分税制改革以后，我国地方财政支出运行中的一个突出问题是财政收入权利和支出义务的不匹配，地方财力困难。为尽可能保证有限的公共资源在各个支出项目间合理配置，地方政府就必须考虑财政支出效率问题。本研究以我国省级地方财政支出效率为研究对象，主要探讨了财政支出效率的内涵、财政支出配置效率、财政支出技术效率以及地方财政支出效率的空间互动性等问题，研究方法和结论为财政支出效率的进一步研究提供了一个参考。

经济学讨论的主题是资源的帕累托效率配置，财政支出效率也是基于帕累托效率进行定义的。财政支出效率是指财政支出以最“经济”的方式来满足社会成员的各种公共需求，即财政支出对公共偏好的“有效回应”或“匹配”。财政支出效率至少包括两个方面：财政支出配置效率和财政支出技术效率。财政支出配置效率是财政支出与居民的公共需求偏好的满足，是从财政支出的最终结果上考察支出效率的。财政支出技术效率是指以最少的财政投入生产出最多的公共商品，可以理解为在一定的公共商品或服务产出量下所投入的财政支出最小，或在一定的财政支出情形下所生产出最多的公共商品或服务，它们被称为投入导向型和产出导向型。另外，现行财政体制安排的一个重要结果就是地方政府间的横向竞争和互动，财政上的横向互动表现为支出和税收，因此若把财政收入考虑到财政支出的竞争中，那么地方间财政支出效率互动和影响就必须加以研究。

本研究以中国31个省（自治区、直辖市）的教育、医疗卫生、社会保障和交通运输四类公共商品为对象进行了财政支出实证研究。

在民主社会，财政支出是通过公共选择而作出安排的，公共选择的结果往往取决于“中间投票人”的需求偏好，因此，财政支出配置效率评估的基本思路是，估计出地方“中间投票人”对公共财政支出的需求，进而比较财政支出实际数与中间投票人的公共支出需求数，如果在一个给定的范围内，则认为财政支出的配置有效；反之，若超出该范围，则认为财政支出的配置无（低）效。这里需要研究的是如何确定合理的“范围”，本研究认为有两个思路：一种是利用离散选择模型中的Ordered Logit模型基本原理，估计出区间值；另一种是利用离散选择模型中的Multinomial Logit模型的原理，事先确定范围值。本研究中认为，地方政府选择是否配置效率不是一个排序模型，因此采用了Multinomial Logit模型，确定的范围是以均方根误（RMSE）作为判断配置是否有效率的标准。利用Multinomial Logit模型的估计结果发现，我国地方公共商品支出的配置效率在2007—2012年经历了先扬后抑的变化状态，2009年实现配置效率的省（自治区、直辖市）最低，约仅有一半的省（自治区、直辖市）达到配置效率。主要的外生变量中地方财政集中的公共资源的多寡程度，是决定公共商品财政支出配置效率的关键。

公共资源具有稀缺性，因此在使用过程中，必须遵循以最少的投入生产出最多的产出（Do more with less）的原则。财政支出技术效率的评价存在困难，主要在于难以确定与财政投入相对应的公共商品或服务产出。因为本书的研究对象基本上都是准公共商品的支出，其成本弥补途径既有公共部门资金也有私人部门资金，从而把公共支出的产出进行分离难度较大。对此，本研究的解决手段是采用了数据包络分析法，该方法是通过构建效率前沿面来评估决策单位的相对效率，这就较大程度上避开了必须准确确定财政支出对应的产出指标的困境。在运用数据包络分析方法评估地方财政支出效率时，遇到的问题是，效率值受到来自内生因素和外生因素的影响，内生因素如财政内部管理制度、外生环境变量如运气和随机因素的冲击，为过滤外生环境因素的影响，本研究采用了三阶段Bootstrapped DEA方法进行了实证评价。另外，为考察地方财政支出效率的动态变化，以及引起变化的原因，本研究采用了Malmquist指数方法。采用投入导向型且控制了环境变量和随机冲击后，DEA计算财政支出效率得分表明，我国地方不同的财政支出在2007年至2012年间，存在11%至40%不等投入浪费。整体而言，效率呈现上升态势，东部地区的技术效率比中西部的技术效率高。另外，大部分地方公共商品支出的规模报酬处于递增状态，这也反映出了我国地方公共商品，如教育、公共卫生等的投入量有待进一步提高。

