• ​Health Care Seeking Behavior and Hospital Choices: A Case Study in Rural China

​

Despite the growth of income and insurance coverage in rural China, the pattern of health facilities choices remained stable for residents, with mostly choosing village clinics. Exploiting the introduction of the New Cooperative Medical Scheme (NCMS), a subsidized public insurance scheme for rural residents, this paper examines the determinants of health care demand in rural China. I estimate a random coefficient logit (RCL) model to explore how such choices are affected by out-of-pocket cost (OOP), the quality of health facilities, and distance to the health facilities, using the China Health and Nutrition Survey (CHNS) data from 1989 to 2011. The RCL model is more flexible as it allows both observed and unobserved heterogeneity in consumer’s preference, which plays a major role in this setting. The results show that OOP and distance have negative impact on the health care decision, while hospital quality has a significant positive effect on patient’s choice. The NCMS insurance, however, has an insignificant impact on the choice. In addition, my results show that there exists significant variation in preferences for both distance and hospital quality across different patients, which results in an interesting substitution pattern. When a nearby or premier healthcare facility increases its price, patients who substitute away will be those who have high marginal utility on quality or distance; and they substitute disproportionately towards other good-quality or short-distance hospitals. 

 

 

• Family Food Insecurity and the Mental Health of Young Children: Can Access to SNAP Break the Vicious Cycle? (with Pamela J. Surkan, Laura Pryor, Maria Melchior and Mauricio Avendano)

​

Mental health problems occur disproportionately among those living in poverty, especially if they are food insecure, and frequently begin in early life, highlighting the need for early prevention. This project examines long-term effects of childhood food insecurity and family SNAP participation on young adult mental health and addictive behaviors. Using data from the Panel Study of Income Dynamics (PSID) Child Development Supplement, we first employ multivariable logistic regression analyses to test associations between childhood food insecurity and psychological distress and addictive behaviors in young adulthood. We then propose an instrumental variable (IV) approach to investigate the effects of family SNAP participation on children’s later psychological distress and addictive behaviors.  We find significant associations between being in the Moderate and High food insecurity groups during childhood with high levels of psychological distress in young adulthood: Adj. ORs (95% CI’s=1.72 (1.59-1.85) and 2.06 (1.81-2.33), respectively. Primary caregiver’s high psychological distress and high food insecurity placed children at greater risk for later high psychological distress. Access to SNAP attenuated this risk. Results from the IV analysis suggest that family SNAP participation during childhood has significantly reduced psychological distress and addictive behaviors in young adulthood. Our findings can guide policies aimed at halting the vicious cycle between low socioeconomic attainment and and mental illness. 

​

​

​

Work in Progress

​

• Pre-policy Trends in the Event Study Design

​

A broad empirical literature, ranging from public to labor to health economics, uses the event study design. In these studies, it is common to report cluster-robust standard errors (CRSE) for the pre-policy parallel trend test to control for the fact that error terms within groups are typically correlated. However, the standard errors, after adjustment by CRSE, generally exhibit a decreasing trend over the pre-policy period, relatively large in the beginning and getting smaller close to the period of policy implementation. Conversely, standard errors have similar values in both pure OLS and robust OLS cases. To shed light on this recurrent issue, I first use NPDUIS data on prescription claims for seniors in Canada and employ a graphical descriptive approach to show the dynamics of the difference in outcomes between the treatment and control groups in pre-policy trends, after clustering the data into groups. The literature suggests that CRSE can lead to biased inference in the case of few clusters (Wooldridge, 2003; Bertrand et al., 2004) or few treated clusters (MacKinnon and Webb, 2017). I summarize a series of examples from the existing studies, and run Monte Carlo simulation with different parameter settings to explore whether the observed pattern is driven by few (treated) clusters, a problem inherent in CRSE. I then compare standard errors based on different approaches that provide consistent inference (CRSE, OLS, robust OLS, bootstrap). My final goal is to develop a checklist that researchers can use to verify the presence of the issue in their setting and use the most appropriate inference approach, the one providing reliable estimates.

​

• Intergenerational Support and Health Outcomes among Older Parents in China

This study explores the relationship between intergenerational support and well-being of older parents in China. Using China Health and Retirement Longitudinal Study (CHARLS) survey, I estimate the effects of functional support, including emotional and financial support, on the physical and psychological well-being of older Chinese parents. I adopt the instrumental variables approach to control for the endogenous problem that health conditions of parents would conversely affect the children’s support of their parents. My results suggest that financial support (money transfer) has significantly improved both physical and psychological well-being of parents, while the emotional care (visiting frequency) only has significant impact on parental physical health outcomes (activities of daily living, ADLs).