UC Riverside

This dissertation presents three chapters that provide insights into the public health issues, such as psychological well-being (PWB) and infant health. In Chapter 1, I apply machine learning methods to predict people’s psychological well-being using a U.S. large dataset. The main outcome variables used to quantify psychological well-being are: general happiness, satisfaction with financial situation, and satisfaction with job. In order to predict PWB, first, I use K-nearest neighbor (KNN) algorithm to select and rank the importance of predictors. I present that marital status has the highest importance score in predicting one’s general happiness. Prestige score of occupation is the most important predictor of satisfaction with jobs. Next, I utilize the Forward Selection algorithm to find the best com- bination of predictions. Using this selected combination to predict people’s PWB, I achieve 70% - 80% classification accuracy when detecting people with low psychological well-being. Lastly, I provide insights that PWB is an important factor that affects people’s behavior by investigating how PWB is associated with physical and mental health, risky goods consump- tion, investment decisions, and working behaviors. I find that happier people have better health conditions, smoke and drink less, have more confidence in financial institutions, and generally work more hours.In Chapter 2, I examine the effects of medical marijuana laws(MMLs) on infant health using Vital Statistics Natality data from 1996 to 2016. I exploit the geographic and temporal variation in the implementation of MMLs using a difference-in-differences estimation framework. I find that MMLs are associated with a 0.251 percentage point (p<0.01, 3.74% of the mean) increase in the incidence of low birth weight (<2500 grams), and a 0.435 percentage point (p<0.05, 4.2% of the mean) increase in the incidence of premature births (<37 weeks). The effects are statistically significant among births of white mothers, and partly significant among births of black mothers. Using an event study design, I show that the effects are persistent and long-lasting. This study suggests that there should be a more cautious use of medical marijuana use among pregnant women. In Chapter 3, we present the evidence on a previously recognized but under-investigated decrease in birth weight in the United States during the first decade and a half of the 21st Century. From 2000 to 2006, mean birth weight for US singletons decreased by 1.53%, and has only partially recovered since 2007. The declines in birth weight occur at all gestational ages, for all races, within all maternal age bins, for both smokers and non-smokers, for vaginal and c-section births and at all quantiles of the birth weight dis- tribution. These trends are of great concerns. We provide some evidence for the declining mean birth weight in the U.S. that could partially explained by changes in gestational length and induction rates.