我国现行分税制财政体制是一个促进地区间横向竞争的体制，追求科学和谐全面发展的目标，在现行地方官员政绩考核机制和财政体制安排下，退化为追求本地GDP 最大化的目标。地方政府为实现本地区经济增长，运用财政支出和税收手段进行跨区竞争，在平衡预算约束下，财政支出效率（主要是技术效率）间的竞争将不可避免，因此研究地方财政支出效率就必须研究地方财政支出效率的跨地区互动。本研究首先用Moran's I考察地区财政支出效率地域集聚性，结果表明各个类别的公共支出效率具有空间相关性。为进一步考察地方间财政支出效率空间特征，本研究借助空间计量模型——空间滞后模型、空间误差模型和空间Durbin模型来刻画。研究结果表明，教育、医疗卫生、社会保障、交通运输财政支出效率在空间上存在集聚现象，财政支出效率高的地区集中在一起，财政支出效率低的地区也较为集中。继而利用空间面板Durbin模型实证研究影响地方政府财政支出效率的因素，研究结果表明：不仅相邻地方的财政支出效率具有空间外溢效应，同时一些外生解释变量除了对本地区的财政支出效率有影响外，还对其他地区的财政支出效率有空间溢出效应。

本研究还存在一些尚未解决的问题。第一，以部门为效率评价对象需要更多讨论。本研究对我国地方财政支出效率的定量分析进行了尝试，评价的对象是特定的公共部门，如教育、医疗卫生等部门。实际上，以部门为评价对象存在一个重要的问题是，未能将部门间效率的流动性考虑进来，以交通运输支出效率为例，除了本研究考虑的一些外生影响因素，还有一个部门间效率互动的问题，交通运输支出效率不高或许不是因为本部门的原因，而是由于其他部门的效率低下造成的。更有可能的是，某个部门财政支出效率排名靠后或许是财政支出整体效率的帕累托效率改进。第二，本研究对效率评价的结果应用只给出了较为直接的政策启示，未能做出深入研究。财政支出效率评价对于了解公共职能部门公共资金配置和落实状况，有着重要的实践意义，财政以及具体职能部门对支出效率评价结果如何解读，以及如何制定政策需要深入探讨。以交通运输支出效率为例，在以投入导向型效率评价方法下知道地方交通运输财政支出效率低下，存在大量的投入浪费，对于该结果，是否应该对交通运输部门的财政支持减少？如果以产出导向型评价方法下知道效率低下，是否应该将产出目标降低？甚至不设具体目标更好。第三，本研究证实了省际间存在财政支出效率的集群特征，但这种空间外溢效应的作用机制尚不非常明确，本研究也没能更深入地给予解释。这些都是值得深入研究的。

总之，财政支出效率是财政支出理论中一个重要方面，完整准确科学地界定和评价财政支出效率并非易事，本研究在众多的相关研究中狗尾续貂，仅在定量研究上给出了一个研究视角和相应的研究结果，以期为我国地方财政支出效率的理论和实践提供部分参考。

关键词：财政支出 配置效率 技术效率 空间互动

Abstract

Since reform of tax assignment（fenshuizhi）,the critical issue in sub-national public finance has been mismatch between fiscal income and spending and fiscal pressure facing subnational governments.To allocate the public resource among items,governments have to pursue public spending efficiency.The dissertation focuses on such public spending efficiency as allocation efficiency,technique efficiency and spatial interaction among local governments.The research methods and conclusions are hoped to provide a reference for further research in the field.

