Publications
[1] “Fear on the Plank? Virtual but more Real!” with Wei Huang, Paul McIntosh, Irwyn Shepherd and Xiaojian Zhao, Experimental Economics, forthcoming.
This paper uses virtual reality (VR) technology to study willingness to pay for safety under emotionally salient conditions in a controlled economic experiment. In our experiment, participants virtually stand on the roof of a hundred-story building. They need to build a plank bridge to reach a monetary reward suspended from a drone on the other side, and then walk through the plank to grab the reward, facing a probability of falling virtually. The planks vary in width and price, with wider planks offering greater security at a higher cost. We compare choices in VR with paper-based elicitation formats and interpret the differences as reflecting context-dependent valuations of safety. This approach extends conventional risk elicitation by measuring willingness to pay for safety in immersive simulations of life-threatening situations.
[2] “Strategic Information Suppression in Borrowing and Pre-Lending Cognition: Theory and Evidence” with Xiaojian Zhao, Games (Special Issue on Economics of Motivated Beliefs), 2023, 14(3), 43.
Editor’s Choice Articles in 2023
Selected Works
[1] “Managing Screen Time: Feedback and (Soft-)Commitment” with Erte Xiao, Jane Zhang and Xiaojian Zhao
Managing screen time appears to be a pressing issue in our digital age. Previous studies suggest that sophistication about present bias is associated with commitment take-up. In a field experiment, we provide participants with information feedback about their over- and underestimation of screen time usage, with the aim of improving sophistication. We investigate the causal effect of this feedback on the adoption of popular soft commitment apps designed to control screen time usage. We observe no evidence that feedback influences commitment adoption. Our data suggest that, instead, feedback has a direct effect on screen time usage for individuals with self-control problems who had underestimated their usage. On the other hand, those who overestimated their usage show no compensatory increase. We also provide a procrastination model with diversely naive individuals, largely accounting for our experimental observations.
[2] “When Bad News Motivates: A Field Experiment in Juvenile Rehabilitation” with Wei Huang
We study selective performance feedback in a field experiment with adolescents in residential juvenile rehabilitation. Students were randomly assigned to see all of their monthly grades, only their Excellent grades, or only their Fair grades, and were asked to enter a self-rating for each withheld item without being told the disclosure rule. Written ratings on blanks cluster near the middle of the three-point scale under both selective policies, although the grades those policies conceal differ sharply. Consequently, disclosing only excellence places the largest optimistic reporting errors on the items with the most room to improve. The exact prediction that ratings are rule-invariant is rejected on the grade concealed by both policies, but the estimated difference is small relative to the difference in concealed grades. Randomised policy-level comparisons of subsequent grades are imprecise and do not reject equality. An exploratory association between current silence and lower next-month grades is concentrated in later months, when the grades that determine silence may themselves reflect earlier feedback, and weakens under calendar-time controls. The experiment therefore establishes how these disclosure protocols distribute errors in completed records; whether that distribution changes behaviour remains open.
[3] “Managing Digital Consumption” with Juanjuan Meng, Erte Xiao, and Xiaojian Zhao
The rise of social media has intensified excessive digital consumption, as lapses in self-control and limited self-awareness jointly sustain overuse. Using evidence from a randomized experiment with Douyin users, this paper examines whether feedback on individuals’ own usage helps them learn about their self-control and (mis)beliefs about it. We find that individuals improve in predicting their future usage and reduce subsequent usage after receiving feedback, while merely asking them to predict future use has no effect, indicating that the feedback effect is not driven by goal setting. Importantly, the asymmetric feedback effect persists: both underestimators and overestimators update their beliefs about future usage, but only underestimators reduce subsequent usage.
[4] “The Road to a Distant Goal: Self-Control, Self-Awareness, and Dynamic Effort Incentives” with Xiaojian Zhao
Can an inaccurate forecast of future self-control improve the pursuit of a distant goal? The answer depends on how the task is completed. For agents who weakly overestimate their future self-control, we compare a task completed by one irreversible action with effort accumulated across dates. In the completion problem, an intermediate forecast can induce immediate completion while both sophistication and full naivete induce delay. We characterize the forecast interval, its welfare gains, and the losses from missing it. The mechanism requires at least three continuation dates: a more optimistic forecast of one future self can make an earlier future self wait longer. In the cumulative-effort problem, under strict present bias and concave reference continuation surplus, every admissible forecast leaves current effort below its efficient level at every nonterminal date. The best forecast for one action therefore remains second best. These results separate the incentive value of self-knowledge from its accuracy and identify when forecasts can substitute for immediate incentives.
[5] “Agreement and Unawareness under Misinterpreted Evidence” with Xiaojian Zhao
Agreement and no-speculation can survive systematic misinterpretation of evidence. On a finite state space, actual signals and their interpreted meanings form two matched partitions. Non-deludedness requires their encodings to agree pointwise; mass balance requires agreement in distribution. Mass balance restores partitional common belief and an Aumann-style agreement theorem despite individual failures of truth and introspection. The ordinary informationpartition theorem is a special case. Once the interpretation rules admit a full-support balancing measure, agreement and the stated no-speculation conclusion are robust to changes in the shared conditioning measure. More generally, a sharp bound controls disagreement by the balance defect relative to the probability of the common-belief region. We also study symmetric higher-order ignorance: under partition range, global KU requires at least five states for nontriviality, while an inclusion-antichain range rules out Boolean closure of awareness. A contracting game links these interpretation errors to strict speculative trade and identifies a second-order belief obstruction to commonly believed execution. All constructions use ordinary events and probability measures on one state space.