| Chapter 1 examines the dynamic competition between Online Retailers and traditional Brick-and-Mortar retailers using the dynamic games framework (Ericson and Pakes (1995), Pakes and McGuire (1994)). The transactions costs of buying online declines stochastically as the amount of goods sold online increases, reflecting the impact of learning-by-doing effects and word-of-mouth effects on consumers' familiarity with online retailing and the online retailer's ability to design user-friendly websites. Using a differentiated goods model, I find that aggressive price discounting and output growth is consistent with rational intertemporal optimization. The model is extended to analyze the impact of changes in the growth rate of the market size and transaction cost externalities, and I find that if web-sites have ‘low stickiness’ and externalities are high, online retailers should be less aggressive in their price discounts and output growth.; Chapter 2 examines tacit collusion and consolidation in the online retail industry using the dynamic games framework. I use the Fershtman-Pakes (2000) algorithm to analyze the sustainability of collusion in the online retailing industry. Collusion is only sustainable when the firms have similar productive efficiencies and when the market is not growing too rapidly. When the disparity in productive efficiency is sufficiently large, the more-efficient firms frequently engage in price wars to force out the less-efficient firms, and the latter frequently exit the industry.; Chapter 3 introduces imperfect information to dynamic games by explicitly incorporating beliefs into the firms' objective functions and providing a mechanism for updating those beliefs dynamically in response to observed changes in the market environment. With learning-by-doing effects, I find that the amount of ‘excess output’ relative to the static equilibrium depends on the firms' current level of efficiency, the current size of the market, and their degree of optimism about the rate at which the market could grow. A simple simulation shows the potential for ‘under-reaction’ and/or ‘over-reaction’ by the firm in response to changes in the environment. The other learning rule examined in this paper is variations of the reinforcement learning approach. |
