社会核算矩阵平衡方法研究

本文针对双比例尺度(RAS)、交叉熵(CE)等方法在平衡社会核算矩阵(SAM)中仅从技术层面机械地进行平衡化处理致使先验信息损失的问题,提出了加权离差熵平方期望最小化方法；并以先验信息为基础,构造了初始加权矩阵和可行加权矩阵.同时,本文以中国2007年的非平衡SAM为例,对比研究RAS、CE和加权离差熵平方期望最小化三种方法对其进行平衡化处理的实际效果.结果表明:RAS方法得到的结果偏差相对较大,而CE方法和加权离差熵平方期望最小化方法得到的结果相对较精准；此外,加权离差熵平方期望最小化方法能够有效利用先验信息,避免有效信息的无谓损失.

Study on the Balance Approach of the SAM

Considering the defects of the RAS and Cross-Entropy(CE) approaches that losing the priori information when they are applied for the balance of the Social Accounting Matrix(SAM), this paper proposes the weighted approach which is based on minimizing the expectation of the deviation entropy square. And it constructs the weighting matrix in accordance with the degree of the prior information. Meanwhile, this paper takes the Chinese unbalanced SAM in 2007 as an example, and compares the real effects of balancing among RAS approach, CE approach and weighted approach based on minimizing the expectation of the deviation entropy square. And the results show that: RAS approach gets the results with a larger deviation, while the results produced by CE approach and weighted approach based on minimizing the expectation of the deviation entropy square are more accurate. Furthermore, weighted approach based on minimizing the expectation of the deviation entropy square can make flexible and effective use of the prior information and avoiding this deadweight loss of effective information, but RAS and CE approaches do not have this advantage.