Economic and measurement issues in diet diversity and obesity in China

Abstract

Using the China Health and Nutrition Survey 2004-2009 this thesis reports the examination of three separate yet related questions: (i) Do levels of income and education affect the diversity of diet among Chinese adults?; (ii) Are the estimated impacts of income and education sensitive to how diet diversity is measured?; and (iii) Are overweight and obesity confined to affluent and better educated Chinese? The thesis is organised as follows. Chapter 2 examines the relationship between income and diet diversity in China, and is the first research to do so. The finding reinforces existing research suggesting that higher income increases diet diversity and, thus, enhances public health. But the analysis reported in the chapter goes beyond replication of previous studies, providing new evidence on the form of the relationship. The estimated marginal effect of income is found to fall sharply from the lowest to the second lowest quintile, as well as over time, suggesting that income might become irrelevant in influencing diet diversity in China as income levels continue to rise. This chapter is also the first study to explore the potential endogeneity in the income-diet diversity relationship. Arguably, endogeneity in this relationship most likely arise from omitted variable bias, and Chapter 2 shows that ignoring the issue apparently understates/overstates the income/education effect. In particular, while estimates of the education effect are positive with both OLS and 2SLS methods, it is much smaller with the latter and even becomes statistically insignificant in 2009. There are numerous indices of diet diversity and this hinders efforts to compare estimates of the income and education effects across studies. Chapter 3 addresses this issue and establishes that the sign, significance, and patterns along both the income distribution and over time are robust with respect to how diet diversity is measured. However, food classification and index construction methods do influence the estimates' magnitudes. Chapter 3 finds that the income elasticity is larger when the index is constructed from the counting method and a disaggregate food classification. The chapter also argues that a count index using a detailed food classification better reflects the health benefits from diet diversity because it captures diversity both within and between food groups. It is suggested that such an index is a more appropriate measure for research and policy focusing on health aspects of diet diversity. Chapter 4 investigates how household income and education attainment influence Body-Mass Index (BMI) and the probability of overweight/obesity. Contradicting previous evidence suggesting that excessive weight is confined to the affluent in China, this chapter reveals a dynamic transition where overweight/obesity risk is trickling down to the poor. Between 2004 and 2009, the data reveal a negative relationship between income and BMI, a result is often observed in developed countries. For year 2009, BMI and the possibility of overweight/obesity are found to decrease as household income rose, especially among women. Moreover, more schooling was associated with higher overweight/obesity risk among men, yet and interestingly, better educated women had a lower risk of being overweight/obese. Show more