A healthy diet is related to a low risk of chronic diseases. A large body of research is devoted to improving social welfare by promoting healthy eating. This dissertation addresses the relation of food and health by analyzing the money and time inputs in food, the food poverty measurement, and a corresponding health outcome.

The second chapter extends the current food poverty measure in headcount and proposes a set of Foster, Greer, and Thorbecke (FGT) indices, which is commonly used in development literature, in food poverty to allow for a more comprehensive understanding in food poverty evaluation. The counter-factual analysis on removing the American Recovery and Reinvestment Act (ARRA) component from the food expenditure shows that the original metrics underestimate the reduction to food expenditure poverty associated with ARRA, whereas the FGT indices indicate a slightly larger impact of ARRA in alleviating food poverty.

The third chapter uses the same FGT indices in food poverty measurement but focuses on the sensitivity of these measurements to a different spatial and temporary food price. We use linear regression to estimate the local level of food poverty thresholds. The results show the spatial and temporal-specific thresholds are higher than the national threshold. The West region shows the most severe poverty situation, indicating the importance of considering spatial and temporal variations in measuring food expenditure poverty. The decompositions of food expenditures show that both the Supplemental Nutrition Assistance Program (SNAP) benefits and money spent on protein play an essential role in reducing food expenditure poverty.

The fourth chapter combines the two datasets used in the previous two chapters to investigate the connection between the resources (money and time) devoted to food and a corresponding health outcome (Healthy Eating Index, HEI). Two-Sample-2-Stage-Least-Square (TS2SLS) model is used to account for the two different datasets in predicting the time spent on food-related activities. After obtaining the time input, a Three-Stage-Least-Square (3SLS) model shows the time input improves the HEI for Non-SNAP households, who are more constrained by time. The decomposition of the impact of education on the HEI shows the indirect impact account for 22% of the total impact. This analysis breaks down the impact of the characteristics on HEI through different channels, thus offers more comprehensive policy recommendations.