This dissertation has four interrelated studies, namely (1) the characterization ofmilk purchase choices which included the purchase of organic milk, both organic andconventional milk and conventional milk only; (2) the estimation of a single-equationhousehold demand function for organic and conventional milk; (3) the assessment ofbinary choice models for organic milk using the Brier Probability score and Yatespartition, and (4) the estimation of demand systems that addresses the censoring issuethrough the use of econometric techniques.In the first paper, the study utilized the estimation of both multinomial logit andprobit models in examining a set of causal socio-demographic variables in explaining thepurchase of three outcome milk choices namely organic milk, organic and conventionalmilk and conventional milk only. These crucial variables include income, householdsize, education level and employment of household head, race, ethnicity and region.Using the 2004 Nielsen Homescan Panel, the second study used the Heckmantwo-step procedure in calculating the own-price, cross-price, and income elasticities by estimating the demand relationships for both organic and conventional milk. Resultsindicated that organic and conventional milk are substitutes. Also, an asymmetric patternexisted with regard to the substitution patterns of the respective milk types.Likewise, the third study showed that predictive outcomes from binary choicemodels associated with organic milk can be enhanced with the use of the Brier scoremethod. In this case, specifications omitting important socio-demographic variablesreduced the variability of predicted probabilities and therefore limited its sorting ability.The last study estimated both censored Almost Ideal Demand Systems (AIDS)and Quadratic Almost Ideal Demand System (QUAIDS) specifications in modeling nonalcoholicbeverages. In this research, five estimation techniques were used whichincluded the usage of Iterated Seemingly Unrelated Regression (ITSUR), two stagemethods such as the Heien and Wessells (1990) and the Shonkwiler and Yen (1999)approaches, Generalized Maximum Entropy and the Dong, Gould and Kaiser (2004a)methods. The findings of the study showed that at various censoring techniques, priceelasticity estimates were observed to have greater variability in highly censored nonalcoholicbeverage items such as tea, coffee and bottled water.
