交互效应面板数据模型的方法论及应用研究

发布时间：2018-06-15 19:56

本文选题：交互效应 + 非线性平滑转换　；参考：《华中科技大学》2014年博士论文

【摘要】：传统面板数据模型通常以可加形式引入个体效应和时间效应,来反映个体异质性与不随个体变化的时间效应。Bai(2009)提出了交互效应的面板数据模型(IEPDM),在传统面板数据模型中引入了个体效应与时间效应的交互项,来揭示共同因素对不同个体的效应差异性。近几年来,带交互效应的面板数据模型备受关注,其理论研究得到了深入发展。但现有的交互效应面板数据模型的研究都是基于二维面板数据结构,单方程的线性模型。基于二维面板数据结构模型无法研究区域间的劳动力流动、资本流动和双边贸易等实际案例；基于线性模型不能反映现实经济中变量之间的非线性关系；基于单方程的交互效应面板数据模型无法考察变量间的反馈机制、动态冲击效应。所以,在全面系统地分析现有交互效应面板数据模型的优点基础上,本文的研究内容分为两个层次：一是模型方法论研究,二是方法论成果对中国现实问题的应用。方法论研究：第一,鉴于线性模型无法考察经济变量之间的非线性关系,本文将非线性平滑转换引入到交互效应的面板数据模型中,建立了非线性交互效应面板数据模型(NIEPDM)。对静态模型提出了非线性迭代OLS(NIOLS)估计；对动态模型构建了非线性迭代GMM(NIGMM)估计,两类估计量均具有一致性。蒙特卡罗仿真结果显示估计量的有限样本性质良好。第二,鉴于交互效应的单方程面板数据模型不能反映变量间的反馈机制,无法测度内生变量的冲击效应,本文将交互效应扩展到了多方程的面板结构VAR(SVAR)领域,建立了带交互效应的面板结构VAR(IEPSVAR)模型,并针对IEPSVAR模型给出了详细的参数估计方法和估计程序。第三,本文在传统三维面板数据模型中引入了交互效应,建立了带交互效应的三维面板数据模型,并对静态模型提出了迭代OLS(IOLS)估计方法；对动态模型提出了迭代GMM(IGMM)估计方法。蒙特卡罗仿真实验结果表明,有限样本性质良好。方法论成果对中国现实经济问题的应用研究：为了对本文所介绍理论模型提供较完整的应用案例,本文分别对人口结构、信贷规模与房地产价格的趋势调整；财政支出的自激励机制、溢出效应与区域不平衡性；经济地理分割与城乡间产业转移三个问题进行了实证研究。第一,对人口结构、信贷规模与房地产价格的趋势调整研究是基于非线性交互效应动态面板数据模型,借以评判中国社会经济转轨形态中人口结构对房价长期趋势的影响以及货币工具在房价调控中的有效性和适用性。实证分析显示：少年抚养比下降将对中国房价产生长期的抑制作用。老年抚养比上升对东部房价的抑制性开始显现,但对中部暂时具有推动作用。伴随着信贷扩张,东部房地产市场投机和泡沫化特征在2004-2007年间开始凸显；2009年以后,投机性需求向西部地区转移；中部投机性需求特征较不明显。东部房地产价格对利率调整最敏感,具有近似单元弹性,西部地区的弹性约0.5左右,中部地区最不敏感。第二,对财政支出的自激励机制、溢出效应与区域不平衡性研究是基于交互效应的面板SVAR模型,系统分析财政支出对地区经济运行效率的动态自激励机制、溢出效应以及对区域平衡的调节功效。实证分析结果显示：地方财政支出对经济效率存在显著的自激励效应和溢出效应,但是与经验判断不同,中西部地区财政支出的溢出效应明显强于东部地区。中央财政支出对经济效率的激励效应也存在显著的地区差异,自东向西逐渐弱化。因而,增强欠发达地区自主的财政支出能力,是提高经济运行效率、促进区域平衡发展的关键。第三,对经济地理分割与城乡间产业转移研究也是基于交互效应的面板SVAR模型。从经济地理两方面构建了城乡产业转移的理论框架和实证分析模型。其研究结果表明：在影响城乡产业转移的诸多因素中,产业基础是最重要的决定因素。资本流动是最主要的引导力量,比劳动力流动更具导向性。地理差距对城乡产业转移的分割效应明显大于收入差距。样本期内,社会政策环境因素更有利于中西部地区城乡间产业转移,对经济发达地区效应不明显甚至有阻滞作用。
[Abstract]:The traditional panel data model usually introduces the individual effect and time effect in addition form to reflect the individual heterogeneity and the time effect that does not change with the individual.Bai (2009). The panel data model (IEPDM) of interaction effect is proposed. In the traditional panel data model, the interaction of the body effect and the time effect is introduced to reveal the common factors. In recent years, the panel data model with interaction effect has attracted much attention, and its theoretical research has been developed deeply. However, the existing interactive panel data models are based on the two-dimensional panel data structure and linear model of single equation. The base Yu Erwei panel data structure model can not be studied in the area. The actual cases of labor flow, capital flow and bilateral trade between regions; the linear model can not reflect the nonlinear relationship between the variables in the real economy, and the interaction effect panel data model based on single equation can not examine the feedback mechanism and dynamic impact effect between variables. Therefore, the existing interaction effect is systematically analyzed. On the basis of the advantages of the panel data model, the research content of this paper is divided into two levels: one is model methodology research, and the two is the application of methodology results to Chinese realistic problems.
Methodological research: first, in view of the inability of linear model to investigate the nonlinear relationship between economic variables, the nonlinear smooth transition is introduced into the panel data model of interactive effect, and the nonlinear interaction effect panel data model (NIEPDM) is established. The nonlinear iterative OLS (NIOLS) estimation is proposed for the static model, and the dynamic model is used for the dynamic model. The nonlinear iterative GMM (NIGMM) estimation is constructed, and the two kinds of estimators are consistent. Monte Carlo simulation results show that the finite sample property of the estimator is good.
Second, in view of the interaction effect single equation panel data model can not reflect the feedback mechanism between variables and can't measure the impact effect of endogenous variables, this paper extends the interaction effect to the panel structure VAR (SVAR) domain of multiple equations, and establishes a panel structure VAR (IEPSVAR) model with interaction effect, and gives a detailed description of the IEPSVAR model. A fine parameter estimation method and an estimator.
