摘要
更新投入产出(input-output,IO)表作为一种非调查技术一直受到普遍重视。已有的文献强调整合目标表的信息,在不同程度上忽视了保持初始表的原有结构。本文提出添加保零约束以保持初始表的零值结构,并采用九种更新方法,将中国1997年和2002年的IO表分别向前和倒推更新到2000年的IO表。通过与现有2000年的IO延长表进行比较检验,我们发现:第一,添加保零约束不仅能继承初始表的零值结构,而且极大提高了各种更新方法计算结果的精度;第二,添加保零约束不仅提高了各种更新方法的局部精度,而且提高了整体精度,其中以交叉熵法的精度提高最大;第三,以1997年为初始表的更新结果好于以2002年为初始表的倒推更新结果暗示着现有的2000年IO延长表有待于必要的修正。
：Updating input-output （IO） tables is always highlighted as a non-survey technique. Existing literatures attach importance to integrating information about target tables, and neglect preserving the structure of the original table. The paper proposes adding zero-preservation constraints so as to keep the zero structure of the original table, and adopts nine updating methods forecasting China＇ s IO table from 1997 to 2000 and backcasting China＇ s IO table from 2002 to 2000. The updating results are compared with available IO table of 2000. The following findings are obtained ： first, zero-pres- ervation constraints can both inherit the zero structure of the original table and upgrade updating accuracy of methods; second, both partitive and holistic accuracy are upgraded, and the method of cross entropy enjoy the largest improvement among all methods; third, the fact that an updating result from 1997 IO table is better than that from 2002 IO table implies that existing IO table of 2000 should be revised.
出处
《预测》
CSSCI
2008年第2期55-61,共7页
Forecasting
基金
国家自然科学基金资助项目(70471073)
关键词
投入产出表
更新
保零约束
input-output table
updating
zero-preservation constraint
