Abstract

Purpose of the Multi-Accounting System (MAS) is to give a comprehensive survey of the stocks and flows of people in education and training, in employment, unemployment, and household activities. The flows between the stocks are represented by transition matrices. These data are planned to be used for theory-based analyses concerning the relationship between the labour market and adjacent statuses. The results are important for a rational design of regional labour market policy.
To estimate the transition matrices within MAS a Bayesian inference model is developed. Given a prior probability distribution on the set of possible transition matrices and information about the actual matrix consisting of a set of linear equality and inequality constraints the so-called ADETON procedure computes the transition matrix with highest posterior probability. By this approach it is possible to specify in addition soft constraints which are observed only up to a certain degree. If only linear constraints have to be obeyed the results of ADETON are equal to those obtained by entropy optimization. In this respect the method is a generalization of the well known Iterative Proportional Fitting Algorithm (used in log-linear models) or of the equivalent RAS method (used in input-output analysis). ADETON is applicable for many purposes, e.g. to estimate contingency tables or input-output and other flow matrices.