Simulating competitive collocation strategies with deep reinforcement learning
Abstract
本文介绍了一个基于深度强化学习的仿真模型,用于研究两家国际化企业的竞争性共址策略。借鉴国际商务(IB)中关于战略互动和竞争性共址的文献,我们探讨了在产品市场中的双寡头进行市场寻求型国际化时,决定共址程度绩效优化的因素。我们的仿真表明,在许多情况下,两家企业之间的高度共址是最优的。中到高水平的共址由低竞争强度、低进入壁垒、小本土市场规模以及高市场规模变异所驱动;而完全共址则出现在高竞争强度、低先发优势以及较小本土市场规模的情形下。高进入壁垒、低市场规模变异以及强烈的外来者劣势相结合,则会驱动完全相反的情形,即完全的相互回避。仿真的一个关键结论是,在全球扩张中聚焦竞争对手:在许多情况下,限制竞争对手的增长选项、并在市场准入变得更加受限之前抢先进入关键市场,至关重要。我们讨论了运用基于计算机仿真的方法就国际化开展理论建构的可能性。 Cet article présente un modèle de simulation fondé sur l'apprentissage par renforcement profond, destiné à étudier les stratégies de colocalisation concurrentielle de deux entreprises en phase d'internationalisation. Nous appuyant sur la littérature de l'interaction stratégique et celle de la colocalisation concurrentielle dans le domaine des affaires internationales, nous explorons les facteurs qui déterminent le degré de colocalisation permettant d'optimiser la performance dans le cadre d'une internationalisation axée sur la conquête de marchés, et ce pour un duopole sur les marchés de produits. Notre simulation suggère qu'un degré élevé de colocalisation entre les deux entreprises est optimal dans de nombreuses circonstances. Alors que les niveaux moyens à élevés de collocation sont conduits par une faible intensité concurrentielle, de faibles barrières à l'entrée, une petite taille du marché domestique et une grande variation de la taille des marchés, la collocation complète apparaît sous une forte intensité concurrentielle, un faible avantage du premier entrant et une plus petite taille du marché domestique. Des barrières à l'entrée élevées, combinées à une faible variation de la taille des marchés et aux handicaps et aux coûts liés au statut étranger de l’entreprise (Liabilities of Foreignness), conduisent à la situation inverse, à savoir une évitement mutuel total. L'un des principaux renseignements tirés de cette simulation est de se concentrer sur l'entreprise concurrente dans le cadre d'une expansion mondiale : dans de nombreux cas, il devient essentiel de limiter les possibilités de croissance du rival et de pénétrer des marchés clés avant que l'accès à ceux-ci ne devienne plus restreint. Nous abordons les possibilités d'utiliser des approches fondées sur la simulation informatique pour élaborer des théories sur l'internationalisation. This paper introduces a simulation model based on deep reinforcement learning to study competitive collocation strategies of two internationalizing firms. Drawing on the strategic interaction and competitive collocation literature in IB, we explore the factors determining the performance-optimizing degree of collocation in market-seeking internationalization for a duopoly in product markets. Our simulation suggests that a high degree of collocation between the two firms is optimal in many circumstances. Whereas mid-to-high levels of collocation are driven by low competitive intensity, low entry barriers, low home-market size, and high market-size variation, complete collocation emerges under high competitive intensity, low first-mover advantage, and smaller home-market size. High entry barriers, combined with low variation in market size and strong liabilities of foreignness, drive the opposite case of complete mutual avoidance. A key takeaway from the simulation is a focus on the competing firm in global expansion: in many cases, it becomes key to limit the rival’s growth options and to enter key markets before access becomes more restricted. We discuss the possibilities of using computer-simulation-based approaches to theorizing on internationalization. Este artigo apresenta um modelo de simulação baseado em aprendizado por reforço intenso para estudar estratégias de colocalização competitiva de duas empresas em processo de internacionalização. Com base na literatura sobre interação estratégica e colocalização competitiva em Negócios Internacionais (IB), exploramos os fatores que determinam o grau de colocalização que otimiza o desempenho na internacionalização orientada para mercados, em um duopólio no mercado de produtos. Nossa simulação sugere que um alto grau de colocalização entre as duas empresas é ideal em muitas circunstâncias. Enquanto níveis médios a altos de colocalização são impulsionados por baixa intensidade competitiva, barreiras de entrada reduzidas, tamanho reduzido do mercado doméstico e alta variação no tamanho dos mercados, a colocalização completa emerge sob alta intensidade competitiva, baixa vantagem do pioneirismo e menor tamanho do mercado doméstico. Barreiras de entrada elevadas, combinadas com baixa variação no tamanho do mercado e fortes desvantagens associadas à condição de estrangeiro, impulsionam o caso oposto: a completa evitação mútua. Uma conclusão-chave da simulação