) does not arise—wasteful effort is eliminated, while efficient job allocation is preserved. Second, competion induces schools to employ two strategies to lure students away from each other: cutting fees and providing more informative signals about abilities to enable high-ability candidates to separate themselves out, which  reviving costly effort. This result contrasts with the usual argument that competition enhances social efficiency. We further show that the role of competition is partially restored under market frictions, such as credit constraints. In this sense, policy prescriptions commonly associated with the signaling critique—skepticism toward subsidies and student loan programs, and efforts to curb the expansion of credentialing intermediaries—may come into tension with one another.

Technical difficulty:  The multiplicity of perfect Bayesian/ sequential equilibria in a typical signaling game propagates upstream, leading to a vast set of potential school policies, with many supported by unreasonable off-path beliefs. We appeal to forward-induction reasoning and invoke an equilibrium refinement that extends the D1 criterion of Cho and Kreps (1987) to our setting. This refinement allows us to discipline off-path play sufficiently to obtain unique equilibrium predictions under monopoly and economically meaningful equilibrium predictions under competition.

3. Motivated Misperception   [Slides]

I propose a model of expectation management to study how the directionality of misperception is perpetuated in a long-term relationship. Leveraging the machinery of misspecified learning in a principal-agent framework, I find that the profitable direction of manipulation for the principal is determined by the output function, namely, how multiple factors contribute to the project output. The central mechanism is, to sustain excessive effort, the principal should downplay factors that affect project output independent of the agent’s effort, thus making the agent misattribute the outcome more to her effort choices and inflating the perceived agency. It provides a novel approach to incentivize effort complementary to the usual monetary or informational incentives, and can be applied to a wide range of interactions, such as mentor-mentee, parent-child, self-manipulation, and emotional abuse in professional or intimate relationships.

4. Multilateral War of Attrition with Majority Rule  with Hülya Eraslan and Kirill S. Evdokimov   [Slides]

In multilateral bargaining with majority rule, a reputation for being stubborn can harm the player's chances of being included in the first place. In this paper, we study a three-player concession game in which only two players, including a veto player (the chair), are needed to reach an agreement on how to split a unit of surplus. In contrast to the bilateral case, delay equilibria are less likely to occur; however, conditional on delay, the chair may be worse off when bargaining with two players instead of only one player.

5. Online Anonymity under Stigmatization: Evidence from Geotagging Mandate in China with Yiqian Wang and Ziang Xiao; new draft coming soon!

Anonymity protects the freedom of speech whereas burgeoning misinformation from anonymous sources urge platforms to practice  identity transparency.  Leveraging recent Chinese social media regulations and text analysis advancements, we find that reduced anonymity substantially decreases politically marginalized groups' representation. We motivate our empirical results using a theoretical framework and investigate how the intergroup gap in media biases affects participation in social media. Our findings highlight that, under media biases, identity transparency intensifies political polarization.

6. Mortgage Refinance Under Present Bias: Evidence from the Use of Lender Credit  with Yunbo Liu

Households often incur substantial losses by refinancing their mortgages too early or failing to refinance altogether. We provide a unified framework that accounts for both types of mistakes. In the model, households with present bias decide both when to refinance and whether to use lender credits to reduce upfront refinancing costs. We characterize a closed-form solution for the refinancing threshold under present bias and show that present-biased households are more sensitive to changes in upfront costs. This combination of present bias and upfront cost choice gives rise to both procrastination (for households with sufficiently high present bias) and precrastination (for households with a moderate level of present bias that use lender credits). We then test our model using U.S. mortgage data from 2010-2019 and identify lender credits usage through FHA loan features. We find that borrowers using lender credits refinance significantly earlier during periods when interest rate decreases and achieve less rate savings compared to those who pay closing costs out of pocket.

Papers in Refereed Conference Proceedings

• Best vs. All: Equity and Accuracy of Standardized Test Score Reporting with Sampath Kannan, Aaron Roth, and Rakesh Vohra

FAccT'22: 2022 ACM Conference on Fairness, Accountability, and Transparency,  574–586   [PDF   Slides   Poster]

This paper emphasizes a source of unfairness that arises from the fact that only some applicants have the resources to take standardized tests multiple times. We find that, compared to super-scoring, requiring that all scores be reported results in superior equilibrium outcomes in admission accuracy and equity across populations. This is the case even though the more highly resourced students can—at their option—either report a more accurate signal of their type, or pool with the lower resourced population under this policy. The study represents an unusual situation where the goals of accuracy and equity are in alignment.

Other Publications

• SIGecom Winter Meeting 2022 Highlights (with Emily Diana and Georgy Noarov), ACM SIGecom Exchanges, 20(1), Nov 2022.   [PDF]

Working in Progress

1. Standoffs with Intertwined Resolve and Public Information with Hülya Eraslan and Kirill S. Evdokimov 

2. Signaling and Observability

3. Addiction under Present Bias

Dormant Projects

• Gaming in Zero-Knowledge Proof System

How to incorporate the notion of Zero-Knowledge Proof (ZKP) in cryptography into economics? Loosely speaking, ZKP requires the interactions between a prover and a verifier that yield no extra information other than the validation of the claim. It guarantees data privacy and information safety in a decentralized economic system, especially with the technology of blockchain.  

• Screening with Delegated Learning

A principal needs to delegate information collection to agents, and agents differ in their privately known ability of information acquisition. What is the optimal approval mechanism for the principal to make the right decision? Is it better to let the principal specify the information collection regime (e.g., when to report)  or allow the agent, who has the information advantage, to have the discretion?

Grants

I am currently partially supported by a CUHK startup grant:  Improvement on Competitiveness in Hiring New Faculties Funding Scheme

I was previously supported by the Azrieli Foundation as an international postdoctoral fellow in the academic year 2024-2025.

Services

I've been a referee for:  Economics Letters, Games and Economic Behavior, International Economic Review, International Journal of Game Theory, Journal of Industrial Economics, Journal of Mathematical Economics, and Review of Economic Design, Social Choice and Welfare.