Three Essays On Demand Analysis With Food Spoilage

In Essay One, "Demand for Perishable Foods: A Cost of Consumption Framework with Policy Implications", we challenge the convention of treating consumption and purchased quantities as identical even for food products that undergo substantial deterioration while in storage. We first prove a theorem that shows how decay processes can be incorporated into any existing demand system in a theoretically consistent way. Our method involves augmenting the prices in an existing neoclassical demand model with a function of shopping frequency and decay parameters. The augmented prices have the intuitively appealing interpretation of being 'consumption prices' which reflect the cost of consuming a unit of food. We apply this method to the Quadratic Almost Ideal Demand System to estimate household-level demand for foods that vary by their degree of perishability. Included composite goods include 'fresh fruit', 'fresh vegetables', 'non-fresh vegetables', 'meat', and miscellaneous food 'other'. We are able to directly estimate the quality-adjusted decay rates that rationalize observed household demands. We ask, does our framework perform better at estimating perishable food demand than 'naive' models that ignore perishability? Because our framework nests the 'naive' model, we are able to conclude that our framework performs better. We argue that our framework is well-suited for studies that inform public health policies which attempt to improve diets through improving access.Essay Two, "Measuring Consumer-Level Quality-Adjusted Food Loss: A Demand System Approach", extends the model developed in essay one by accounting for 1. the endogeneity of six variables including five shopping frequencies and total expenditure, and 2. household heterogeneity. Household heterogeneity is introduced using demographic translating to account for theoretical consistency. Necessary parameter constraints for economic regularity, namely, adding-up, zero-degree demand homogeneity in prices and income, and Slutsky symmetry, are imposed on the system of translated demand equations. Because of the highly non-linear form of the demand system, control functions (cf. instrumental variables) are used. We use our model results to infer the quality-adjusted food waste that rationalizes observed household market behavior to be around 55%.Essay three, "Store-Format Choice: A Competing Risk Approach", introduces a novel model of food retail store choice. The time since the last shopping trip is central to our competing risk framework. When a household shops, their clock starts, ticking down the time until they shop again. At each tick, they choose to shop at 'competing' store formats or may choose not to shop at all. The hazard functions of these choices, which fully characterize shopping probability distributions, are jointly estimated. This approach has many advantages over the multinomial logit models which dominate the store-format choice literature. First, we use the information of when the household does not shop. Second, by construction, shopping timing is accounted for in a way that avoids any possibility of endogeneity. Third, our non-parametrically estimated household-specific baseline hazard functions can take virtually any functional form. Indeed, we find that household hazard functions are bimodal, with households feeling increasing pressure to shop at all store formats until a peak around 7-10 days, feeling decreasing pressure until a nadir around 15-20 days, and increasing pressure thereafter. The peak at 7-10 days suggests that weekly schedules and perishable food shelf-lives strongly impel households to shop. The richness of our framework allows us to draw many other conclusions.