|
Abstract:
|
In this project, we firstly focused on visualization for the correlation on consumer behaviors of different subpopulation towards certain groups of items, based on Consumer Expenditure(CE) public-use microdata files data, similar works had been done such as Discrimination in new car purchases, see Pinelopi Koujianou Goldberg (1996), or Chinese tourists habit to Guam, see Connie and Thomas (2000). We considered to partition population based on income status and/or age and/or family size, etc. At the same time, we classified items in CE data based on Universal Classification Code(UCC). Main categories include (but not limited to) Food, Housing, etc. Given the visualization results, we choose specific pairs of subpopulation and expenditure group to build up a time series model. Category for each subpopulation was treated as annual time series and further analysis had been done for both mean trend and residual parts. Based on analysis result and demographic structure, one-year ahead predictor was given to estimate the next year national wide demand.
|
