基于结构关系识别的中国汽车销量预测

Chinese automobile sales forecasting based on structural relationship identification

准确的销量预测有助于汽车企业合理安排生产计划.提出采用包含单位根检验、格兰杰因果检验、弱外生性检验以及协整检验的结构关系识别方法来研究中国汽车销量与宏观经济变量之间的动态联系.其中,宏观经济变量主要考虑了汽油价格、消费者信心指数、居民消费指数和钢材产量,收集了2007~2016年的月度数据,构建了用于实证的数据集.研究表明中国汽车销量与识别的内生变量间存在着长期的协整关系,基于此,构建向量误差修正模型以量化这些变量对中国汽车销量的长期影响.与传统时间序列方法的比较表明,所提方法能提高预测精度,更好地反映中国汽车销量与宏观经济变量之间由短期偏离向长期均衡调整的动态过程.

Accurate prediction helps automobile companies arrange production plans.A structural relationship identification approach that contains a battery of statistical unit root,Granger-causality,weak exogeneity and cointegration tests,is presented to research the dynamic couplings among Chinese automobile sales and macroeconomic variables.The monthly data of China for the period from2007 to 2016 are built as the data sets for empirical study.Macroeconomic variables such as gasoline price,consumer confidence index（CCI）,consumer price index（CPI）and steel production are selected.Research result shows that there are long-term cointegration relationships among Chinese automobile sales and identified endogenous variables.A vector error correction model（VECM）is built to quantify long-term impact of macroeconomic variables on Chinese automobile sales.Compared with other classical time-series methods,the presented approach can improve the prediction accuracy and reflect the dynamic process of adjusting from short-term departure to long-term equilibrium well among Chinese automobile sales and macroeconomic variables.