统计与数据科学学院知识库

  新冠肺炎疫情的全球持续蔓延和不断增大的不确定性，剧烈冲击全球经济，严重影响国际贸易。面对国内外风险挑战明显上升的复杂局面，中国把握发展大势，积极统筹国内外的市场和资源，努力构建以国内大循环为主体、国内外双循环相互促进的新发展格局。新发展格局的动力源泉是高水平对外开放，贸易规模的大小决定着贸易投资的开放程度，在此背景下，本文旨在准确预测出口贸易，对未来出口贸易发展和精准调控贸易政策提供依据。

  本文构建并论证分位数扩散指数自回归模型在中国出口预测中的优越性，同时进行趋势外推预测。首先，基于111个宏观经济月度变量信息集，融合分位数回归和因子模型构建分位数因子模型，在不同分位点上提取中国宏观经济公共因子。其次，利用提取出的中国宏观经济公共因子构建分位数扩散指数模型和分位数扩散指数自回归模型，并比较论证分位数扩散指数自回归模型的预测能力高于分位数扩散指数模型的预测能力。再次，将选出的预测最优的分位数扩散指数自回归模型与自回归模型、扩散指数自回归模型进行对比预测，结果表明，在预测我国月度出口同比增长速度时，分位数扩散指数自回归模型优于扩散指数自回归模型优于自回归模型。最后，为验证模型的稳健性，本文将固定窗口滚动预测中的固定窗口由80个月调整为215个月，即以金融危机为切点调整为以新冠肺炎疫情为切点，得到的结论仍然是分位数扩散指数自回归模型的预测能力是最强的。

  鉴于此，本文采用分位数扩散指数自回归模型进行样本外一步预测，即对2021年10月我国月度出口同比增长速度进行预测，结果显示，该模型在0.75分位点上预测最优，预测值为26.9%，结合实际，国家统计局公布的2021年10月我国月度出口同比增长速度为27.1%，相差大约0.2%。因此，分位数扩散指数自回归模型在中国出口预测方面的预测能力是非常强的，同时该模型也可以用来预测其他经济变量。

    The continuing global spread of COVID-19 and growing uncertainties have severely impacted the global economy and international trade. Facing the complex situation of rising risks and challenges both at home and abroad, China has grasped the trend of development, actively coordinated the market and resources at home and abroad, and worked hard to build a new development pattern in which the major domestic cycles play the main role and the double cycles at home and abroad reinforce each other. The driving force of the new development pattern is the high-level opening to the outside world. The size of trade determines the degree of opening to trade and investment. In this context, this thesis aims to accurately forecast export trade and provide a basis for the future development of export trade and precise regulation of trade policies.

    This thesis establishes and demonstrates the superiority of quantile diffusion index autoregressive model in China's export forecast, and makes trend extrapolation prediction. Firstly, based on the information set of 111 monthly macroeconomic variables, the quantile regression and factor model were combined to construct the quantile factor model, and the common macroeconomic factors of China were extracted at different quantile. Secondly, the quantile diffusion index model and quantile diffusion index autoregression model are constructed by using the common factors extracted from China's macro economy, and the prediction ability of quantile diffusion index autoregression model is higher than that of quantile diffusion index model. Again, will choose to predict the optimal quantile regression model and the diffusion index regression model, the diffusion index back to the regression model, and compared the prediction results show that the forecast monthly year-on-year growth in China, the spread of quantile autoregressive model is better than that of diffusion index since the regression model is better than the regression model. Finally, in order to verify the robustness of the model, the fixed window of the fixed window rolling forecast was adjusted from 80 months to 215 months, that is, the financial crisis was adjusted to COVID-19 as the cut-off point. The conclusion was still that the prediction ability of the autoregressive model of quantile diffusion index was the strongest.

    In view of this, this thesis adopts the quantile diffusion index autoregression model to make an out-of-sample prediction, that is, to forecast the year-on-year growth rate of China's monthly export in October 2021. The results show that the model is the best to predict the growth rate of China's monthly export at 0.75 points, and the predicted value is 26.9%. According to the National Bureau of Statistics, the year-on-year growth rate of China's monthly export in October 2021 was 27.1%, with a difference of about 0.2%. Therefore, the quantile diffusion index autoregressive model is very strong in the prediction of Chinese export, and the model can also be used to predict other economic variables.