Three Essays on Food Environment, Food Demand, and Health

Open Access
Chen, Danhong
Graduate Program:
Agricultural, Environmental and Regional Economics
Doctor of Philosophy
Document Type:
Date of Defense:
March 04, 2014
Committee Members:
  • David Gerard Abler, Dissertation Advisor
  • David Gerard Abler, Committee Chair
  • Edward C Jaenicke, Dissertation Advisor
  • Edward C Jaenicke, Committee Chair
  • Stephan J Goetz, Committee Member
  • Stephen Matthews, Committee Member
  • Food Environment
  • Food Demand
  • Self-rated Health
  • Obesity
  • Meta-analysis
  • Structural Equations Model
  • Multilevel Model
There have been a multitude of studies estimating food demand elasticities in China, but no published study has conducted a quantitative review of this literature for China. In the first essay, a database was assembled of 670 own-price elasticities, 917 cross-price elasticities, and 855 income elasticities from 85 studies estimating demand elasticities for 19 major food categories consumed in China. Instead of using raw elasticity estimates extracted directly from these studies, we converted all values to unconditional elasticity estimates if necessary. That is, the converted elasticities are all dependent on income rather than expenditures on either food or subcategories of food. Weighted least squares (WLS) was then used to examine the determinants of those elasticity estimates accounting for the possibility of heteroskedastic errors. The results showed that this approach to data conversion substantially reduced differences between conditional and unconditional elasticity values. They also indicate that own-price, cross-price and income elasticity estimates are significantly influenced by data characteristics, the budgeting framework employed in a study, type of demand model used in a study, and the mean per capita income of the sample analyzed in study. Using the meta-analysis results, this essay estimates elasticities of food demand in China in 2011. Although there has been extensive research on the adverse impacts of perceived discrimination on health, it remains unclear how perceived discrimination gets under the skin. The second essay develops a comprehensive structural equation model (SEM) by incorporating both the direct effects of perceived discrimination on self-rated health (SRH), a powerful predictor for many health outcomes, and the indirect effects of perceived discrimination on SRH through health care system distrust, neighborhood social capital, and health behaviors and health conditions. Applying SEM to 9,880 adults (ages between 18 and 100) in the 2008 Southeastern Pennsylvania Household Health Survey, this essay not only confirmed the positive and direct association between discrimination and poor or fair SRH, but also verified two underlying mechanisms: 1) perceived discrimination is associated with lower neighborhood social capital, which further contributes to poor or fair SRH; and 2) perceived discrimination is related to risky behaviors (e.g., reduced physical activity and sleep quality, and intensified smoking) that lead to worse health conditions, and then result in poor or fair SRH. Moreover, the results indicated that perceived discrimination is negatively associated with health care system distrust, but there is not a significant relationship between distrust and poor or fair SRH. Previous studies often employ two approaches to evaluate diet quality: one is to calculate dietary indexes/scores reflecting compliance with recommended nutrient or food intakes, while the other is to derive dietary patterns using factor or cluster analyses. Assessments of diet quality relying on data collected through individuals’ recall of past consumption frequencies for specified food items are subject to various measurement errors. The third essay evaluates a new dietary assessment method and its association with diet-related health outcomes. Households in the recent IRI Consumer Scanner data (2008-2012) recorded their food purchases on each shopping trip by scanning Universal Product Codes (UPCs). The healthfulness score for each household’s shopping cart measures the discrepancy between the household’s monthly expenditure shares of 24 aggregated food categories defined by USDA’s Center for Nutrition Policy and Promotion and the recommended values calculated based on the USDA Food Plan. In addition to household-level data derived from the IRI Consumer Scanner data, data was obtained on self-reported medical profiles and dietary components from household members, constituting the individual-level variables. Data from County Business Patterns and American Community Surveys were used to create food environment variables and poverty rates at the county level. Using this multilevel data set assembled from different sources, a series of multilevel models were estimated to examine the impact of individual diet quality, household shopping basket healthfulness, and food environments on overweight and obesity status. To mitigate potential endogeneity, densities of food stores and restaurants were calculated with a lag of one year. Model results showed that a higher shopping basket healthfulness score is significantly associated with reduced risk of being overweight and obese. Higher densities of supercenters in non-metro neighborhoods are found to be positively related to individuals’ overweight and obesity risks, whereas greater availability of full-service restaurants and limited-service restaurants per capita is associated with lower probability of overweight and obesity status. However, these results are not consistently found across models utilizing all various population subsamples. Nevertheless, an increasing relationship between food deserts and the likelihood of being overweight and obese was more consistently found across samples, which is consistent with contextual effects on weight outcomes.