Aviv and Federgruen (1998), with the aim of investigating impacts of information
sharing, consider a single vendor plus multiple retailers. They assume a VMI agreement that
leads to a fully centralized planning model where the vendor minimizes the system-wide total
cost of inventory holding and distribution. Using a combination of Markov decision process
and non-linear programming, they construct approximate policies for the vendor and the
retailers under both information-sharing alone and information sharing in conjunction with
VMI. They find that VMI (with that sharing) is always more beneficial than information
sharing alone.