导　　师： 许宪春;许鹏

学科专业： 020208

授予学位： 博士

作　　者： ;

机构地区： 湖南大学

摘　　要： GDP作为国民经济核算的核心内容总量指标之一,其数据质量受到广泛的关注。从数据的匹配性角度研究GDP数据质量,构建基于匹配性的GDP数据质量评估方法体系,并对中国GDP的数据质量进行实证分析,有利于全方位认识GDP数据质量和更为有效地提出数据质量的控制方法。 从GDP数据质量广义内涵出发,将GDP数据的匹配性与GDP数据质量内涵结合起来,总结基于匹配性的GDP数据质量具体内容,即逻辑匹配性、方法匹配性和经济意义匹配性。根据数据类型的划分,利用对应数据类型的基本假设,将基于匹配性的GDP的数据质量内容转化为结构、时间和空间三个维度的数据质量评估,进而选择适应性的统计模型,构建基于匹配性的GDP数据质量评估方法体系。 GDP作为衡量产出的重要指标,其与经济系统中很多变量具有近似的比例关系,即结构关系,这些结构关系主要体现在投入产出结构、实物量结构和经济生产的内部结构三个方面。利用相适应的模型,对这三个方面的验证进而对中国GDP进行数据质量评估。评估结果显示,中国GDP数据从1981年至2010年,逻辑上存在较大缺陷,方法上并不是十分健全,经济结构匹配性上相对稳定。总体上,GDP数据质量在结构匹配性方面并不理想,存在较大的改进空间。 GDP数据在时间维度上具有显著的阶段性特征,这种阶段性特征体现在与经济运行的相关性和GDP数据内部/(即GDP和GDP增长率/)的相关性上。在GDP时间序列上的渐进稳定、GDP数据绝对数和相对数在时间维度上相关联等基本假设的基础上,构建GDP和GDP增长率的确定性和随机性时间序列模型,利用这些模型的误差率对GDP时间维度的数据质量进行评估从实证分析的结果来看,GDP数据在时间维度上基本匹配,对应各个内容上,时间维度GDP在经济意义的匹配性最高,而逻辑上匹配性次之,但两者均达到80/%以上,在方法匹配性上较低,为53.33/%。从时间维度年份可疑点来看,可疑年份主要体现在1991年以前的年份,用不同方法得到GDP数据基本上存在可疑特征的年份为1990、1991和1992及其附近的年份。 中国不同地区GDP之和与全国GDP之间的差额近年来差距越来越大,基于此,从空间角度分析中国GDP数据质量不能用各省份的总量数据去解释全国GDP数据质量高低。构建基于空间匹配性的GDP数据质量评估时,依赖于各地区结构化数据失真程度较低、地区汇总结构化数据与全国结构数据具有较强各联系、地区结构化数据体现在不同环节和综合效应三个基本假设,同时对这些假设利用典型相关分析进行验证。通过构建系统的向量自回归模型和分布滞后回归模型,利用误差率和综合评价方法,对中国GDP数据进行空间维度上评估。实证结果表明,GDP数据在空间维度上匹配性较差,对应各个匹配性内容上,空间维度GDP在逻辑匹配性最高,为76.08/%,而经济意义匹配性次之,仅为45.16/%,方法健全匹配性非常低,仅为29/%。从空间维度年份可疑点来看,可疑年份主要体现在1992年至1999年之间,用不同方法得到GDP数据基本上存在可疑特征的年份为1990、1991和1992及其附近的年份。 在基于结构、时间和空间三个维度的GDP数据质量评估的同时,从两个方面确定基于匹配性的GDP数据质量评估结果。一方面是以前述个章节评估的误差率为基础,利用层次分析法确定权重,使用综合评价方法,得到各个年份的综合误差率,作为GDP数据质量评估的结果。另一方面为了细化GDP数据质量高低的比较,利用数据变换的方式,对基于匹配性的GDP数据质量评估的误差率进行指数化计算,得到GDP数据质量评估的指数化结果,便于比较不同年份GDP数据质量高低。对综合评估结果的分析从总体特征、阶段性特征、相关维度的关联性、各个维度对GDP综合评估值的影响四个层面进行分析。通过总体特征的考察,发现中国GDP数据质量各个维度和总体的平均值大体处于7.5分左右,而结构匹配性维度得分最低,同时各个维度的变异程度较大,结构维度变异程度为最大。从阶段性特征来看,中国GDP数据质量尽管具有白噪声的特征,但依然存在明显的阶段性特征,这种阶段性特征与国民经济核算改革历程紧密相连。从各个维度的关联性进行分析,中国GDP数据质量的各个维度相关性较低,说明用模型对中国GDP数据进行质量评估时,各个维度的相关性较低,能够充分地考虑各个维度对总体数据质量的影响。从各个维度对GDP数据质量综合评估的影响来看,各个维度对GDP数据综合评估质量具有动态的特征,而且这种动态特征在1984年以前具有较强的差异性,而在1987年以后逐渐变为平坦。 针对GDP数据质量存在的问题,分析原因,并提出相应的控制体系,进而提出对策。在中国GDP数据质量不高的原因分析中,本文认为应归结为制度上的缺失、技术上的障碍和1978年以来社会的复杂变化三个主要原因。而针对这些原因,提高GDP数据质量应该从GDP数据运行环节,即GDP数据生产、发布和评估三个主要环节建立数据质量控制体系。要使GDP数据质量控制体系得以有效运行,需要抓紧数据管理人才培养、建立高效的国家统计信息网络体系、建立GDP数据质量组织保障体系。 GDP as one of the aggregate indicators of the core content of national accounts, the quality of its data has received widespread attention. This paper studies the quality of GDP data from the perspective of data matching, constructs a system of evaluation methods of GDP data quality based on matching, and does an empirical analysis for GDP data quality, which is conducive to all-round understanding of the quality of GDP data and to have a more effective presentation of data quality control methods. From the generalized connotation of the GDP data quality, this article combines the matching of the GDP data with the GDP data quality connotation, and summarizes the specific content based on matching, which are logic matching, method matching and the significance of economic matching. Dividing according to the data type and using the basic assumptions of the corresponding data type, this paper converts the quality of the GDP data, based on matching, into structure, time and space of three-dimensional data quality assessment, and then selects the adaptive statistical models to construct the GDP data quality assessment methodology based on matching. GDP, as an important indicator for the measurement of output, has an approximate proportional relationship with many variables in the economic system, namely structural relationship which is mainly reflected in three aspects-the input-output structure, the structure of the physical quantity, and the internal structure of the economic production. Through using the adaptive models to verify these three aspects and assessing China's GDP data quality, evaluation results show that China's GDP data from1981to2010has larger defects in logic matching; its methods are not very sound; GDP data is relatively stable in the economic structure matching. Overall, the GDP data quality is not ideal in structure dimension matching; there is greater room for improvement. GDP data in the time dimension has a significant phase characteristics which are reflected in the correlation of the economic operation and within GDP data-namely GDP and GDP growth. Based on the basic assumptions of asymptotically stability in GDP time series, correlation in the time dimension between the absolute and relative GDP data and so on, this article constructs the deterministic and stochastic time series models for GDP and GDP growth, then uses the error rate of these models to assess the quality of the GDP data in the time dimension. From the empirical analysis results, GDP data is basically matching in the time dimension; corresponding to each of the contents in the time dimension, GDP matches highest in the economic significance, the logic matching followed, but both are above80/%, and the method matching is low, at53.33/%. From doubtful year points in