The Crowding-out Effect of Formal Insurance on Informal Risk Sharing: An Experimental Study, Games and Economic Behavior, 2014. (joint with Wanchuan Lin and Juanjuan Meng) (SSRN)
Optimal Goal-setting with Present Bias: Theory and Experiment (Job Market Paper) (pdf)
This paper develops a theory of optimal goal bracketing with present bias and tests the predictions of the model in an online real-effort experiment. In our multi-selves model, the long-run self who faces a self-control problem caused by present bias sets goals as reference points to motivate the short-run selves. Narrow bracketing is defined as setting one goal for each short-run self and broad goal-setting is defined as setting one broad goal for several short-run selves to jointly achieve. The main trade-off is between commitment and flexibility. Narrow goal-setting provides more commitment but broad goal-setting is better in terms of flexibility. In line with the predictions of the model, the results of our online experiment show that 1) "Nudging" subjects to set narrow goals facilitates self-control. 2) The assumption that goals work as reference points is supported by empirical evidence. 3) Subjects who are more present-biased benefit more from goal-setting. 4) Broad goal-setting does not work. It also causes procrastination: subjects exert more efforts on the later date under broad goal-setting. 4) Surprisingly, but consistent with the model, we find that narrow goal-setting always outperforms broad goal-setting from the long-run self's perspective regardless of the degree of present bias. However, the gap between the two bracketing methods shrinks as present bias decreases, indicating that the trade-off between commitment and flexibility does exist.
Responsibility Shifting through Delegation: Evidence from China's One Child Policy (joint with Yi Han) (pdf)
There is a growing body of experimental evidence indicating that delegation can foster the shifting of responsibility for unpopular actions from a principal to an agent. Using the well-known episode of the one-child policy in China (OCP), we provide first field evidence for responsibility shifting through delegation.
Income Inequality and Political Polarization
In last four decades, party polarization and income inequality have experienced rising comovement. In this paper, I build a link between the two. I consider a Spatial model of redistribution with heterogeneity in voters' wealth. Two ex ante identical candidates compete for a public office by proposing redistributive taxes. Voters care about taxes and candidates' valence. Valence is endogenously determined by campaign spending financed by voter's individual contributions. Rich voters have the strongest incentive to contribute, because they are mostly affected by taxes. In equilibrium, policy polarization arises when income inequality level is high enough. In this case, median voter's ideal tax is defeated by a lower tax that can attract enough campaign contributions from the rich. In contrast, when inequality level is low, median voter theorem holds. With low inequality, the ideal tax rates of the rich and the rich are not that different. The candidate who deviates from the median voter's ideal tax cannot collect enough funding from the rich because their incentives to change the election result are too weak.
RESEARCH IN PROGRESS
Correspondence Bias (joint with Yi Han and George Loewenstein)(Draft Coming Soon)
People often draw inferences about other's enduring characteristics from behaviors that can be entirely explained by the environment/incentives. This bias is simple yet challenging to identify in an online experiment.
Performance Self-assessment and Matching Outcomes in a High-stakes Environment: Evidence from Chinese Entrance Exams (joint with Stephanie Wang)
It has been well documented that human beings have tendency to be overconfident. However, we find that in a high-stakes environment, there is no under- or over-confidence in the overall sample. The results suggest that people enjoy the consumption utility of optimism to the extent that the mistakes caused by it are not too costly (Brunnermeier and Parker, 2005).
Major Choice in the Chinese Education System: a Machine Learning Approach (joint with Stephanie Wang)
Students in China need to choose their academic track (STEM vs Non-STEM) in the freshman year of high school. Using a unique administrative data from 4 large high schools over two provinces from 2014-2016, we find that women are more responsive to relative strengths than men, which leads more women who have good grades in science choose the Non-STEM track. A counterfactual analysis using machine learning reveals that being “too responsive” decrease the chance that a female student will be admitted by elite universities.