Third, in this paper, the interaction effect is introduced in the traditional 3D panel data model. A three-dimensional panel data model with interaction effects is established, and an iterative OLS (IOLS) estimation method is proposed for the static model. The iterative GMM (IGMM) estimation method is proposed for the dynamic model. The Mont Carlo simulation results show that the finite sample is of good properties.
The application of methodology results to China's real economic problems: in order to provide a more complete application case for the theoretical model introduced in this paper, the trend adjustment of population structure, credit scale and real estate price, self incentive mechanism of fiscal expenditure, spillover effect and regional imbalance; economic geography segmentation and urban and rural areas The three problems of inter industrial transfer are studied.
First, the research on the trend adjustment of population structure, credit scale and real estate price is based on the dynamic panel data model of nonlinear interaction effect, which is used to judge the effect of population structure on the long-term trend of housing price in China's social and economic transition form and the effectiveness and applicability of monetary instruments in house price regulation and control. The decline in the juvenile dependency ratio will have a long-term inhibitory effect on house prices in China. The inhibition of housing prices in the eastern part of the upbringing ratio began to appear, but it has a temporary effect on the central region. With the credit expansion, the characteristics of speculation and foam in the Eastern real estate market began to highlight in the 2004-2007 years; after 2009, speculative demand The western region transfer; the characteristics of the central speculative demand are not obvious. The Eastern real estate price is most sensitive to the interest rate adjustment, with approximate unit elasticity, about 0.5 in the western region, and the most insensitive in the central region.
Second, the research on the self incentive mechanism of fiscal expenditure, the study of spillover effect and regional imbalance is a panel SVAR model based on the interaction effect, and systematically analyzes the dynamic self incentive mechanism of fiscal expenditure on regional economic efficiency, the spillover effect and the regulation effect on the regional balance. There is significant self incentive effect and spillover effect in economic efficiency, but different from experience, the spillover effect of fiscal expenditure in the central and western regions is obviously stronger than that in the eastern region. The incentive effect of central fiscal expenditure on economic efficiency also has significant regional differences, which gradually weaken from east to west. Thus, the independent finance in the underdeveloped areas is strengthened. The ability to pay is the key to improving economic efficiency and promoting balanced regional development.
Third, the study of economic geography division and urban industrial transfer is also a panel SVAR model based on interaction effect. The theoretical framework and empirical analysis model of urban and rural industrial transfer are constructed from two aspects of economic geography. The results show that among the factors affecting the industrial transfer of urban and rural industries, the industrial base is the most important determinant. Capital flow is the main guiding force, which is more guiding than labor flow. The segmentation effect of geographical gap on urban and rural industrial transfer is obviously greater than income gap. In the sample period, social policy and environmental factors are more conducive to the transfer of urban and rural industry in the middle and western regions, and have no obvious or even block effect on the economic developed areas.
【学位授予单位】：华中科技大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：F224;F124

【参考文献】