Evaluation of the empirical performance of two-stage budgeting AIDS, QUAIDS and Rotterdam models based on weak separability

Microsimulation models for indirect taxation require detailed underlying demand systems, in order to be policy relevant. A possible solution for the econometric problem (lack of necessary degrees of freedom) is the separability concept and the closely related notion of two-stage budgeting. In this paper, weak separability is applied on the Almost Ideal Demand System (AIDS), its quadratic extension QUAIDS and the Rotterdam model. These two-stage budgeting demand systems were estimated on Belgian time series data and were evaluated by means of a comparison of their elasticities (both partial and total), goodness-of-fit measures and their forecasting accuracy. Though the rank three QUAIDS model does not dominate the others in every respect (at least for time series data), it has nice theoretical properties which can on their own be a justification for the use of the system.