【作者】 刘超；

【导师】 葛虹；

【作者基本信息】 哈尔滨工业大学 ， 运筹学与控制论， 2006， 硕士

【摘要】 生产函数分析中急需解决的问题之一是如何选择合适的函数形式。根据前沿生产函数的定义,本文利用投入产出对数形式和数据包络分析方法求出了相对指数最大产出,并证明了在一定的条件下,这种数据结构即所谓的支撑面板数据具有分段近似Cobb-Douglas性。由此说明在不违背前沿生产函数经济学含义的前提下,函数形式是可以初步确认的,而无需通过后续的假设检验进行验证。利用Cobb-Douglas生产函数作为聚合前沿生产函数研究行业经济或跨国经济的必要条件是所有经济的要素边际产出应该相等。尽管人们已经注意到由于各个经济技术条件和技术水平之间的差异使这个条件无法满足,但并没有给出一个有效的方法从根本上解决这个问题。为了反映行业(跨国)之间的技术条件和技术水平的差异,本文利用聚合前沿生产函数和行业生产前沿之间的关系规范了技术系数的概念。在此基础上,利用支撑面板数据的近似Cobb-Douglas性和变异系数回归模型给出了一种估计聚合前沿生产函数(它能够反映各个经济体技术条件和技术水平之间的差异)和技术系数的方法,从而为宏观经济领域中的研究工作提供了一个更加有效的计量分析工具。最后我们运用SAS软件对2000年我国33个行业的截面数据和2000-2003年33个行业的面板数据进行了回归分析,采用变异系数生产前沿测算了技术系数,并进行了分析、给出了某些政策性的建议。更多还原

【Abstract】 One of problem unsettled in the analysis of production function is how to choose the appropriate form of function. The frontier production function, a parametric function of any form, can be estimated by sample data. This study shows that the projected panel data, which is the inputs and relative exponential maximum output data set, obtained by conducting the data envelopment analysis with the log inputs and output, is approximately piecewise Cobb-Douglas under a certain assumption. Thus the expression of production function can be determined preliminarily without violation of its economic implication and no further test should be followed.The Cobb-Douglas production function can be accumulated to be an aggregate production function if the marginal products are all equal across industries or countries. But this is not always satisfied since the technology developed and production technique employed varies from one industry (country) to another. In this study, the technical coefficients are defined so as to capture the difference in technology and production techniques across industries (countries). Then, the varying coefficients regression model is applied in estimating aggregate production function and the technical coefficients with the projected panel data. Thus, the derived best production practice, which exhibits the difference in technology and production technique across industries (countries), could be used more effectively in the field of macroeconomics.With the SAS software, we regress the cross-section data for 33 industries in 2000, and the panel data from 2000 to 2003. Then we can estimate the technical coefficient by using the varying coefficients frontier model, and give some suggestion for our country.更多还原