Health and education are two important issues in developing economies. Field and natural experiments provided me with great opportunities for identifying the effects of health insurance and incentive on doctors' prescribing behaviors and the peer influences among students.

The first chapter examines whether doctors write more expensive prescriptions for insured patients and if so, why. I conducted a randomized audit experiment using undercover visits to Chinese hospitals. The results show that prescriptions for insured patients are 43% more expensive than those for uninsured patients when doctors expect to obtain a proportion of their patients' drug expenditures. The differences in prescriptions are largely explained by a differential agency problem hypothesis that doctors act of self-interest and prescribe more unnecessary or expensive drugs to insured patients, rather than by a considerate doctor hypothesis that doctors consider the trade-off between drug efficacy and patients' ability to pay.

The second chapter studies peer effects in a natural experiment generated by an unusual change in college admission policy at a prestigious Chinese university. The change in admission policy brought a large number of low-scoring students into several academic departments which only admitted high-scoring students in usual years. Exploiting the large variations in peer characteristics and the strong interactions among peer groups, the analysis finds that specially admitted low-scoring students significantly reduced the performance of regular students in standardized English tests. This detrimental effect of specially admitted students is concentrated among students with English ability below average.

Most research on peer effects in pre-college education focuses on the class or school level and assumes that students are influenced by class- or school-level averages. The third chapter examines peer effects within small groups inside classrooms by exploiting an experiment with random seat assignments inside Chinese classrooms. We define peers as either deskmates or neighboring students who sit at desks directly in front of or behind the students. The results suggest different patterns of impacts from deskmate and other neighboring students: female deskmates improve test scores for both boys and girls while the proportion of females among other neighboring students has strong positive effects on girls but no impact on boys. Overall, the results suggest that organizing students into small groups of homogeneous gender inside a classroom—a small-scale version of single-sex education—can significantly improve test scores.