the time dimension, they mainly embody in the years before1991; with different methods, concerning GDP data, there exist basically suspicious characteristics in the year1990,1991and1992and around. There is a growing gap between the total GDP of different areas and the national GDP in recent years; based on this, from the spatial perspective while analysing of the quality of China's GDP data, the total amount of GDP data of all provinces cannot be used to explain the quality level of the national GDP data. When building GDP data quality assessment based on space matching, it can depend on three basic assumptions-the low level of distortion of structured data in various regions, the strong association between structured data of regional summary and national structural data, the reflection in the different segments and the combined effects of the regional structural data, at the meantime, uses the canonical correlation to verify these assumptions. This paper evaluates the China's GDP data from the spatial dimension by constructing the systematic vector autoregressive models and the distributed lag regression model and using the error rate and the comprehensive evaluation methods. The empirical results show that GDP data matching in the spatial dimension is poorer; corresponding to each of the contents, in the spatial dimension GDP matches highest in the logic matching, about76.08/%; the economic significance matching followed, only45.16/%; the sound methods matching is the lowest, only29/%. From doubtful year points in the spatial dimension, they mainly embody in the years between1991and1999; with different methods, concerning GDP data, there exist basically suspicious characteristics in the year1990,1991and1992and around. Based on the structure, time and space the three dimensions of GDP data quality assessment, meanwhile, this paper determined the GDP data quality evaluation results based on the matching from two aspects. On the one hand it takes error rate of the evaluation in the foregoing sections as the foundation, and then uses the analytical hierarchy process /(ahp/) to identify weight, and the comprehensive evaluation method for the comprehensive error rate of each year which is took as the results of GDP data quality evaluation. On the other hand, to refine GDP data quality high or low comparison, this article uses data transform, taking index calculation to error rate of the GDP data quality based on matching, and we get the result of indexation of GDP data quality evaluation to compare GDP data quality high or low in different years. The analysis of the results of comprehensive evaluation from four aspects are analyzed including overall characteristics, phasic characteristics, the relevance of relevant dimensions, the influence for the comprehensive value of GDP from each dimension. Through the investigation for overall characteristics, it finds that China's GDP data quality is7.5points or so at an average level in various dimensions and general and gets the lowest point in the structure dimension, and at the same time it varies largely among the dimensions, especially in the structural dimension with the biggest variation. From the feature of stage, although China's GDP data quality has the characteristics of white noise, it still exists obvious gradual characteristics; this gradual characteristic and national economic accounting reform processes are closely linked. From the analysis of the correlation among the dimensions, China's GDP data quality is with a low correlation in each dimension, and this indicates that when using models to evaluate China's GDP data quality, it has a low correlation among dimensions, so it is able to fully consider the influence from each dimension to the overall data quality. From the influence for the comprehensive evaluation of GDP from each dimension, various dimensions have a dynamic characteristic of comprehensive evaluation quality of the GDP data, and the dynamic characteristic enjoys a strong difference before1984, and becomes flat gradually after1987. For GDP data quality problems, this paper analyzes the causes, and proposes appropriate control systems and countermeasures. When analyzing the reasons for which China's GDP data is not of high quality, this article thinks it should come down to three main causes including institutional deficiency, technical barriers, and social complex changes since1978. For these reasons, improving GDP data quality should be from the GDP data running link, that is, GDP data production, dissemination and evaluation of three main links to establish data quality control system. To make the quality control system for the GDP data run effectively, we need to seize the training for persons with the data management ability, construct the efficient national statistical information network system, and establish the GDP data quality organization system.

关 键 词： 数据质量 匹配性 数据质量评估 数据质量控制

分 类 号： [F224 F124]

领　　域： [经济管理] [经济管理]