This paper discusses the accuracy and stability of an aggregation procedure based on extreme point generation. Numerical results obtained from an empirical model show that the aggregation procedure is highly satisfactory in tenns of aggregation errors and the aggregate model is very stable under objective function parameter changes.