Research
Publications
2026
- with Wenbo ZouJournal of Economic Behavior & Organization, Feb 2026
Generative artificial intelligence (AI) technologies are widely regarded as having the potential to enhance human creativity. In a pre-registered lab experiment, we assigned 302 university students tasks involving the creation of ideas for generating creative pictures, with a randomly selected subset granted access to ChatGPT. The results indicate that access to ChatGPT improves participants’ creative performance. The treatment effect follows an inverted-U relationship with respect to participants’ baseline creative potential, leading to a reduction in the inequality of their creative performances. Comparing the creative performance of human-AI teams to ChatGPT standalone, we find no evidence of complementarity. However, by using measures construed from the human-AI conversations to explain the variations in the observed treatment effects, we find suggestive evidence that collaborative involvement of ChatGPT is associated with more positive effects. Finally, we also observe significant treatment effects on participants’ opinions about AI, their ratings of subjective experiences during the tasks, and their self-reported creative and problem-solving abilities.
2025
- with Chiara Lo Prete , Rong Rong , Anthony Kwasnica, Cody Hohl, and Jiaxing WuThe Electricity Journal, Dec 2025
The integration of renewable energy technologies into electricity markets presents an important challenge. This paper examines the performance of alternative short-term electricity market designs for managing the uncertainty associated with wind power generation. The markets vary by their opportunity for wind power plants to respond to changing wind forecasts prior to final supply. In the two-stage market, subjects receive information about their forecasted wind availability in the day-ahead stage, and choose a quantity offer to maximize their profit for the day before the actual wind output is realized. In the multi-stage market, subjects receive an additional signal on their forecasted wind availability in the intraday stage, and may revise their quantity offer to maximize their profit before the actual wind output is realized. We develop a model of rational profit-maximizing behavior in sequential markets, derive theoretical predictions for individual behavior and market outcomes, and design laboratory experiments to test the model’s predictions. Subject decisions in the day-ahead stage of the multi-stage market rely on the prior probability of the wind state and mostly ignore the potential value of the upcoming intraday signal. This is consistent with findings of the experimental economic literature showing that decision makers may fail to make rational decisions when faced with uncertainty. Our results also suggest that the multi-stage market leads to lower uplift payments and higher revenues for the wind power producers. These efficiency gains mainly stem from the fact that subjects correctly update their quantity offers after receiving new information on wind availability in the intraday stage.
- with Jiaxing Wu , Chiara Lo Prete , Rong Rong , and Burçin ÜnelEnergy Economics, Oct 2025
We compare generation capacity investment and unserved energy in experimental electricity markets under three designs: (1) an energy-only market serving as baseline, (2) an energy-only market with scarcity pricing where firms earn revenues from the provision of an additional product, reserves, and (3) a capacity market that rewards capacity investment and is followed by an energy-only market. Subjects in the experiments act as firms, choosing generation capacity and competing to meet electricity demand in the market. Based on simulation results from complementarity-based equilibrium models, our predictions suggest that an energy-only market incentivizes the lowest capacity investment, which is insufficient to satisfy peak electricity demand. Further, we expect the capacity market to promote more investment than the energy-only market with scarcity pricing, avoiding unserved energy across all demand scenarios. In contrast, the energy-only market with scarcity pricing experiences unserved energy during peak demand periods. In line with the predictions, we find that the experimental baseline leads to insufficient capacity investment to meet peak demand. The alternative designs encourage additional capacity investment, with firms in the capacity market generally investing in more generation capacity. Yet, both alternative designs are unable to satisfy peak electricity demand because their additional capacity falls short of expected levels. Further, unserved energy is reduced, but results are not statistically different under the alternative designs.
- with Yohanes E. Riyanto and Siqiang YangJournal of Economic Behavior & Organization, Aug 2025
This paper presents a laboratory experiment investigating contest design and contestant behavior in settings where participants have different cost functions and regions of advantage, a topic that is relatively unexplored. Utilizing lab experiments based on Siegel’s (2009, 2010) models, we discover that bidding strategies and payoffs largely follow theoretical predictions. The weakest contestants tend to over-participate due to their cost advantage in low-bid regions. Adding a prize or increasing the prize value benefits stronger participants and increases the designer’s revenue. While decreasing the number of contestants also favors stronger participants, the designer’s revenue decrease in response.
- with Xiangyi ChenIEEE Transactions on Automatic Control, May 2025
We present an adaptive form of the simultaneous long-short (SLS) trading strategy, which employs feedback control principles for stock trading. Prior work established the “robust positive expectation” (RPE) theorem for the SLS strategy, asserting positive expectations of trading profits for stock prices following geometric Brownian motions with unknown parameters. We extend the RPE property to adaptive SLS strategies, which guarantee strictly positive expected profits even in round-trip price scenarios where traditional SLS strategies fail.
2024
- with Xueying LyuApplied Economics, Dec 2024
Weather and climate changes have been found to affect various socioeconomic outcomes, but the mechanisms that underlie these effects are not fully understood. This article examines the effect of temperature on risk-taking behaviour using daily weather and Powerball sales data in the U.S. The results show that a 1 degree Fahrenheit increase in daily maximum temperature, on average, leads to a 0.30% increase in the Powerball sales in the county. We also find that individuals tend to opt for riskier options on higher-temperature days. Alternative explanations for these effects, such as avoidance behaviour and the income effect, are ruled out as possible drivers of these effects. Our article provides new insights into a potential yet undiscovered channel through which temperature, or more broadly, weather and climate changes can influence socioeconomic outcomes. The findings also have important implications for policymakers concerned about weather and climate changes and their effects on human behaviour.
2021
- Feng ZhuJournal of Mathematical Economics, Aug 2021
I analyze the optimal favoritism in a complete-information all-pay contest with two players, whose costs of effort are weakly convex. The contest designer could favor or harm some contestants using one of two instruments: head starts and handicaps. I find that any given player’s effort distribution is ranked in the sense of first-order stochastic dominance according to how (ex post) symmetric the players are in terms of competitiveness. Consequently, as long as the designer values effort from both contestants, ‘‘leveling the playing field’’ is optimal regardless of which instrument is used.