The theme of economics is Pareto efficiency,so is public spending.Public spending efficiency refers to public spending meet the various public demands of social members in a most“economic”way,or in other words,best respond to（match）public preference.Actually,public spending efficiency at least comprises two different parts: allocation efficiency and technique efficiency.The form refers to public spending meet public demand.It examines the efficiency from outcome perspective.The latter refers to supply the most public goods by the least public resource.It examines efficiency from process.Besides,current fiscal system results in horizontal interaction among governments,so the paper analyzes the interaction of public efficiency.

The paper empirically studies such four 31 provincial public spending categories as education,healthcare,social security,and transportation and their efficiency respectively.

In democratic societies,public spending decisions are made by public choice and are determined by preference of“median voters”.Therefore,the best way to evaluate public spending allocation efficiency is estimate demand of“median voters”for public spending and compares actual public spending and preferred spending by median voters.If the residuals fall on between certain ranges,it can be deemed efficiency.Otherwise,inefficiency.So the reasonable range is critical.

The paper proposes two clues: one is estimates the critical values by using discrete ordered Logit Model.The other is to predetermine the critical values by using discrete multinomial Logit Model.Actually,provincial governments' choosing to allocate efficiently does not apply an ordered model,so the latter model is applied and RMSE is used as a measurement to judge whether efficiency.The results show Chinese provincial governments spending allocation efficiency experienced bell-shaped routes from 2007 to 2012.In 2009,especially,only half provinces achieved allocation efficiency.Owned public resources,among others,is decisive in spending allocation efficiency.

Due to scarcity,public resources are supposed to do more with less.The main difficulty with appraisal of technique efficiency of public spending is to determine exact input and output.In the paper,the public spending categories examined are semi public goods,so theoretically the supply costs are compensated by public as well as private sector fund.Therefore,it is hard to isolate output resulted from public fund.To deal with this,the paper uses Data Envelopment Analysis.DEA evaluates DMU ´ S efficiency by constructing efficiency frontier.It calculates the“relative”efficiency so that it is unnecessary to determine exact public input and output.DEA efficiency score,however,is influenced by endogenous like management efficiency and exogenous factors like lucky and random disturbances.To filter exogenous factors,three stages Bootstrapped DEA technique is used.Besides,the paper examines the dynamic changes of the public spending technique efficiency and reasons by using Malmquist index.The empirical results show there are 11% to 40% input waste in provincial governments spending categories from 2007 to 2012 when using input oriented DEA.The technique efficiency experienced upward trend at large and efficiency scores in east provinces are higher than that in west ones.Meanwhile,public spending return of scale economy is increasing in most provinces and public goods categories,which reflects that more inputs are needed to invest in such public services as education and healthcare.

Current fiscal system is a system to promote horizontal competition across governments.The faults of performance evaluation Chinese officials deteriorate the scientific and harmonious development goal to a goal of maximising GDP of local jurisdictions.In order to trigger economic growth,subject to a certain degree of balance budget,governments spending efficiency compete by using fiscal and tax instruments.So public spending efficiency interaction among jurisdictions needs analysing.To analyse interaction of public spending across provinces,the paper firstly examines clustering of the spending efficiencies using Moran's I technique,the results show all the spending categories efficiencies spatially interact.Secondly,the paper analyses in-depth efficiency interaction with spatial econometrics model,like spatial lag model,spatial error model and spatial Durbin model.The empirical results show the spending efficiency of public services examined spatially clustered and more efficient jurisdictions tended to cluster,and less efficient jurisdictions tended to cluster.Furthermore,spatial panel Durbin model are used to study the influencing factors to the efficiencies.The results show not only public spending efficiency spatially spill over,and some exogenous independent variables have spill over effect.

