Research
Research
Directional AI Advice: Experimental Evidence from Healthcare (joint with Yuyu Chen, Hongbin Li, Lingsheng Meng, and Qingxu Yang)
Generative AI is fast becoming the first place people turn for expert advice. The advice it provides can be directional rather than neutral, shaped in part by the choices of its designers and regulators. When clients consult AI before meeting an expert, they carry this directional advice into a relationship that once rested on the expert's judgment alone. We study its consequences in healthcare through a large-scale preregistered field experiment at a Chinese hospital, where we randomize patients' access to an AI chatbot before their outpatient visit. Examination of the conversation logs shows that the chatbot routinely cautions against the use of medications, especially Traditional Chinese Medicine and antibiotics, while issuing clean recommendations for diagnostic testing, consistent with the liability-driven guardrails encoded in AI training. This directionality propagates into clinical practice. Prescription rates decline among treated patients while diagnostic testing increases, and these effects are more pronounced among physicians who are receptive to patient input and those with more intensive prescribing styles. Beyond shifting healthcare utilization, survey results show that AI access reduces patient compliance and satisfaction, shifting the balance of authority between patients and physicians.
Choosing What's Familiar: Information Inequality in College Major Choices
revise & resubmit at the Economic Journal
I study disparities in college major choices across socioeconomic backgrounds and their implications for intergenerational mobility. Using administrative data from China's centralized college application system, I document that students from low socioeconomic backgrounds are 3.13 percentage points (21.4%) more likely to choose majors that are familiar to them from high school, even though these majors have lower labor-market returns on average. I present evidence from a survey experiment and from spillovers among high school classmates that is consistent with unequal access to information contributing to this disparity. I also discuss alternative explanations.
Misaligned Calendars, Missed Internships: Labor Market Costs of Coordination Frictions in Higher Education
Administrative choices in higher education can have unintended consequences for students' careers. This paper examines the choice of academic calendar and the coordination frictions it creates when universities deviate from the prevailing norm. Using quasi-experimental variation in academic calendars across Ohio universities and data from LinkedIn profiles, I find that switching to the more prevalent semester system increases internship participation by 16%, with gains concentrated in labor markets where the semester calendar dominates. These additional internships improve students' post-college outcomes, including higher starting salaries and better job quality. The findings highlight the labor market consequences of seemingly internal institutional choices.
Firm Pricing Strategy and Consumer Dishonesty: Evidence from the Bike-Sharing Industry (joint with Guangyu Cao, Wei Dai, and Juanjuan Meng) revise & resubmit at Management Science
This study examines the impact of firms' pricing policies on consumer dishonesty in the context of the sharing economy. We analyze over two million trip records from a bike-sharing company and develop measures of customers' dishonest behaviors at the trip level. Leveraging price variations from randomly distributed coupons and an exogenous price shock induced by a promotional campaign, we find that higher prices significantly increase the likelihood of dishonest behavior. Heterogeneity analysis reveals that this effect is more pronounced among male, non-student, and new users, as well as during nighttime hours. We also find suggestive evidence that the promotional campaign triggers reciprocal behavior from consumers, who demonstrate reduced dishonesty after experiencing the promotion, with prices held constant.
Family Spillover Effects of Marginal Diagnoses: The Case of ADHD (joint with Petra Persson and Maya Rossin-Slater)
American Economic Journal: Applied Economics, 17(2): 225-256, April 2025.
The health care system uses patient family medical history in many settings, and this practice is widely believed to improve the efficiency of health care allocation. This paper provides a counterpoint by documenting that reliance on hereditary information can amplify the misallocation of low-value care. We study Attention Deficit Hyperactivity Disorder, and show that reliance on family medical history generates a "snowball effect"—the propagation of an original marginal diagnosis to a patient's relatives. This snowball effect raises the private and social costs of low-value care.
Heuristics in Self-Evaluation: Evidence from the Centralized College Admission System in China (joint with Hongbin Li)
Review of Economics and Statistics, 07 (6): 1724–1733, November 2025.
Using administrative data on the Chinese National College Entrance Examination, we study how left-digit bias affects college applications. We find strong discontinuities in students' admission outcomes at 10-point thresholds. Students with scores just below multiples of 10 make more conservative college application choices that place them into less selective colleges and majors. In contrast, students who score at or just above multiples of 10 aim and achieve higher but are at greater risk of overshooting. The discontinuity reveals that, despite the educational and labor market consequences, students' self-evaluation based on exam scores is subject to information processing heuristics.