Estimating Models with Product Differentiation and Endogenous Product Characteristics

提案摘要在许多市场中，产品不是同质化的，而是差异化的。近年来，在分析和估计产品差异化模型方面取得了很大进展。差异化产品行业的完整模型包括：(1)消费者对产品不同特性的评价；(2)不同产品生产成本的描述；(3)生产者如何根据产品特性做出生产和定价决策的审查；(4)产品特性如何随时间推移而被选择的模型。该提案包含一些项目，这些项目扩展了现有工作，以便更好地对生产成本、产品空间和产品特性的内生性进行建模和估算。第一个项目开发了一个简单的模型，将工厂层面的成本数据与工厂生产的产品的特征联系起来，从而揭示了“享乐成本函数”。该模型将根据汽车业的人口普查数据进行评估。我们探讨了在对产品特性的差异进行调整后，面对不断变化的经济和监管环境，生产成本是如何发生变化的。该项目还侧重于生产过程的基本原理，如规模经济和边做边学。最后，新的成本方面的估计被纳入现有的需求方面的研究。第二个项目关注的是差异化产品模型中产品空间的正确性质。现有的计量经济模型隐含地假设产品空间的维度随着产品数量的变化而变化。这与许多保持产品空间维度固定的理论模型形成了鲜明对比。这项研究发展了用固定维积空间估计产品差异化模型的方法，并检验了这些估计量的性能。对产品特性的选择进行建模要困难得多。这个项目从非常简单的产品空间和进入这些空间的简单模型开始。这里的一个经验例子是无线电行业。长期以来，人们一直声称广播行业存在过度进入，因为新电台经常与现有电台的节目重复，因此抢走了客户。一系列越来越现实的产品空间模型与入门模型相结合，作为实现现有产品内生化的第一步。实证应用考察了针对广播进入的不同政策制度。最简单的进入模型使用的是横截面数据，实际上是一次性的“静态”进入模型。最后一组项目考察了进入差异化产品市场的完全动态模型。这些动态模型在存在沉没成本的情况下尤为重要。最近在动力学模型的理论和计算方面有了重要的进展。这些改进使计算和估计相对真实的状态空间的模型成为可能，例如无线电格式列表和不同的电台质量水平。这项研究完成了使用新可能的动态行业框架进行的首批实证研究之一；拟议的经验性应用再次适用于包括无线电在内的各种行业。

Proposal Summary In many markets, products are not homogeneous, but differentiated. Recently, there has been much progress in analyzing and estimating models of product differentiation. A complete model of a differentiated products industry would include (i) consumers' valuation of the differing characteristics of products, (ii) a description of the costs of producing different products, (iii) an examination of how, conditional on product characteristics, producers make production and pricing decisions and (iv) a model of how product characteristics are chosen over time. This proposal contains a number of projects that extend existing work to better model and estimate production costs, the product space and the endogeneity of product characteristics. The first project develops a simple model that relates plant-level cost data to the characteristics of products produced at the plant, thereby uncovering the `hedonic cost function`. The model is to be estimated on Census data for the automobile industry. We explore how, after adjusting for differences in product characteristics, production costs have changed in the face of a changing economic and regulatory environment. The project also focuses on the fundamentals of the production process, such as economies of scale and learning-by-doing. Finally, the new cost-side estimates are incorporated into existing demand-side studies. The second project focuses on the correct nature of the product space in a differentiated products model. Existing econometric models implicitly assume that the dimensionality of the product space changes with the number of products. This contrasts with many theoretical models that hold the dimension of the product space fixed. The proposed research develops methods for estimating models of product differentiation with fixed dimensional product spaces and examines the performance of these estimators. Modeling the choice of product characteristics is much more difficult. This project begins with very simple product spaces and simple models of entry into those spaces. One empirical example here is the radio industry. It has long been alleged that there is excess entry into the radio industry, because new stations often duplicate the programming of existing stations and therefore steal customers. A series of increasingly realistic models of the product space are combined with entry models as a first step toward endogenizing the set of existing products. The empirical applications examine different policy regimes toward broadcasting entry. The simplest entry models use cross-sectional data and are really one-shot `static` models of entry. The last set of projects examines fully dynamic models of entry into differentiated product markets. These dynamic models are especially important in the presence of sunk costs. There have been important recent advances in the theory and computation of dynamic models. These advances make it possible to compute and estimate the model for relatively realistic state spaces, such as a list of radio formats and different station quality levels. This research completes one of the first empirical studies using the newly possible dynamic industry frameworks;. The proposed empirical applications are, again, to a variety of industries including radio.