Atsushi Kimura (atkimura@nstac.go.jp, kimura@ta2.so-net.ne.jp)
National Statistics Center, Japan
ABSTRACT
This paper shows for the first time that regional characteristics of food purchasing behavior in Japan can be visualized by cluster analysis of published data from the “Family Income and Expenditure Survey” conducted by Statistics Bureau of Japan. We also show that these regional characteristics are common and stable structures by analyzing data from several different years of the household survey.
The “Family Income and Expenditure Survey” data contains no geographic information such as regional adjacency or distance information between regions, and the data set consists purely of purchase amounts. However, as this paper will show, regional characteristics are inherent in the food item purchasing behavior of Japanese households.
We also discuss the possibility of applying this stable regional characteristic inherent in the “Family Income and Expenditure Survey” to the Analyse phase (GSBPM: Analyse phase (6.2 Validate outputs sub-process)).
Finally, we introduce several new methods that are useful in interpreting the results of different survey-year data analyses. These methods make it possible to quantify the contribution of each variable to the dissimilarity between agglomerating clusters in a multivariate cluster analysis. It also greatly reduces the researcher’s effort when intercomparing the results of analyses of data from different survey years.
Keywords: Food item purchasing behavior, regional characteristics of food item purchasing behavior, LIV method, cluster agglomeration contribution, Ward method hierarchical cluster analysis, dissimilarity
JEL Classification: JEL: D10, C38
