Local aggregation in a dynamic setting (Q911971)

Although it is not possible in general to give a macro description of a macro model over the entire state space of the micro model, it is possible to aggregate exactly over a subspace. This set of micro states where aggregation is possible forms the aggregation set. The paper considers the local (linear) case. The macro model is chosen such that it gives rise to an aggregation set which is approached by the micro movements as fast as possible. This is done technically by selecting appropriate eigenvalues of the micro model and by construing a macro model based on these eigenvalues. As a result, the inferred macro model will describe trajectories of the aggregate quantities towards which the ``true'' trajectories of these quantities, as generated by the micro model, are tending. It is proved that the inferred macro model operates in the real domain (rather than in the complex domain) under weak conditions. The paper concludes with some more general observations on the interpretation of macro relations. A numerical example illustrates the procedure.