The aggregation theory of mathematical programming is used to study decentr
alization in convex programming models. A two-level organization is conside
red and a aggregation-disaggregation scheme is applied to such a divisional
ly organized enterprise. In contrast to the known aggregation techniques, w
here the decision variables/production planes are aggregated, it is propose
d to aggregate resources allocated by the central planning department among
the divisions. This approach results in a decomposition procedure, in whic
h the central unit has no optimization problem to solve and should only ave
rage local information provided by the divisions.