An economic analysis of food safety and public health: Milk pasteurization in the United States

My dissertation is a comprehensive economic history study to the public health impacts of milk pasteurization in the United States. It has four major focuses which are included into four chapters. Chapter I is a case study to the public health impact of Chicago’s pasteurization ordinance. This study sets up the causal relationship between milk pasteurization and health outcomes. Chapter II extends the new econometric tool, the synthetic control methods, from a single unit to multiple treated units. This chapter also measures the impacts of pasteurization ordinances in a group of cities. Chapter III is written from an econometric perspective. It concerns how the synthetic control method can be transformed into a linear regression based model, which has more potential for empirical policy evaluations. Chapter 4 takes an alternative view to milk pasteurization. It discusses how the extent of pasteurization could make difference to public health. It also compares estimations of regular least square model and robust panel data model. Using Chicago’s 1916 pasteurization ordinance as a comparative case study, the first chapter focuses on how to measure the health impacts of food safety interventions. Empirical evidence suggests there was a clear causality relation between milk pasteurization and variations in the health outcomes of interest in Chicago. Thus, I applied the non-parametric synthetic control approach to capture causal health effects of this ordinance. The results suggest that the effect of this policy intervention was more pronounced in Chicago than in its 20 comparison cities, so I conclude that Chicago’s 1916 pasteurization ordinance had positive health effects. The second chapter examines causal health effect of mandatory city pasteurization ordinances in the United States. I apply the synthetic control methods to multiple treated units (MTSCM). Results indicate noticeable health benefits are observed in some cities but not all. For inferences, non-parametric rank-sum tests are preferred because of non-normal outcomes in the control group. This study also suggests regression based Difference-in-Difference (DD) models lead to different results than SCM, since SCM reveals more information like unit-varying and time-varying treatment effect. The third chapter aims to provide a robustness test for major conclusions obtained from prior chapters, e.g. the effect of Chicago’s 1916 milk pasteurization ordinances. Using the synthetic control methods (SCM), I found a significant treatment effect. To verify SCM results, I use a linear regression based cross-sectional time series model (CTM) to re-estimate this intervention. CTM results confirm major findings in my prior SCM studies. In addition, I use the 1989 California cigarette sales tax as an “out-of-sample” robustness check for CTM. Again, CTM results are similarly significant as SCM. The last chapter measures health impacts of variations of extent of pasteurization. Empirically, I choose the Fixed-Effects model to control unobserved intra-city variations. With respect to influential observations, I use robust estimators to validate least squares estimations. Compared with OLS estimate, robust estimates of the coefficients are smaller in absolute value. But their standard errors are even lower. In sum, my FE regressions also support the positive health effect of pasteurization