Some problems stillremain unsolved.First of all,efficiency appraisal by public sectors may be further discussed.Actually,there could be a big problem with efficiency appraisal by sectors,which it is does not take into account of“liquidity”issue.Take transportation spending efficiency for example,except for exogenous factors examined in the paper,efficiency interaction across public sectors should be considered.That is because low transportation spending efficiency may result from other public sectors rather than transportation sectors itself.Even more,lower spending efficiency in certain sectors can be good to public sector overall.Secondly,the paper proposes direct policy implications from the empirical appraisal instead of further research.Appraisal of public spending efficiency means much to understand public resource allocation and implementation.Therefore,how to interpret appraisal results and make policies for fiscal and certain function sectors needs to further research.Take transportation for example,we may know provincial government spending efficiency for transportation is lower than we expect by using input oriented DEA technique.Does it mean fiscal resource allocated to public transportation sectors should be cut?Then what if by using output oriented technique?Does it mean output goal should be lowered or even cancel goal?Thirdly,this study confirmed there were cluster characteritics of public spending efficiency among provinces.The machanism of spacial spillover effect,however,still remain unanswered.These questions worth more researching.

To sum up,public spending efficiency is one of the important parts in public spending theory.It's not easy to complete and define scientifically and appraise efficiency.The paper proposes an insight and some results using econometric instruments and hopes to be a supplement for the excellent related literature,and hope to provide a reference for related research in China as well.

Key words: Public Spending; Allocation Efficiency; Technique Efficiency;Spatial Interaction

