【作者】 陈林兴；

【导师】 黄祖辉；

【作者基本信息】 浙江大学 ， 农业经济管理， 2012， 博士

【摘要】 福利评价是公共政策分析的重要基础。福利评价如果出现问题,政府将无法对现有公共政策的效果作出客观评价,也无法对拟出台政策的预期效果作出正确预测,结果会导致公共资源的浪费,甚至还可能颠倒是非,导致严重的社会问题。然而,遗憾的是,现有福利评价(特别是福利不平等衡量)普遍存在两个明显的缺陷：(1)衡量标准问题。传统西方经济学对福利的衡量基本上是基于收入但是,近年来越来越多的学者指出把收入作为评价福利的标准存在各种缺陷,认为消费比收入更加直接地反映了福利状况,并建议用消费标准代替收入标准；(2)衡量指标问题。现有文献基本上采用均值、方差(或标准差)、基尼系数(Gini Coefficient)、泰尔系数(Theil index)或Atkinson指数等指标来衡量福利不平等状况,但是,这些指标往往导致研究得出模棱两可的结果。鉴于这些缺陷,有效的福利不平等衡量不但要调整评价标准,而且还要在方法上寻求突破。一个很自然的突破方向是直接研究消费的趋同性,即研究消费分配到底是趋同还是趋异,因为这比现有的离散性指标和不平等指标提供更加丰富的信息,从而更具一般性。这正是本文选题的基本出发点。另一方面,最近二十几年,收入趋同研究领域出现了许多新方法,因此,作为研究问题与研究方法交叉融合的结果,一个很自然的想法便是把收入趋同的研究方法运用于消费趋同研究。但是,即便是在收入趋同研究领域,通用的研究方法也存在许多局限。鉴于此,本文拟在现有收入趋同研究方法的基础上,采用更加新颖的研究方法,包括核条件密度、有限混合分布模型和分位数回归模型,并构建一个综合的方法体系,来研究消费趋同问题。此外,最近二十年,随着大众传播媒体的发展和人口流动性的提高,地区之间的交流日益频繁,导致地区间的联系越来越密切,因此,研究我国大陆31个省市区农村居民的消费趋同问题就不得不考虑地区之间消费行为的相互影响。这也正是Tobler的“地理学第一定律”所揭示的规律。实际上,从空间计量经济学的角度来讲,即便不存在这种实质性的相互影响关系,只要有空间自相关或空间异方差的存在,就必须考虑空间效应,否则结果是有偏的。因此,本文最终决定基于空间视角来研究我国省际农村居民的消费趋同性,以揭示我国农村居民消费支出分布的演变规律,以便为相关部门的决策提供参考。全文共计7章。第1章为绪论；第2章为文献综述；第3章为空间自相关检验,分别基于空间截面数据和空间计量经济学模型对我国农村居民消费支出的空间效应进行检验；第4章为基于核条件密度的探索性研究,分别从人均消费总支出和人均分类消费支出两个层面,采用核条件密度来研究我国省际农村居民消费支出的趋同性,以及空间因素和收入对消费趋同的影响；第5章为基于有限混合分布模型的证实性研究,分别从人均消费总支出和人均分类消费支出两个层面,采用有限混合高斯分布模型来研究消费趋同性,以及空间因素和收入对消费趋同的影响；第6章为基于分位数回归模型的解释性研究,分别基于总量消费模型和分类消费模型,对我国省际农村居民的消费趋同性进行统计意义上的正式解释,同时考虑了空间效应；第7章为结论和政策启示,在总结全文的基础上提出政策建议,并客观地指出本文存在的局限性。研究结果表明：(1)我国省际农村居民的各项消费支出总体上呈现双峰锁定分布,也可以称之为双峰俱乐部趋同；(2)我国省际农村居民的人均消费总支出、人均分类消费支出和人均纯收入都存在显著的空间自相关；(3)空间自相关对消费趋同产生一定程度的影响,但是,这种空间效应不足以改变消费趋同的整体性质；(4)人均纯收入是决定消费趋同的首要因素,基本上解释了消费趋同性,但不是唯一因素,收入和空间因素综合在一起几乎解释了所有的消费趋同信息；(5)农村居民人均消费总支出和人均分类消费支出之所以呈现双峰俱乐部趋同态势,是因为边际消费倾向不但不递减,反而是递增的,这不同于凯恩斯的边际消费倾向递减的假定,但是,这个表面上反常的研究结果正好解释了双峰俱乐部趋同；(6)农村居民各类消费支出的收入弹性呈非线性变化,这也不同于传统的假定。本文的研究结果具有一定的政策启示和实际应用价值：(1)在宏观层面,可以为国家制定转移支付政策、区域发展政策以及扩大内需政策提供参考；(2)在中观层面,可以为各级政府制定产业规划和调整产业结构提供参考；(3)在微观层面,可以为企业制定业务组合策略、市场预测、市场细分、目标市场选择以及新产品上市和市场拓展决策提供参考。更多还原

