The demand partitioning method for reducing aggregation errors in the solut
ion of aggregated p-median problems is introduced in this paper. The method
consists of eliminating source A and B aggregation errors using the Curren
t and Schilling (Geographical Analysis 1987;19:95-110) weighting method and
then partitioning the basic spatial units to eliminate source C errors. Th
ese two steps are repeated until all cost estimate error is eliminated in t
he solution of the problem. Data from the Central Valley of Costa Rica are
used to test this demand partitioning method. Specifically, population cens
us data are used to represent demand for services while current health clin
ics locations are the potential service supply points. The demand partition
ing method outperforms current published methods for reducing source C erro
rs.