计量经济模型适用于评估GDP数据质量吗?

Can Econometric Models be Applicable to Evaluating the Accuracy of GDP Statistics?

针对中国GDP数据质量评估研究中广泛采用的计量经济模型方法,本文从其技术原理的局限性以及GDP核算误差发生机制的影响效应两个方面,系统考察了该方法在具体评估实践中的适用性。研究表明,基于计量经济模型的两类评估方法(参数可靠性分析与异常数值识别),在实际操作中易于陷入“数据质量不佳”或者“模型结构设定不当”的对立解读困境;因忽视了GDP核算误差的发生机制,根据模型估计结果的异常特征来反推GDP数据序列中误差结构(效应),其可行性有所不足。计量模型方法之于GDP数据质量评估研究的适用性的提升,还有待于对GDP核算误差机制的显性识别;本文在此方面提供了一个相对完备的概念性基准模型,有助于相关研究的进一步深化拓展。

By means of an analysis on methodological limitations and a stochastic numerical simulation following several hypotheses about the generation mechanisms of GDP accounting errors,the paper aims at examining the applicability of those approaches of adopting econometric models(AEM for short)into evaluating the accuracy of GDP statistics.Our results indicate there is a dilemma for AEM to interpret biased parameters and abnormal residuals got from the fitted econometric models as poor data quality or inappropriate model specifications.Meanwhile,it is difficult to infer the componential and compositive effects of GDP error from the fitted econometric models,since the generation mechanisms of GDP accounting errors are always neglected.In order to improve the applicability of AEM,the explicit identification of the generation mechanism of GDP accounting errors is needed.In this aspect,the paper develops a relatively complete conceptual model,which provides a benchmark for further research.