考量平行輸入之多區域價格競爭

Abstract

本研究探討在全球運籌下，各地之間因平行輸入而形成的價格競爭問題。不同於過往相關研究，本研究建構大型賽局模型描述多決策者下的價格競爭與平行輸入議題。此外，本研究證明純策略均衡存在於此賽局模型中，並發展以學習理論為基礎的演算法，有效求解此多決策者模型中的均衡解。 跨國公司往往透過授權地方事業群銷售其產品至各地區，然而所有地方事業群在銷售其產品時，應該考量平行輸入所造成的影響，因此需了解其他事業群的價格決策與彼此相對應的互動，方能制訂其產品的零售價格。本研究以混合羅吉特模型作為需求模型，考量地方事業群的價格決策與描述平行輸入產品在各地區的分佈情況。此外，本研究透過賽局模型建構各地方事業群之間的價格競爭，並透過學習理論為基礎的演算法進行有效求解。此模型與演算法可提供適當的方法來理解平行輸入帶來的影響與各地方事業群的訂價行為。 本研究的數值結果指出平行輸入有益於整體供應鏈利益。雖然整體利益提升，但各決策者可能有所獲利或損失。當整體利益提升時，各決策者的獲利與否取決於其本地市場的相關特性。此外，平行輸入對於各決策者的影響於數值分析中有進一步的探討。

This research studies price competition problems in global operations with parallel trades between different regions. Different from other research in literature, a large number of decision makers are simultaneously considered in a proposed game-theoretic model. It is shown that a pure strategy Nash equilibrium exists in this game-theoretic model. Furthermore, a learning-based algorithm is developed to efficiently search the Nash equilibrium among the decision makers. When a multinational company sells its product through local business units (LBUs) in different regions, all LBUs should be aware of the effect of parallel trades and adjust their price decisions accordingly. To model the interaction between LBUs, the demand of each LBU is modeled by a mixed multinomial logit model that considers price decisions of all LBUs. A game-theoretic model is then constructed to capture the interaction and price competition among decision makers. To derive the equilibrium between LBUs, a learning-based algorithm with computational efficiency is proposed. The model and the algorithm provide as useful tools for grasping the influence of parallel trade and the behavior of LBUs. In the numerical analysis, numerical results suggest that parallel trade benefits whole supply chain. However, while the total profit increases, parallel trade is either beneficial or detrimental to individual decision makers, according to the market characteristics of their home market. The influences of parallel trade on individual decision makers are analyzed by numerical studies as well.