This thesis examines serial supply chains in a decentralised, market-basedenvironment. It tries to explain the existing world of decentralized supplychains through a novel modelling approach. The actors in the chain are independentand interact through a “Cournot” game in inventory policies (orderlevels) and pricing (mark-ups). The chain is serial so that the most downstreamactor – the retailer –sells to external customers. The thesis examines the effectof decentralisation on the actors’ inventory policies, upstream overstocking,supply chain inefficiency and the easiness of accepting vendor managed inventories.Each actor aims to maximise his own profit only. However, when demanduncertainty is proportional to mean demand, it turns optimal for each upstreamactor to set his inventory policy to minimize the total cost of the entire downstreamchain, given the downstream inventory policies. Thus, the costs of hiscustomers are as important to as his own costs. Consequently, vendor managedinventories becomes easily acceptable, as the vendor has no interest in overstockinghis customer.Numerical studies show that the multiple marginalization of the decentralizedchain causes inefficiency, but generally very little, typically a few tenths ofa percentage of total chain turnover or profit. Thus, Adam Smith’s invisiblehand seems to watch over real-life supply chains in general. Considerable inefficienciesrequire high holding and shortage costs compared to the production/transportation costs of the chain. This is most likely to happen when not directlysatisfied sales are lost, and the retailer’s part of the total margin is small.With backlogging it only happens when upstream profits are unrealisticallyhigh. Subsidies of retailer stockholding is remedy against inefficiency.The margins charged by upstream actors cause too low retailer stock whichis compensated by upstream overstocking (through upstream cost minimization).The studies further show that this process very easily can cause substantialupstream overstocking. This contributes to understand the often observedreal-world phenomenon of manufacturer and wholesaler overstocking.The now described model is the thesis’ major explanations of the existingworld of decentralised supply chains. However, it also investigates a variantwhere demand uncertainty is modelled independently of mean demand. Then,iiithe actors have no interest in the costs of their customers, but gain from downstreamoverstocking, so that vendor managed inventories seem unreliable.Moreover, upstream overstocking is not typical, and inefficiency is generallymuch greater than in the major model. The thesis argues that this model variantis less credible, but truly empirical investigations are missing.In the study of the major model, inefficiency typically increases when marginsmove upstream – always so when unsatisfied demands are backlogged, butsometimes minor upstream moves can improve efficiency when unsatisfieddemands are lost. Moreover, in a three-stage model the wholesale inventory increaseswhen the wholesaler merges with the manufacturer and decreases whenhe merges with the retailer.The results seem to hold also when decisions are made heuristically, as aproxy for real-life decision making.