Takashi Unayama, "The Engel Curve for Alcohol and the Rank of Demand Systems," Journal of Applied Econometrics, Vol. 21, No. 7, 2006, pp. 1019-1038. The data used in this paper are the property of the Ministry of Internal Affairs and Communications (Japan) who do not permit open access. Hence the data cannot be lodged here. The JFIES is designated as important statistics compiled by the government (Designated Statistics No.56). It is carried out in accordance with the Family Income and Expenditure Survey Regulation (Prime Minister's Office Ordinance No.71 of November 12, 1975) issued on the basis of the Statistics Law. The researcher should have a permission to use from the Japanese Government. The list of variables and code book will be given from the statistical bureau when the data can be used. The list of variable only shows the order of variables and number of digits for each variable. In other words, the dataset does not include variable names, and the variables have to be identified personally. The analysis in the paper is based on data for October and November of 1997-2000. The analysis in Section 4 also uses data from 1989-1990 and 1994-1995. The sample size of the data for "Worker's household" in October and November of 1997-2000 is 38,331. Firstly, I limited the sample only to the nuclear households. This means that households for which "Household Type" is coded as "01", "02", "91" or "92." This reduced the sample size to 30,686. Second, I chose households who have their own house, which is shown with the "Tenure of House" code being "01." The sample size then become 18,313. Third, "Number of Household Members" was limited to two, and the sample size became 4,747. Finally, I removed some exceptional cases such as where the household head or wife is a student ("Type of School" is not empty), paying school fees (expenditure on "School Fees" is not zero), paying land rent (expenditure on "Rents for dwelling and land" is not zero), and purchasing children's clothes (expenditure on "Child Clothes" is not zero). These criteria made the sample size 4,602. In addition to these criteria, I have excluded observations that lie outside of 5% quantile points for total expenditure and yearly income, and the sample size become 3,995. For the other periods, I use the same sample selection. To define the "total expenditure" in the paper, I have subtracted expenditures on"Repairs and maintenance of Housing", "Automobiles" and "Pocket money", "Social Expenses", and "Remittance" from "Total Consumption Expenditure." Takashi Unayama unayama [at] kier.kyoto-u.ac.jp