【作者】 郭国强；

【导师】 王少平；

【作者基本信息】 华中科技大学 ， 数量经济学， 2013， 博士

【摘要】 空间计量经济学是研究空间地理分布对经济现象的影响的理论,从空间理论被引入到经济理论的研究中,空间计量理论逐渐快速发展,迅速形成计量经济学的一种重要的现代分支,产生了许多重要的理论和实践成果。空间计量理论通过空间加权矩阵将空间因素对经济现象的影响的模式引入到模型中,空间因素的加入为模型估计和检验带来了新问题,学者们应用各种计量理论从不同的角度出发,采用不同的方法来解决这些新问题。本文的目标是理解和学习基础的空间计量理论和模型,估计方法,重点研究动态空间计量理论和模型,然后应用到解释我国经济问题中。本文首先回顾了静态空间计量理论模型,包括空间自回归模型、空间误差模型、空间德宾模型、一般模型和定性数据空间计量模型的结构特征,空间加权矩阵的定义和设定,极大似然方法估计模型的原理,以及计算机上实现估计过程的步骤。本文理论研究的重点是动态空间计量理论模型,首先分析一个内生变量的动态空间计量模型的种类和分类,特别是时间-空间动态类型的空间计量模型的具体结构设定。动态空间计量模型的特点是模型右边包含被解释变量的滞后期,由于是面板数据,所以还存在时间效应和固定效应,以及面板数据所固有的内生性问题,针对这些问题,本文分析了拟极大似然方法估计模型的原理和过程,估计量的性质。针对在实际应用中经常会遇到多个内生变量的情形,本文引入了另一种多个内生变量的动态空间计量模型,即空间面板VAR模型,这种模型将所有的变量都认定为内生的,解释变量除了自身的滞后时期,还包括空间加权矩阵表示的空间因素项。在估计这类模型的时候所要解决的最主要问题是内生性问题,本文采用估计普通面板VAR模型时采用的GMM方法,采用GMM方法来估计空间面板VAR模型,具体分析了GMM估计空间面板VAR模型的原理和过程,并在最后通过计算机编程实现模型的估计。在此基础上,我们从理论上揭示了空间面板VAR模型与普通面板VAR模型的差异,由此体现了本文的理论创新。我们的创新体现在：普通面板VAR模型中,在做脉冲响应的时候,冲击源全部来自于内生解释变量,并且其响应结果对所有的横截面都是一样的,但是在空间面板VAR模型中,冲击源除了内生变量之外,冲击源发生在不同的地区会导致冲击的效果不一样,因为冲击会通过空间加权矩阵发生不同横街面之间的相互传递和扩散,而由于每个横截面空间分布上的差异,所以最终导致同一个内生变量的冲击,当起源地发生在不同的横截面时,带来的冲击效果也不一样。本文应用上的创新是结合上面的两种动态空间计量模型：一个内生变量的时间-空间动态类型空间计量模型和空间面板VAR模型,将其应用到解释我国实际经济现象中。针对我国的数据特征和背景,本文建立了我国30个省市猪肉价格波动的动态空间计量模型,在解释价格波动的因素中加入空间影响因素,估计和检验结果揭示了空间因素对价格波动的影响程度在0.85到0.90单位之间；基于我国30个省市地区不同的物价指标,建立了30个省市居民消费价格、食品消费价格、商品零售价格、食品零售价格、工业品出厂价格、原材料,燃料和动力购进价格六个变量之间的空间面板VAR模型,研究30个省市价格波动之间的相互影响关系。其主要结论为不同省市的不同类型价格会通过空间从一个市场传递到另一个市场,相互造成影响,同一类型的价格冲击当发生在不同的省市的时候,由于不同省市的空间分布不同,所造成的影响也会不同,这一结论从文献的角度来看,第一次基于空间面板VAR模型,反映了我国不同省市之间不同物价指标之间相互影响的关系和特征。总之,本文在详细解读空间计量模型的基础上,从理论上研究了一个和多个内生变量的动态空间计量模型,体现理论创新意义的研究为空间面板VAR模型。针对我国的背景和现实特征,本文建立了30个省市猪肉价格波动的动态空间计量模型和不同物价指标的空间面板VAR模型,估计了空间因素对猪肉价格波动影响程度,考察了不同地区物价之间相互影响的关系和特征,其意义在于揭示出空间分布对经济现象的影响规律,提出了显著的应用性创新。更多还原

