Qing Gong
Interests: Public Economics, Health Economics

Job Market Paper

Physician Learning and Treatment Choices: Evidence from Brain Aneurysms

Physicians often choose among alternative treatment options based on their beliefs over the treatment effectiveness and their skills in delivering the treatment. I examine how two kinds of physician learning jointly shape their treatment choices: Bayesian learning that updates beliefs about treatment-patient match values and learning by doing that improves surgical skills. Using case-level data on the history of brain aneurysm treatments by over 200 physicians, I find that both kinds of learning are present and that physicians are forward-looking. In light of these empirical patterns, I develop and estimate a dynamic structural model of physician learning and treatment choices for heterogeneous patients. I then disentangle the impacts of the two kinds of learning and explore to what extent forward-looking physicians deviate from myopic best choices. Physicians are more than twice as likely to experiment on unhealthy patients than healthier ones, which hurts short-term outcomes but improves overall treatment success rates by 13-17%. I also evaluate the impacts of several alternative payment schedules. Uniform payments across treatments facilitate the adoption of the new treatment while outcome-contingent payments have heterogeneous effects across physicians. The heterogeneity highlights the coexistence of two opposing effects: the incentive to exploit the myopic best option and the incentive to experiment with less familiar options due to the increased return from learning.


Detecting Potential Overbilling in Medicare Reimbursement via Hours Worked (with Hanming Fang)

American Economic Review, 2017, 107(2), pp. 562-591

We propose a novel and easy-to-implement approach to detect potential overbilling based on the hours worked implied by the service codes physicians submit to Medicare. Using the Medicare Part B Fee-for-Service (FFS) Physician Utilization and Payment Data in 2012 and 2013, we construct estimates of physicians’ hours spent on Medicare beneficiaries. We find that about 2,300 physicians, representing about 3% of those with 20 or more hours of Medicare Part B FFS services, have billed Medicare over 100 hours per week. We consider these implausibly long hours and find suggestive evidence that the coding patterns of the long-hour physicians respond to financial incentives: within code clusters with different levels of intensity for the same service, they tend to submit more high-intensity codes, especially when the marginal revenue gain from doing so is higher.

Polluting Thy Neighbor: Unintended Consequences of China’s Pollution Reduction Mandates (with Hongbin Cai and Yuyu Chen)

Journal of Environmental Economics and Management, 2016, 76(3), pp. 86-104

We study how China’s pollution reduction mandates in 2001 triggered unanticipated responses from its provinces. We apply the difference-in-differences-in-differences (DDD) method to a unique dataset on industry-level activities in counties along 24 major rivers in China from 1998 to 2008. We find that the most downstream county of a province has up to 20% more water-polluting activities than otherwise identical counties starting from 2001. Moreover, we find that the enforcement of pollution fee collection is more lenient in the most downstream counties, and that private firms contribute more to the downstream effect of pollution than state-owned and foreign firms. These findings suggest that provinces respond to the pollution reduction mandates by shifting enforcement efforts away from the most downstream county, thereby meeting the mandate requirements but polluting the downstream neighbor province.

Ex Post Moral Hazard in Automobile Insurance Markets with Experience Rating

Accounting for unreported accidents due to ex post moral hazard is important for studying asymmetric information in insurance markets. In this paper, I study ex post moral hazard in an automobile insurance market with experience rating. I develop a dynamic model in which policyholders with private information about their risk types choose accident prevention efforts (ex ante moral hazard) and make claim filing decisions when accidents happen (ex post moral hazard). I then estimate the model using a detailed policy-level panel dataset from China. I find that policyholders do not report 24% of all accidents, which account for about 5% of total monetary losses. The degree of ex post moral hazard varies by experience rating: policyholders with the best rating hide 40% of all accidents. Finally, I use counterfactual experiments to evaluate the welfare implications. I find that experience rating improves policyholder welfare, mainly by inducing higher preventive efforts and reducing accidents. When ex post moral hazard is restricted and policyholders are forced to report all accidents, the benefit from increased accident prevention efforts barely outweighs policyholders’ loss from increased premiums.

Retention Effects of Employee Stock Options: Evidence from Bunching at Vesting Dates (with James Liang, Hong Zhang, and Li-An Zhou)

Whether employee stock options bring the firm enough benefit to justify the cost is at the center of much debate. We study the retention effect of stock options to reconcile their popularity and high granting costs. We use the bunching design to address the endogeneity problem and identify the causal effects of options on retention. Using a novel administrative panel dataset at the individual employee level, we find option owners delay quitting until options vest to minimize the opportunity cost of quitting. The bunching at the vesting dates is significant and sizable: the quitting of option owners more than doubles shortly after options vest. We find the bunching effect is robust to alternative explanations, is stronger when option values are high, and is unique to voluntary quitters but absent among owners who leave involuntarily. We also verify that the retained option owners have superior performance, and conduct a simple benefit-cost analysis with the baseline estimates. We find that the retention benefits contribute greatly to the total benefit of options, which exceeds the granting cost by 95-275%. Accounting for the retention effects of options can avoid underestimating the benefit-cost ratio, thereby helping to reconcile the popularity of options and the high granting costs.

Learning by Doing and Endogenous Patient-Physician Matching: Evidence from Cardiac Surgeons in New York State (in progress, with Hanming Fang)

We examine the dynamic patient allocation problem of hospitals whose goals are to improve patient outcomes while training new physicians. We develop a dynamic model of endogenous patient-physician matching with heterogeneous patient conditions. Hospitals in our model try to strike the optimal balance between the learning by doing of junior physicians and the superior performance by experienced physicians. We conduct the empirical analyses on an administrative panel data of cardiac surgeons in New York State.

Teaching Experience

Public Finance, Teaching Assistant for Professor Hanming Fang (Fall 2016)
Industrial Organization, Teaching Assistant for Professor Anne Duchene (Spring 2016)
Microeconomics, Recitation Instructor for Professor Anne Duchene (Fall 2015)
Econometrics, Recitation Instructor for Professor Frank Schorfheide (Spring 2014)
Econometrics (graduate), Teaching Assistant for Professors Frank Schorfheide and Xu Cheng (Fall 2013)


Conference Presentations

International Health Economics Association (iHEA) World Congress

Econometric Society North American Summer Meeting
American Society of Health Economists (ASHEcon) Biennial Meeting
Beyster Symposium for Employee Ownership and Profit Sharing


Hanming Fang (advisor)
Department of Economics
University of Pennsylvania

Ashley Swanson
Health Care Management
Wharton School
University of Pennsylvania

Petra Todd
Department of Economics
University of Pennsylvania


I am on the job market and will be available for interviews during the AEA meetings in Philadelphia from 1/5 to 1/7.