Managing and reallocating inventory across two markets with local information

Abstract

Consider a firm that controls inventory centrally for two separate markets that are managed by regional managers having local demand information. The central planner provides a dedicated inventory level to each market, to ensure a minimum service level, but can reallocate inventory once associated demands are filled. How should such a firm solicit regional managers to share their local information so inventory levels can be set efficiently? In this paper, we study a forecast sharing game in such a firm, where demand information lies within regions and inventory is managed by a central decision maker. We show that incentives are misaligned and truthful information sharing is not an equilibrium. We study two potential ways to improve the system: a) regional managers directly place orders (rather than simply passing demand forecasts), or b) a pricing scheme for inventory reallocations is imposed. We show that under direct ordering a unique pure strategy Bayesian Nash equilibrium exists and makes retailers place orders based on their true forecasts, although the requested order quantities are not system optimal (e.g., some inventory pooling benefits are lost). Under an endogenous transfer pricing mechanism set by the central planner, we find that a first best truth telling equilibrium is possible under certain conditions. We conduct a numerical study to examine the different cases and find that when the critical fractile is high, direct ordering results in lower inventories while the reverse is true when the critical fractile is low. Expected profit comparisons show the value of local information.