【Abstract】 Spatial econometrics is the study of the theory of impact of space geographic distribution on economic phenomena, from space theory was introduced, spatial econometric theory has been rapidly developed, and quickly formed an important modern branch of econometrics, obtained a number of important theoretical and practical results. Through the spatial weighting matrix, the impact of space factors on the economic phenomena is introduced into the model, some estimation and testing problems yield because of this, from different point of view, scholars have used a lot of different methods to solve these new problems. The goal of this paper is to understand basic theories and models of spatial econometric, model estimation methods, at last focusing on dynamic spatial econometric theories and models, then applied to the interpretation of the economic problems in our country.This paper first reviews the static spatial econometric theories and models, including: the spatial autoregressive model, spatial error model, spatial Durbin model, general spatial model, and spatial model using qualitative data, significance of spatial weighting matrix and its structure, the principle of maximum likelihood estimation method, as well as how to achieve this estimation process steps on the computer.The focus of the theoretical study of this article is dynamic spatial econometric models. We first study the type and classification of dynamic spatial econometric model with one endogeneous variable, especially the structure of the time-space dynamic spatial model. The feature of dynamic spatial econometric model is that the right side of the model contains explanatory variables lag term, and because of the panel data, time effects and fixed-effects also exist, and panel data inherent endogenous. To solve these problems, we study Quasi-Maximum Likelihood estimation method, its principles and process, as well as the nature of the estimators.In practice, we often encounter situations of multiple endogenous variables, so we introduce another spatial dynamic panel data model, is that the spatial panel VAR model. The variables in this model are all endogenous, the explanatory variables include its own lag period term, also contain the spatial effect that expressed by spatial weighted matrix. The most important problem about the estimation of such model is endogeneity, so this paper introduce GMM method to solve this problem, and this method is also used in the estimation of ordinary panel VAR model. We use the same GMM method to estimate the spatial panel VAR model, study its principles and processes, and finally achieve this process through computer programming. On this bais, we theoretically reveal the differences of spatial panel VAR model and panel VAR model, which reflects the throretical innovation. Our innovation embodies here, When we do the impulse response, all the source of impulse in ordinary panel VAR model come from the endogenous explanatory variables, and its response is same across all the cross sectionals. But in spatial panel VAR model, in addition to the endogenous variables, when the source of impulse happens in different regions, the response will be different. Because the shock will transfer and spread among these cross sections through the spatial weighting matrix, and the spatial distribution of every cross section is different, the impulse happens in different cross section will cause different effect.The practical innovation of this paper is using the two kinds of dynamic spatial econometric models to explain China’s actual economics phenomenon. Contrary to Chinese data characteristics and background, we construct a dynamic spatial econometric model about the fluctuations of pork price in China’s30provinces, we join spatial factors to explain the fluctuations of pork price. The results of the model show that the spatial effect on pork price is about0.85to0.90units. We also construct a spatial panel VAR model of total price level of China’s30provinces and cities, the model contain six endogeneous variables, these are consumer prices, food consumer prices, retail prices of commodities, food retail prices, producers’ prices for manufactured goods, raw materials, fuel and power purchase prices, we use this model to study the relationship of the interaction of the price volatility among the30provinces and cities. Its main conclusions is that the price in different province can be affect, from one market to another, even the same type of impulse but happen in different province, the response will also be different. From the point of view of literature, this conclusion is the first time to reflect the relationships and characteristics of the interactions among the different price indifferent provinces, basing on the spatial panel VAR model. In short, based on the detailed interpretation of spatial econometrics model, this article study the dynamic spatial econometric model with one and more than one endogeneous variables, and the spatial panel VAR model is the theoretical innovation. For our background and practical characteristics, this paper established a spatial dynamic model of the fluctuations of pork price of30provinces, and a spatial panel VAR model of different price indicators, estimated the impact of spatial effect on pork price, investigate the mutual interaction relationship and characteristics between different price in different province, the significance is that it reveals the effect of spatial distribution on economic phenomena, and present significant application innovation.更多还原