ITR/PE (SES): Empirical Studies of Network Effects

当一个代理基于有多少其他代理使用这种商品来评价一种商品时，就存在网络效应。例如，在VHS和Beta VCR之间进行选择的消费者不太可能根据技术性能的优劣来选择。相反，消费者选择其他消费者选择的标准，很可能是为了能够访问大量预先录制的电影库。网络效应存在于许多市场，对许多高科技和信息技术产品尤其重要。网络产品可能导致给公共政策带来难题的市场结果。这个项目研究了两个这样的结果：第一个是市场可以在单一的专有商品上进行协调，从而导致该商品所有者的市场力量。第二，市场可能无法协调，因此消费者不确定最终将采用哪种标准。这种混乱会导致新产品的采用率非常低。本项目分析了表现出这些问题的两个具体行业。该项目通过在面对网络效应时出现的估计问题进行工作，并使用估计结果来分析政府干预这些市场的影响，无论是通过反垄断还是通过标准制定。项目的第一部分估计了黄页广告市场中积极反馈循环的重要性。我估计有多少消费者更喜欢有更多广告的目录，有多少广告商愿意在有更多读者的目录上做广告。我使用所得的估计来考虑市场是受益于垄断（利用网络效应）还是寡头垄断（减少市场力量）。初步估计表明，更有竞争力的市场是可取的。这一结果支持了最近促进目录市场竞争的政策。此外，这里开发的技术可以有效地应用于其他市场。例如，正如消费者和广告商更喜欢在单一目录上进行协调一样，消费者和软件开发人员更喜欢在单一操作系统上进行协调。我从黄页市场得到的结果的一个缺点是，这些估计预测了从向市场添加目录中获得的不合理的高收益。这个问题在使用离散选择模型时很常见。该项目的第二部分（与加州大学洛杉矶分校的Daniel Ackerberg合作）提出了一种新的估计方法，该方法捕捉到这样一个事实：当产品进入市场时，产品在研究人员无法观察到的特征方面变得更加相似。我们的研究提出了估计器的结构论证，以及何时估计器最合适的蒙特卡罗研究。这个估算器有很多应用，将这个技术应用到黄页市场应该会导致更准确的估计，无论是网络效应的强度，还是竞争和标准化之间的福利权衡。该项目的第三部分是与安吉利克·奥格罗（麦肯锡）和谢恩·格林斯坦（凯洛格）共同工作。我们将分析高速调制解调器的引入以及市场如何未能在单一标准上进行协调。最初，56K调制解调器的两个相互竞争但相当的标准导致了消费者和互联网服务提供商不得不协调调制解调器技术选择的局面。总的来说，采用非常缓慢，最终，一个行业组织开发了一个被广泛采用的单一标准。我们对缓慢的初始采用率的解释是，与协调的动机相反，ISP也有在两个标准之间区分的动机，因为这样做可以减少价格竞争。消费者随后会推迟采用，因为他们不知道哪种标准会“胜出”。我们开发了一个理论模型来形式化这些点。此外，初步的实证分析表明，在新标准出台之前，即使在单个市场内，采用ISP的选择在两种标准之间分布均匀。正如我们的理论所预测的，ISP是完全分化的，而不是协调的。我们使用来自该市场研究的见解和结果来讨论采用选择的竞争性决定以及干预标准战争的好处的含义。

A network effect exists when an agent values a good based on how many other agents use the good. For instance, consumers choosing between VHS and Beta VCR's are unlikely to choose based on superior technical performance. Instead, consumers choose the standard that other consumers choose, most likely in order to have access to a large library of pre-recorded movies. Network effects are present in many markets and are especially important for many high technology and information technology goods. Network goods can lead to market outcomes that present difficult problems for public policy. This project studies two such outcomes: The first is that a market can coordinate on a single proprietary good, leading to market power for the owner of the good. The second is that markets can fail to coordinate, so consumers are unsure of which standard will ultimately be adopted. Such confusion can lead to very slow adoption rates for new products. This project analyzes two specific industries that exhibit these problems. The project works through estimation issues that arise in the face of network effects and uses the results of the estimation to analyze the implications for government intervention into those markets, either by antitrust or standard setting. The first part of the project estimates the importance of a positive feedback loop in the market for Yellow Pages advertising. I estimate how much consumers prefer directories with more advertising and how much advertisers are willing to pay to advertise in a directory with more readers. I use resulting estimates to consider whether the market benefits from monopoly (which takes advantage of network effects) or oligopoly (which reduces market power). The preliminary estimates imply that a more competitive market is preferable. This result supports recent policy promoting competition in the directory market. Furthermore, techniques developed here could be usefully applied to other markets. For instance, just as consumers and advertisers prefer to coordinate on a single directory, consumers and software developers prefer to coordinate on a single operating system. A drawback of my results from the Yellow Pages market is that the estimates predict unreasonably high benefits from adding directories to the market. This problem is a common one when using discrete choice models. The second part of this project (which is joint work with Daniel Ackerberg, UCLA) proposes a new estimator that captures the fact that as products enter a market, products become more similar with regards to characteristics that are unobservable to the researcher. Our research presents structural justifications of the estimator as well as Monte Carlo studies of when the estimator is most appropriate. This estimator has many applications, and the application of this technique to the Yellow Pages market should lead to more accurate estimates, both of the strength of the network effect and of the welfare tradeoffs between competition and standardization.The third part of the project is joint work with Angelique Augereau (McKinsey) and Shane Greenstein (Kellogg). We analyze the introduction of high-speed modems and how the market failed to coordinate on a single standard. Originally, two competing but equivalent standards of 56K modems led to a situation in which consumers and Internet service providers had to coordinate their choice of modem technology. Overall, adoption was very slow and eventually, an industry group developed a single standard that was widely adopted. Our explanation of the slow initial adoption rate is that, in contrast to the incentives to coordinate, ISP's also have an incentive to differentiate across the two standards because doing so reduces price competition. Consumers then postpone adoption because they do not know which standard will "win". We develop a theoretical model that formalizes these points. In addition, preliminary empirical analysis suggests that, before the arrival of the new standard, adopting ISP choices were evenly distributed between the two standards even within individual markets. As our theory predicts, ISP's perfectly differentiated instead of coordinated. We use insights and results from the study of this market to discuss the competitive determination of adoption choice and implications for the benefits of intervention into a standards war.