Appended Note 3-3 Estimation by the cohort analysis

1. The cohort analysis is a method of grasping the respective influences on changes in time-series variables of the age effect typical of age groups, the period effect typical of periods (business cycles, trends, etc.) and the generation effect typical of cohorts classified by the year of birth. Here, this approach is explained using the analysis of consumption expenditure and the savings rate in the Family Income and Expenditure Survey as an example.
2. In a specific analysis, the results of the indicator concerned by age group of household heads are estimated by the least-square method setting up dummy variables for age, period and generation. Dummy variables were set using the generation of those born between 1941 and 1945 (a generation ahead of the baby-boom generation), aged between 35 and 39 in 1980, as the benchmark. The analysis was conducted with the assumption that the generation effect is the same for those aged 24 or below and those aged between 25 and 29 in 2004 as a constrained condition.
3. In order to remove the trend of the period effect, the zero-sum constraint of parameters is applicable along the line of Deaton (1997). For items taken in for health maintenance, however, the zero-sum constraint is not applicable because data are available only from 1995.
4. With the age classification by five years, results for the same generation can be obtained with the use of results every five years between 1980 and 2000. The latest evaluation for the identical generation would require the results for 2005, but in this analysis, the results of 2004 are used as a substitution because they represent the closest cohort classification. For this reason, the age effect and the generation effect were estimated with each generation out of alignment by one year. (Example: the results were obtained for the generation of those born between 1971 and 1975 up to 2000, but in 2004, the results for those born between 1970 and 1974 were used as a substitution.)
5. For the estimation for 2010, the period effect for 2010 was assumed to be the same as for 2004. The generation effect for the generation of those who would be 24 or under (the generation of those born in 1986 or later) was assumed to be the same as for the generation of those aged between 25 and 29 (those born between 1981 and 1985). In estimating the results for all households by the household composition by the age group of household heads in 2010, we used the future household estimate by the National Institute of Population and Social Security Research.
6. In the analysis of the consumer propensity and expenditure in Section 2, the estimation was made using annual data for 1980, 1985, 1990, 1995, 2000 and 2004 of the Family Income and Expenditure Survey of the Ministry of Internal Affairs and Communications. For the propensity to consume and the savings rate, we used the results for wage-earners' households (households with two or more members, excluding households in the agriculture, forestry and fisheries sector), and for the ratio of each consumption item to total expenditure, we used the results for all households (households with two or more members, excluding households in the agriculture, forestry and fisheries sector). For an analysis of residential house acquisition in Section 2, we used the "statistical survey on housing and land" by the Ministry of Internal Affairs and Communications, and for an analysis of the labor force participation rate in Section 3, we used the "Labor Force Survey" by the Ministry of Internal Affairs and Communications.