更新社会核算矩阵(Social Accounting Matrix,SAM)作为一种非调查技术一直受到普遍重视。本文用十种更新方法,包括RAS法、交叉熵法、广义交叉熵法等,以江苏省1997年的宏观和细化SAM为初始表,将其分别更新到2002年的宏观和细化SAM。笔者通过保号检验、方向检验和接近程度检验比较了各种方法的更新结果。研究结论显示:第一,基于商的更新方法要优于基于距离的更新方法;第二,保号类方法不仅具有保号的功能,而且在方向检验和接近度检验中有着良好表现;第三,对某些重要系数采用适当的外生估算方式,可以提高SAM更新的精度。
Updating social accounting matrix (SAM),as a non-survey technique,is always stresses. The paper uses 10 methods to update Jiangsu province's macro and micro SAM from 1997 to 2002,including RAS,generalized RAS and cross entropy methods.We adopt sign-preservation test,direction test and closeness test to compare the methods,and draw the following conclusions.Firstly,quotient-based updating methods are better than distance-based methods.Secondly,sign-preservation methods also have good performance in direction test and closeness test.Thirdly,external estimation on key coefficients is necessary for upgrading SAM updating accuracy.