After the reform of the tax-sharing system, a prominent problem in the operation of local fiscal expenditure in China is the mismatch between fiscal revenue rights and expenditure obligations, and local financial resources are difficult. In order to ensure that limited public resources are allocated as reasonably as possible among expenditure items, local governments must consider the efficiency of fiscal expenditure. This study takes the efficiency of provincial local fiscal expenditure in China as the research object, and mainly discusses the connotation of fiscal expenditure efficiency, fiscal expenditure allocation efficiency, fiscal expenditure technical efficiency and spatial interaction of local fiscal expenditure efficiency, and the research methods and conclusions provide a reference for further research on fiscal expenditure efficiency. The topic of economic discussion is the Pareto efficiency allocation of resources, and fiscal expenditure efficiency is also defined based on Pareto efficiency. Fiscal expenditure efficiency refers to the "effective response" or "matching" of fiscal expenditure to public preferences in the most "economic" way. The efficiency of fiscal expenditure includes at least two aspects: the efficiency of fiscal expenditure allocation and the technical efficiency of fiscal expenditure. The efficiency of fiscal expenditure allocation is the satisfaction of fiscal expenditure and residents' public demand preferences, and the expenditure efficiency is examined from the final result of fiscal expenditure. The technical efficiency of fiscal expenditure refers to the production of the most public goods with the least amount of financial input, which can be understood as the smallest financial expenditure invested under a certain amount of public goods or services, or the most public goods or services produced under a certain fiscal expenditure situation, they are called input-oriented and output-oriented. In addition, an important result of the current fiscal system arrangement is the horizontal competition and interaction between local governments, and the horizontal interaction of finance is manifested in expenditure and taxation, so if fiscal revenue is taken into account in the competition of fiscal expenditure, then the interaction and impact of fiscal expenditure efficiency between local governments must be studied. In this study, an empirical study of fiscal expenditure was conducted on four types of public goods: education, medical care, social security, and transportation in 31 provinces (autonomous regions and municipalities directly under the central government) in China. In a democratic society, fiscal expenditure is arranged through public choice, and the result of public choice often depends on the demand preference of "intermediate voters", therefore, the basic idea of fiscal expenditure allocation efficiency assessment is to estimate the local "intermediate voters" for public financial expenditure, and then compare the actual number of fiscal expenditure and the public expenditure needs of intermediate voters, if within a given range, the allocation of fiscal expenditure is considered to be effective; Conversely, if it exceeds this range, the allocation of fiscal expenditure is considered to be inefficient (low). What needs to be studied here is how to determine a reasonable "range", and this study believes that there are two ideas: one is to use the basic principles of the Ordered Logit model in the discrete selection model to estimate the interval value; The other is to use the principle of the Multinomial Logit model in the discrete selection model to determine the range value in advance. In this study, it is believed that the local government's choice of allocation efficiency is not a ranking model, so the Multinomial Logit model is adopted, and the scope of determination is based on root mean square error (RMSE) as the criterion for judging whether the allocation is efficient. Using the estimation results of the Multinomial Logit model, it is found that the allocation efficiency of local public goods expenditure in China experienced a state of change from 2007 to 2012, and the provinces (autonomous regions and municipalities directly under the central government) achieved the lowest allocation efficiency in 2009, and only about half of the provinces (autonomous regions and municipalities directly under the central government) achieved allocation efficiency. Among the main exogenous variables, the degree of public resources concentrated by local finances is the key to determining the efficiency of public goods fiscal expenditure. Public resources are scarce, so in the process of use, the principle of producing the most output with the least input must be followed. There are difficulties in evaluating the technical efficiency of fiscal expenditures, mainly because it is difficult to determine the output of public goods or services corresponding to fiscal inputs. Because the object of study in this book is basically the expenditure of quasi-public goods, the cost compensation method is both public and private sector funds, so it is difficult to separate the output of public expenditure. In this regard, the solution of this study is to adopt the data envelopment analysis method, which evaluates the relative efficiency of decision-making units by constructing the efficiency frontier, which largely avoids the dilemma of accurately determining the output indicators corresponding to fiscal expenditure. When using the data envelopment analysis method to evaluate the efficiency of local fiscal expenditure, the problem encountered is that the efficiency value is affected by endogenous factors and exogenous factors, endogenous factors such as internal financial management system, exogenous environmental variables such as luck and random factors, in order to filter the influence of exogenous environmental factors, this study uses the three-stage Bootstrapped DEA method for empirical evaluation. In addition, in order to examine the dynamic changes in the efficiency of local fiscal expenditure and the causes of the changes, the Malmquist index method was used in this study. After adopting input-oriented and controlling for environmental variables and random shocks, the DEA calculation of fiscal expenditure efficiency scores shows