Research
Research Interests
Causal Inference, Applied Microeconomics, Applied Econometrics, Development Economics, Survey Methodology
Working Papers
Social Media Usage and the Level of Depressive Symptoms in the United States (Job Market Paper)
The use of social media can either decrease the level of depressive symptoms by providing support or increase the level of depressive symptoms by putting social pressure on users. This paper leverages fixed effects and instrumental variable models to find that using social media (Twitter and Facebook) decreases the level of depressive symptoms on average. This result explains why social media usage in the US grows steadily even though most studies find that more usage correlates with higher levels of depressive symptoms. Three explanations reconcile the negative causal effect and positive correlation: confounder, social media improvement, and heterogeneity.
Selection Bias Reduction by Matching and Weighting Estimators in Social Media Data Collection
Social media can be a powerful tool for surveys and data collection. Whether rigorous academic surveys and data collection can be done through social media remains to be a question. This paper focuses on the potential selection bias issue in social media survey. I examine performances of matching estimators (matching on propensity scores or distances), weighting estimators, and covariate balancing propensity score estimators in eliminating bias between Twitter users and non-users. I find that matching on distance performs the best if computation capability allows, and adding machine learning in model selection in first stage improves estimators’ performances. All examined methods reduce selection bias between users and non-users.
Publications
Riehm, Kira E., Calliope Holingue, Luther G. Kalb, Daniel Bennett, Arie Kapteyn, Qin Jiang, Cindy Veldhuis et al. "Associations between media exposure and mental distress among US adults at the beginning of the COVID-19 pandemic." American Journal of Preventive Medicine (2020).
Work in Progress
School Starting Age Effect and Noncompliance behavior: Evidence from China, with Yinan Liu
Most countries have specific rules indicating at what age children should enter primary school and how long primary education should last. Evidence from international comparisons and psychological research shows advantages of a later start to formal education. Thus many parents postpone the time children starting primary school. This behavior is called red-shirting. For example, in the US, upper-income and highly-educated parents red-shirt at the highest rate. In China, however, the opposite behavior of sending children to school early is more prevalent (more than 40%). This raises the question why Chinese parents make such choice. For example, is it because effect of relative age in developing countries and developed countries are different? Is it because information is more asymmetry in developing countries, i.e. parents don't realize the rational behavior should be postponing?