Econometric Analysis of Dynamic Games with Limited Information

该项目将提供经验性研究企业在动态市场中的行为所需的计量经济学方法，在动态市场中，企业可能对周围环境和异构信息处理能力的信息有限，就像在数字市场中发生的那样，并为该领域的政策提供信息。实证研究动态市场中企业行为的现有计量经济学方法要求企业获取和吸收大量信息。企业需要知道竞争对手的特征和过去的行为，竞争对手在每个时间点观察到什么，任何其他相关的过去和现在的市场特征，以及这些特征在概率方面的未来演变。然而，当企业在现代网络市场上运作时，这种完美假设就变得不切实际了。通过在复杂的、分层的、不断变化的和模块化的环境中托管多个利益相关者，像亚马逊和eBay这样的在线市场已经不可比拟地增加了公司必须获得的信息量，才能完全知情。根据在市场上的经验和在储存和分析数据方面的复杂程度，一些公司可能对市场活动有全面的了解；其他人可能需要时间来适应、反应和实验，这也得益于人工智能学习算法。在处理和使用数字平台上可访问的大量信息时，这种摩擦构成了竞争监管挑战，因为它们阻碍了我们使用标准的计量经济学方法来估计公司的收入和成本，而这对于研究市场力量和反垄断案件至关重要。反过来，我们对在线市场中企业激励的了解主要局限于描述性证据，不足以指导政策。因此，开发计量经济学方法来处理信息有限的动态博弈，是为严肃的实证研究铺平道路的最重要问题，也是本项目的重点。新方法的严密理论将得到发展。新方法的性质将在模拟研究中进行检验，并将用实际市场数据说明它们的实现。开源代码将被发布，使从业者能够直接采用新工具。除了学术界，这些实践者还包括政府、公共机构和私营部门的专业经济学家。这些研究结果将引起计量经济学和产业组织学者的兴趣，他们将受益于对以前不存在的现代动态环境进行实证研究的方法。此外，该项目将首次为政策制定者提供适当的工具，以分析市场力量，发现反竞争行为，并保护在线市场中的消费者。新方法还将帮助数字平台增强对用户动机的了解，这对于设计保护收入和其他关键利益的平台规则至关重要。

This project will provide the econometric methods needed to empirically study firms' behaviour in dynamic markets where firms may have limited information about the surrounding environment and heterogeneous information-processing capabilities, as happens in digital markets, and inform policy in this area.The available econometric methods to empirically study firms' behaviour in dynamic markets require firms to access and absorb a large amount of information. Firms need to know competitors' characteristics and past actions, what competitors observe at each point in time, any other relevant past and present market features, and the future evolution of such features in probabilistic terms. However, this perfection assumption becomes unrealistically demanding when firms operate in modern online markets. By hosting multiple stakeholders in intricate, layered, constantly changing, and modular environments, online markets such as Amazon and eBay have incommensurably increased the amount of information firms must acquire to act as perfectly informed. Depending on the experience in the market and the sophistication in storing and analysing data, some firms may have a full view of the market activities; others may need time to adapt, react, and experiment, also helped by artificial intelligence learning algorithms. Such frictions in processing and using the sheer volume of information accessible on digital platforms pose competition regulatory challenges because they prevent us from applying standard econometric methods to estimate firms' revenues and costs, which are essential for studying market power and antitrust cases. In turn, our knowledge of firms' incentives in online markets is mainly confined to descriptive evidence and is insufficient to guide policy. Developing econometric methods to handle dynamic games with limited information is, therefore, an issue of utmost importance to pave the way for serious empirical studies and will be the focus of this project. A rigorous theory for the new methods will be developed. The properties of the new methods will be examined in simulation studies, and their implementation will be illustrated with real market data. Open-source codes will be published, making the adoption of the new tools straightforward for practitioners. In addition to academics, these practitioners include professional economists in governments, public agencies, and the private sector. The research findings will interest scholars in econometrics and industrial organisation, who will benefit from having methods to empirically study modern dynamic environments that did not exist before. Further, this project will be the first to give policymakers appropriate tools to analyse market power, detect anti-competitive practices, and protect consumers in online markets. The new methods will also help digital platforms enhance their knowledge of users' incentives, which is crucial to designing platform rules that safeguard revenues and other key interests.