that between 2007 and 2012, there was an input waste of 11% to 40% of different local fiscal expenditures in China. On the whole, efficiency is on the rise, and the technical efficiency in the eastern region is higher than that in the central and western regions. In addition, the scale remuneration of most local public goods expenditure is in an increasing state, which also reflects that the input of local public goods, such as education and public health, needs to be further improved. China's current tax-sharing fiscal system is a system that promotes horizontal competition between regions, pursues the goal of scientific, harmonious and comprehensive development, and degenerates into the goal of maximizing local GDP under the current performance appraisal mechanism and financial system arrangement of local officials. In order to achieve regional economic growth, local governments use fiscal expenditure and taxation to compete across regions, and under balanced budget constraints, competition between fiscal expenditure efficiency (mainly technical efficiency) will be inevitable, so to study local fiscal expenditure efficiency, we must study the cross-regional interaction of local fiscal expenditure efficiency. In this study, Moran's I is first used to investigate the regional agglomeration of regional fiscal expenditure efficiency, and the results show that each category of public expenditure efficiency is spatially correlated. In order to further investigate the spatial characteristics of fiscal expenditure efficiency between places, this study uses spatial econometric models, namely spatial lag model, spatial error model and spatial Dulbin model. The results show that there is a spatial agglomeration in the efficiency of fiscal expenditure on education, medical and health, social security, and transportation, with areas with high fiscal expenditure efficiency concentrated together, and areas with low fiscal expenditure efficiency also concentrating. Then, the factors affecting the fiscal expenditure efficiency of local governments are empirically studied by using the spatial panel Durbin model, and the results show that not only the fiscal expenditure efficiency of neighboring places has a spatial spillover effect, but also some exogenous explanatory variables have a spatial spillover effect on the fiscal expenditure efficiency of other regions in addition to the fiscal expenditure efficiency of the region. There are still some unresolved issues in this study. First, sectoral efficiency assessments require more discussion. This study attempts to quantitatively analyze the efficiency of local fiscal expenditure in China, and evaluates specific public sectors, such as education, medical and health departments. In fact, an important problem in the evaluation object of the department is that the mobility of inter-sectoral efficiency is not taken into account, taking the efficiency of transportation expenditure as an example, in addition to some exogenous factors considered in this study, there is also a problem of efficiency interaction between departments, and the inefficiency of transportation expenditure may not be due to the reasons of this department, but due to the inefficiency of other departments. It is more likely that a sector's lower ranking in fiscal expenditure efficiency may be a Pareto efficiency improvement in the overall efficiency of fiscal spending. Second, this study only gives more direct policy enlightenment on the application of efficiency evaluation results, and fails to conduct in-depth research. The evaluation of fiscal expenditure efficiency has important practical significance for understanding the allocation and implementation of public funds in public functional departments, and how to interpret the results of expenditure efficiency evaluation and how to formulate policies by finance and specific functional departments needs in-depth discussion. Taking the efficiency of transportation expenditure as an example, under the input-oriented efficiency evaluation method, it is known that local transportation fiscal expenditure is inefficient and there is a lot of waste of input, should the financial support for the transportation sector be reduced? If inefficiencies are known under an output-oriented evaluation approach, should output targets be lowered? It is better not even to set specific goals. Third, this study confirms the cluster characteristics of fiscal expenditure efficiency between provinces, but the mechanism of this spatial spillover effect is not very clear, and this study has not been able to explain it in more depth. These are all worth studying in depth. In short, fiscal expenditure efficiency is an important aspect of fiscal expenditure theory, complete accurate scientific definition and evaluation of fiscal expenditure efficiency is not easy, this study in many related studies dog-tailed mink, only quantitative research to give a research perspective and corresponding research results, in order to provide some reference for the theory and practice of local fiscal expenditure efficiency in China. Keywords: fiscal expenditure, allocation efficiency, technical efficiency, spatial interaction, AbstractSince reform of tax assignment (fenshuizhi), the critical issue in sub-national public finance has been mismatch between fiscal income and spending and fiscal pressure facing subnational governments. To allocate the public resource among items,governments have to pursue public spending efficiency. The dissertation focuses on such public spending efficiency as allocation efficiency,technique efficiency and spatial interaction among local governments. The research methods and conclusions are hoped to provide a reference for further research in the field. The theme of economics is Pareto efficiency,so is public spending. Public spending efficiency refers to public spending meet the various public demands of social members in a most“economic”way,or in other words,best respond to（match）public preference. Actually,public spending efficiency at least comprises two different parts: allocation efficiency and technique efficiency. The form refers to public spending meet public demand. It examines the efficiency from outcome perspective. The latter refers to supply the most public goods by the least public resource. It examines efficiency