社会核算矩阵SAM表的调平处理方法

文章基于EGRAS模型对社会核算矩阵进行乘数分解,并对SAM表中的先验信息进行"保号、保零、保凸"统计处理,重点针对未通过方向检验与接近度检验的矩阵系数进行调平。数值模拟的结果表明,EGRAS方法具有无偏、避免正负向相抵等技术优势,且无须事先定义0/0与0 ln 0数值,可在交叉熵差值较低的条件下求得全局最优解。为验证这一结论的可靠性,将基于EGRAS的调平方法应用于经济系统中活动账户、要素账户、部门账户等内生账户对外生变量敏感性程度的实证研究中。实证发现,EGRAS方法不仅对部分初始零值流量调整系数进行了调平修正,也增强了社会核算矩阵统计结果的可靠性与精度,显著提升了SAM表的解释效力。

EGRAS Technology-based SAM Table Leveling Processing Method of Social Accounting Matrix

This paper uses the EGRAS model to decompose the multiplier of social accounting matrix(SAM),and makes statistical treatment of prior information in SAM table"guarantee number,zero and convex",with the focus on leveling the matrix coefficients that fail the direction test and the proximity test.The results of numerical simulation show that the EGRAS method has the advantages of unbiasedness,avoidance of the positive and negative offset,etc.and does not need to define the values of 0/0 and 0 ln 0 in advance,able to obtain the global optimal solution under the condition of low cross entropy difference.In order to verify the reliability of this conclusion,the paper applies the leveling method based on EGRAS to the empirical study on the sensitivity of endogenous accounts such as active account,factor account and department account to exogenous variables in the economic system.The empirical results show that the EGRAS method not only makes some initial zero-value flow adjustment coefficients leveled and corrected,but strengthens the reliability and accuracy of the statistical results of the social accounting matrix,significantly improving the explanatory power of SAM tables.