【Abstract】 Welfare assessment lays the foundations of public policy analysis. If something went wrong in this process, a government would not be able to measure the effects of current public policies effectively, nor would it be able to give an accurate prediction to a policy it is going to launch. As a result, public resources would be painfully wasted, and, moreover, right and wrong would be confused, leading to a serious social security problem.Unfortunately, two important things do usually go wrong in this process, especially when the measurement of welfare inequality is concerned:(1) Inappropriate measurement base. In the tradition of classic economics welfare was measured mainly based upon income. But in recent years more and more economists have pinpointed the defects of this type of measurement. They believed that welfare condition would be measured more direct by consumption than by income, and argued that income-based measurement should be replaced by consumption-based measurement. (2) Improper measurement indicators. In current economic literature, mean, variance (or standard deviation), Gini coefficient, Theil index and Atkinson index have been commonly used to indicate welfare inequality. But researches based upon these indicators tend to lead to confusing conclusions. In view of these defects, effective welfare measurement necessitates modification of measurement base as well as improvement in methodology. As a logical way of thinking, it would be more effective to study consumption convergence direct, since it can provide richer information than traditional way of measurement based upon discrepancy or inequality indicators, and thus represents a more general alternative to current procedures. This precisely constitutes the starting point of this dissertation.Meanwhile, in the past two decades, many new methods have found their way in the studies of income convergence. As a logical step forward, after interweaving practical problems and methodologies, it would be natural to apply these methodologies to the study of consumption convergence. But even in the field of income convergence study, there are still many limitations in terms of methodology. In view of this, based upon current methodology, this dissertation tries to apply some new methods to consumption convergence study in a comprehensive methodological framework composed by kernel conditional distribution, infinite mixture distribution model, and quantile regression.In the past two decades, with the rapid development of mass media and mobility of people between regions, inter-regional relationships have become increasingly closer, therefore, the interaction of consumption behavior should have to be taken into consideration in studying consumption convergence of rural residents across 31 provinces in China. And this is just the central idea of Tobler’s First Law of Geography. In fact, from an econometric perspective, even in the absence of substantive influence across regions, as long as spatial autocorrelation or spatial heteroscedasticity exists, it should have to be taken into consideration or else the results would be biased. Therefore, to avoid misleading results, this dissertation studies consumption convergence across 31 provinces from a spatial perspective, in order to uncover the rule behind the dynamics of consumption distribution and thus to provide some reference to related policy-making departments in China.The whole dissertation is composed of 7 chapters. Chapter 1 is the preface. In Chapter 2, literature in closely related fields is concisely reviewed. In Chapter 3, spatial autocorrelation is tested in terms of spatial cross-sectional data and spatial econometric models to determine whether there exists spatial effect in various expenditures. Chapter 4 represents an exploratory study in which, based upon kernel conditional density, consumption convergence is studied by means of stochastic kernel in terms of per capita total expenditure and classified expenditures respectively, and spatial and income effect are explored. Chapter 5 serves as a confirmatory study in which consumption convergence is tested by means of infinite mixture models in terms of per capita total expenditure and classified expenditures respectively, and spatial and income effect are explored. Chapter 6 is as an explanatory study in which consumption convergence is explained by means of quantile regression models in terms of per capita total expenditure and classified expenditures respectively, and spatial effect is also explored. Chapter 7 constituteis the conclusion in which a summary is made and limitations of this paper are pointed out.Results show that:(1)The various cross-provincial per capita expenditures are all distributed as two locked modes, indicating a bi-modal club convergence; (2)The various cross-provincial per capita expenditures and per capita net income of the 31 Chinese provinces show significant spatial autocorrelation; (3)Spatial autocorrelation influences consumption convergence in some degree, but is not strong enough to change the whole nature; (4)The per capita net income is the first important cause of consumption convergence, and can explain consumption convergence fairly well. But even though, it is not the sole cause.Spatial effect and income combined together can almost explain all the information about consumption convergence; (5)In contrast to the traditional assumption of decreasing MPC(marginal propensity to consume), the MPC of the Chinese rural residents is increasing. Interestingly, this seemingly unusual result explains the special bi-modal converging consumption pattern of the rural residents across the 31 Chinese provinces; (6)The income elasticities of various expenditures display a non-linear pattern, which is also different from traditional assumption.Research conclusions derived from this dissertation are of some value to both policy-makers and industrial practitioners at various levels:(1) At the macro level, they can provide reference to policy-making in terms of transfer programs, regional development as well as domestic demand expansion; (2)At the medium level, they can help governments at various levels in designing industrial development and adjustment plans; (3)At the micro level, they can aid companies of various size in designing business portfolio, predicting sales volumes, segmenting and targeting customers as well launching new products and expanding market coverage.更多还原