é o foco na empresa concorrente durante a expansão global: em muitos casos, é essencial limitar as opções de crescimento do rival e entrar em mercados estratégicos antes que o acesso se torne mais restrito. Discutimos as possibilidades de utilizar abordagens baseadas em simulações computacionais para teorizar o processo de internacionalização. This study explores how competition influences multinational firms' decisions to enter the same markets, known as collocation. The research aims to understand the dynamic strategic interactions that lead to collocation or avoidance between firms over time. The study uses a computer-simulation model to analyze the internationalization strategies of two competing firms in a market, focusing on industries with high concentration and limited synergies between locations. The simulation examines firms' market-entry decisions across 50 countries over 50 years, using 10,000 scenarios with varying conditions. This article uses a computer simulation combined with machine-learning techniques, specifically neural networks and deep reinforcement learning (DRL), to study firms' market-entry strategies. The simulation models a duopoly, where two firms compete in a market with no synergies between locations. The firms aim to optimize their internationalization strategies based on their competitor's moves. The simulation runs for 50 time periods, with firms starting in a randomly assigned home country and expanding to other countries over time. The model considers factors like market size, competitive intensity, and entry barriers to determine firms' market-entry decisions. The DRL algorithm helps firms adapt their strategies dynamically, allowing the study to explore the complex relationships between internationalization, performance, and competition. The simulation results suggest that firms often collocate in markets, especially when competitive intensity is low, entry barriers are low, the home-market size is small, and there is high variation in market size. Complete collocation occurs under high competitive intensity, low first-mover advantage, and when firms originate from smaller home markets. Conversely, complete avoidance happens when entry barriers are high, market-size variation is low, and firms face strong liabilities of foreignness (challenges faced by foreign firms in a new market). The study concludes that firms from smaller home markets can benefit from collocating with rivals from larger home markets to limit their growth opportunities. The research contributes to international business (IB) and strategy literature by providing insights into the dynamic strategic interactions that drive collocation and avoidance. It also highlights the potential of using machine learning and simulation methods to build theories in IB and strategy research. Future implications include using these insights to inform policy decisions and strategic planning for firms considering international expansion. This text was initially drafted using artificial intelligence, then reviewed by the author(s) to ensure accuracy. Este artículo presenta un modelo de simulación basado en aprendizaje profundo por refuerzo para estudiar las estrategias de co-localización competitiva de dos empresas en proceso de internacionalización. Basándonos en la literatura sobre interacción estratégica y co-localización competitiva en negocios internacionales (IB), exploramos los factores que determinan el grado óptimo de co-localización para maximizar el desempeño en la internacionalización orientada a la búsqueda de mercados en un duopolio de mercados de productos. Nuestra simulación sugiere que un alto grado de co-localización entre ambas empresas resulta óptimo en numerosas circunstancias. Mientras que los niveles medios y altos de co-localización están impulsados por una baja intensidad competitiva, bajas barreras de entrada, un mercado doméstico de menor tamaño y una elevada variación en el tamaño de los mercados, la co-localización completa emerge en contextos de alta intensidad competitiva, escasa ventaja del pionero y un mercado doméstico más pequeño. Por el contrario, altas barreras de entrada, combinadas con una baja variación en el tamaño de los mercados y fuertes desventajas de extranjería, conducen al caso opuesto de evitación mutua completa. Una conclusión fundamental de la simulación es la importancia de prestar atención a la empresa competidora durante la expansión internacional: en muchos casos, resulta crucial limitar las opciones de crecimiento del rival y entrar en mercados clave antes de que el acceso a ellos se vuelva más restringido. Finalmente, analizamos las posibilidades de utilizar enfoques de teorización sobre la internacionalización basados en simulación computacional.