空间hedonic模型理论建模技术的研究.doc

摘  要  定量分析方法很多都是采用的对分析对象直接建模的方式。一般而言，个体表 现会被定为被解释变量，并被一组解释变量表示成一个方程的形式。但是在很多情 况下而言，个体的表现并不会完全取决于个体特征，而是取决于社会结构以及个体 之间的相互作用。这种现象其实在其他的很多领域也同样适用，例如房地产领域。 对于房地产定价而言，最常见的方式莫过于 hedonic 价值方程，也就是特征向量法。 如果房地产资产样本是相互关联的，那么单由房地产的属性和特征对房地产进行定 价将是有偏误的，因为它忽略了周边房地产的价格对其的影响。 对房地产受到周边的影响进行建模的技术即为空间计量的主要研究领域。值得 注意的是，已经有很多实证研究证明，样本可能表现出一定空间上的相关性。本文 主要对四种空间效应进行分析：空间依赖性、空间异质性、空间外溢性以及呈现出 空间依赖的自变量遗漏问题进行建模的技术，主要集中于空间滞后模型，空间误差 模型、广义空间模型和空间杜宾模型在这些空间效应存在的情况下进行建模的技术 的讨论和比较。 本文的工作在于，通过理论推导分析了不同空间模型对不同类型空间效应进行 建模的方式的区别及其限制条件，并对每个模型的建模特征在理论上进行了分析和 比较，同时也讨论了这些模型在实践中的适用程度。依据此研究思路以及本文研究 结论，本文认为对于以上四种空间效应而言，空间杜宾模型的适用范围最广且限制 条件最少，与其他三种模型相比其优势最为明显。 关键词：空间 hedonic 模型 空间依赖性 空间异质性 空间外部性 变量遗漏 I Abstract It is the common methodology for quantitative methods to build models directly on studying target. Generally, the targets are defined as the explanatory variables and interpreted by a variety of independent variables with an equation. However, sometimes the target not only depends on its attributes, but is also influenced by its neighbors. This phenomenon is applicable in a wide range of areas, and the real estate market is not excluded. The most prevailing approach to value properties is to use the hedonic pricing model, which divide the house price into its attributes and regard the whole price of property the sum of the values of them. But if the prices of the properties are connected with each other, merely interpreting the house prices by its attributes would lead to biases, as it overlooks the influence exerted by the adjacent house prices. To quantify the influences from the near observations are the main focus of spatial econometrics. Worth to be concentrated is that it has been widely proved by empirical evidences that data samples are likely to present some sort of connections among its neighborhoods, which we name it spatial effects altogether. In this article, we focus on four kinds of spatial effects: the spatial dependence, spatial heterogeneity, spatial externality and the omission of sp