Abstract
A generating system of future household's attributes is developed in this study. The system creates the weight for each sample case obtained from the disaggregate survey to match present multiple aggregate marginal distributions, and simulates the demographic changes in each sample household along the time dimension. The simulation results applied to Osaka City suggest that the distributions of the number of trips or vehicle-holdings differ from those obtained by a conventional weighting method, that thesystem predicts population and number of households for the year 1995 fairly well using the data from 1990, and that the distribution of the predicted 2020 populations by repeated simulations has a high kurtosis.