from process. Besides,current fiscal system results in horizontal interaction among governments,so the paper analyzes the interaction of public efficiency. The paper empirically studies such four 31 provincial public spending categories as education,healthcare,social security,and transportation and their efficiency respectively. In democratic societies,public spending decisions are made by public choice and are determined by preference of“median voters”. Therefore,the best way to evaluate public spending allocation efficiency is estimate demand of“median voters”for public spending and compares actual public spending and preferred spending by median voters. If the residuals fall on between certain ranges,it can be deemed efficiency. Otherwise,inefficiency. So the reasonable range is critical. The paper proposes two clues: one is estimates the critical values by using discrete ordered Logit Model.The other is to predetermine the critical values by using discrete multinomial Logit Model.Actually,provincial governments' choosing to allocate efficiently does not apply an ordered model,so the latter model is applied and RMSE is used as a measurement to judge whether efficiency. The results show Chinese provincial governments spending allocation efficiency experienced bell-shaped routes from 2007 to 2012.In 2009,especially,only half provinces achieved allocation efficiency. Owned public resources,among others,is decisive in spending allocation efficiency. Due to scarcity,public resources are supposed to do more with less. The main difficulty with appraisal of technique efficiency of public spending is to determine exact input and output. In the paper,the public spending categories examined are semi public goods,so theoretically the supply costs are compensated by public as well as private sector fund. Therefore,it is hard to isolate output resulted from public fund. To deal with this,the paper uses Data Envelopment Analysis.DEA evaluates DMU ´ S efficiency by constructing efficiency frontier. It calculates the“relative”efficiency so that it is unnecessary to determine exact public input and output. DEA efficiency score,however,is influenced by endogenous like management efficiency and exogenous factors like lucky and random disturbances. To filter exogenous factors,three stages Bootstrapped DEA technique is used. Besides,the paper examines the dynamic changes of the public spending technique efficiency and reasons by using Malmquist index. The empirical results show there are 11% to 40% input waste in provincial governments spending categories from 2007 to 2012 when using input oriented DEA. The technique efficiency experienced upward trend at large and efficiency scores in east provinces are higher than that in west ones. Meanwhile,public spending return of scale economy is increasing in most provinces and public goods categories,which reflects that more inputs are needed to invest in such public services as education and healthcare. Current fiscal system is a system to promote horizontal competition across governments. The faults of performance evaluation Chinese officials deteriorate the scientific and harmonious development goal to a goal of maximising GDP of local jurisdictions. In order to trigger economic growth,subject to a certain degree of balance budget,governments spending efficiency compete by using fiscal and tax instruments. So public spending efficiency interaction among jurisdictions needs analysing. To analyse interaction of public spending across provinces,the paper firstly examines clustering of the spending efficiencies using Moran's I technique,the results show all the spending categories efficiencies spatially interact. Secondly,the paper analyses in-depth efficiency interaction with spatial econometrics model,like spatial lag model,spatial error model and spatial Durbin model. The empirical results show the spending efficiency of public services examined spatially clustered and more efficient jurisdictions tended to cluster,and less efficient jurisdictions tended to cluster. Furthermore,spatial panel Durbin model are used to study the influencing factors to the efficiencies. The results show not only public spending efficiency spatially spill over,and some exogenous independent variables have spill over effect. Some problems stillremain unsolved. First of all,efficiency appraisal by public sectors may be further discussed. Actually,there could be a big problem with efficiency appraisal by sectors,which it is does not take into account of“liquidity”issue. Take transportation spending efficiency for example,except for exogenous factors examined in the paper,efficiency interaction across public sectors should be considered. That is because low transportation spending efficiency may result from other public sectors rather than transportation sectors itself. Even more,lower spending efficiency in certain sectors can be good to public sector overall. Secondly,the paper proposes direct policy implications from the empirical appraisal instead of further research. Appraisal of public spending efficiency means much to understand public resource allocation and implementation. Therefore,how to interpret appraisal results and make policies for fiscal and certain function sectors needs to further research. Take transportation for example,we may know provincial government spending efficiency for transportation is lower than we expect by using input oriented DEA technique. Does it mean fiscal resource allocated to public transportation sectors should be cut? Then what if by using output oriented technique? Does it mean output goal should be lowered or even cancel goal? Thirdly,this study confirmed there were cluster characteritics of public spending efficiency among provinces. The machanism of spacial spillover effect,however,still remain unanswered. These questions worth more researching. To sum up,public spending efficiency is one of the important parts in public spending theory. It's not easy to complete and define scientifically and appraise efficiency. The paper proposes an insight and some results using econometric instruments and hopes to be a supplement for the excellent related literature,and hope to provide a reference for related research in China as well. Key words: Public Spending; Allocation Efficiency; Technique Efficiency; Spatial Interaction（AI翻译）