Collaborative Research on Dynamic Antitrust Policy: Predatory and Limit Pricing in a Model of Learning-by-Doing and Organizational Forgetting

本项目考察了掠夺性定价和限制性定价在马尔可夫完全均衡中出现的程度，在马尔可夫完全均衡中，企业面临着“边做边学”和组织遗忘的前景。掠夺性定价适用多种标准，包括传统的Areeda &amp； Turner（1975）定价低于平均可变成本的标准和许多学者（如Ordover &amp； willis（1981）和Cabral &amp； Riordan(1997)）提出的经济测试。迄今为止，还没有研究表明掠夺性定价的经济和法律定义如何在具有进入和退出的价格竞争动态模型中运作，也没有对掠夺性定价在这种模型中均衡产生的程度进行全面分析。此外，还没有在完全动态模型中研究禁止掠夺性定价对福利的影响。本研究试图填补这些空白。掠夺性定价一直是反垄断政策中一个有争议的领域，一些学者（如McGee 1980）认为掠夺性定价是罕见且无效的，而其他学者（最著名的是Kreps， Milgrom， Roberts和Wilson在几篇论文中）表明，如果企业关注建立强硬的声誉，掠夺性行为可以作为一种均衡现象出现。当学习型经济存在时，区分掠夺性和非掠夺性行为尤其具有挑战性，因为公司既有掠夺性的理由，也有非掠夺性的理由，收取低于自付成本的价格。非掠夺性的原因是，低于成本的定价使公司更有可能在学习曲线上走得更远；由于这个原因，即使是垄断者也可能采取低于成本的定价。掠夺性的原因是，通过比竞争对手更快地降低其学习曲线，一家公司可能会诱使其竞争对手退出该行业，从而使该行业对新进入者失去吸引力。在20世纪80年代和90年代评估日本公司在DRAM芯片市场上的定价行为时，难以区分掠夺性定价和非掠夺性定价是一个问题（Flamm 1996）。更广泛的影响：本研究有望产生更广泛的社会影响，超出其对产业组织（IO）理论文献的贡献。在过去几年中，IO文献在分析行业动态方面取得了相当大的进展，现在可以相当容易地计算大型动态随机博弈的均衡（即使存在多个均衡）。此外，学者们在估计这类动态模型的原语方面也取得了很大进展。鉴于这些进展，这一分析预计将是向前迈出的重要一步，为负责制定和执行反垄断政策的机构提供一种工具，以确定在特定情况下公司是否可能从事掠夺性行为以及政策干预可能产生的福利后果。更一般地说，这项研究是朝着使动态分析成为回答公共政策问题的标准方法的目标迈出的一步。

This project examines the extent to which predatory pricing and limit pricing arise in the Markov-perfect equilibrium of a model of industry dynamics in which firms face the prospect of learning-by-doing and organizational forgetting. A variety of standards for predatory pricing are applied, including the traditional Areeda & Turner (1975) standard of pricing below average variable cost and economic tests proposed by a number of scholars (e.g., Ordover & Willig (1981) and Cabral & Riordan (1997)). To date, there has been no work showing how economic and legal definitions of predatory pricing can be operationalized in a dynamic model of price competition with entry and exit, nor has there been a comprehensive analysis of the extent to which predation arises in equilibrium in such a model. In addition, there has been no work on the welfare effects of a ban on predatory pricing in a fully dynamic model. This research attempts to fill these gaps.Predatory pricing has been a contentious area of antitrust policy, with some scholars (e.g., McGee 1980) suggesting that predatory pricing is rare and ineffective, and other scholars (most notably, Kreps, Milgrom, Roberts, and Wilson in several papers) showing that predatory behavior can arise as an equilibrium phenomenon if firms are concerned with establishing a reputation for toughness. When learning economies are present, distinguishing predatory from non-predatory behavior is especially challenging because firms have both predatory and non-predatory reasons to charge a price below out-of-pocket cost. The non-predatory reason is that below-cost pricing makes it more likely that the firm moves further down its learning curve; even a monopolist may engage in below-cost pricing for this reason. The predatory reason is that by moving down its learning curve faster than its rivals, a firm may induce its rivals to exit the industry and make it unattractive for new entrants to come into the industry. The difficulty distinguishing predatory from non-predatory pricing was an issue in evaluating the pricing behavior of Japanese firms in the market for DRAM chips in the 1980s and 1990s (Flamm 1996).Broader impacts: This research is expected to have a broader social impact beyond its contributions to the theoretical literature in industrial organization (IO). Over the past few years, the IO literature has made considerable progress in analyzing the dynamics of an industry, and the equilibria of large dynamic stochastic games can now be computed fairly easily (even when there are multiple equilibria). In addition, scholars have made great advances in estimating the primitives of such dynamic models. In light of these advances, this analysis is expected to be an important step forward in providing agencies charged with developing and enforcing antitrust policy with a tool to determine in particular cases whether firms are likely to engage in predatory behavior and what the welfare consequences of a policy intervention are likely to be. More generally, this research is a step toward the goal of making dynamic analysis a standard methodology to answer questions of public policy.