Nonparametric and parametric analyses of food demand in the United States.

摘要

The objective of this study is to develop complete demand systems for 19 food categories in the United States. Monthly time-series data are constructed from the continuing Consumer Expenditure Surveys data from 1980 to 1986. This study employs several advanced methodologies including nonparametric testing of data consistency, rank condition of demand systems, factor and cluster analyses, flexible functional forms, translating methods, as well as nonlinear system estimation.;The nonparametric analyses are conducted to ensure that the data base used are consistent with the underlying microeconomic theory. The weak separability and functional forms are also tested and verified. The data consistency test reveals that the 19 food categories as a group satisfies the assumptions of utility maximization by a representative consumer. The rank test results suggest that the minimum number of forming groups from 19 food categories is three. Furthermore, based on the clustering analysis, the 19 food categories can be classified into six subgroups, and weak separability conditions are satisfied for these subgroups.;The PIGLOG demand system with dynamic factors, demographic factors and first-order autocorrelation is shown to be the most appropriate model. Seven alternative versions of the linear approximation of Almost Ideal Demand System (LA/AIDS) are estimated for the 19 food categories under a one-stage budgeting procedure. In addition, three PIGLOG models, i.e., Lewbel's full model, translog and LA/AIDS are compared under a two-stage budgeting procedure. The empirical results show that the estimated elasticities are sensitive to the assumptions of underlying preference structure and model specifications. The parametric results detect a unstable demand structure during the sample period for beef, milk, other dairy products, processed vegetables, sweets, miscellaneous foods, and food away from home. Among 19 food categories, only food away from home has a strong negative elasticity of family size. The empirical results also indicate seafood, sweets, miscellaneous foods and food away from home are expenditure-elastic. The dynamic LA/AIDS models for the entire 19 food categories generally have a good predictive performance and are better than the static models.