统计与数据科学学院知识库

党的二十大报告和中央经济工作会议多次强调稳增长、稳物价和防风险的重要性，坚持稳中求进的工作总基调，保持经济稳定增长和物价总体稳定，着力实现稳增长和物价合理回升的优化组合，这一战略导向凸显宏观经济稳健运行对国家治理现代化的核心价值。当前，在内部结构性矛盾与外部环境不确定性叠加的复杂形势下，对宏观经济风险进行深入研究，有助于防范化解系统性经济风险，促进“新安全格局保障新发展格局”的战略协同，实现经济质的有效提升与量的合理增长有机统一。

宏观经济平稳发展的理想状态体现为适度增速与温和通胀的动态均衡，二者形成协同共生关系，共同构成宏观经济稳态的核心表征与政策调控的锚定基准。因此，精准测度经济增长和通货膨胀风险，识别当前宏观经济脆弱点，构成监测宏观经济稳态及优化政策工具箱的基础性前提。据此，本文将经济增长和通货膨胀纳入统一的在险框架，依托数据驱动的研究范式，改进并拓展相关研究方法，提升方法与应用问题的适配性，构建更精准的模型，科学全面地测度宏观经济风险，深入剖析潜在风险的驱动因素，立足研究结论，提出契合的宏观经济政策建议。具体而言，本文从方法改进与拓展，以及宏观经济风险测度与溯源两个方面展开研究。

论文的主要工作和结论如下：

第一，针对具有异常值与尖峰厚尾特征的同频高维数据情境，本文提出分位数因子增广分位数自回归分布滞后模型，该模型能够捕捉公共因子的异质性特征，具有可解释性，可以刻画高维同频变量间的异质性关系。通过数值模拟发现，该模型在预测方面具有独特优势。

第二，针对具有异常值与尖峰厚尾特征的混频高维数据情境，本文提出分位数因子增广混频分位数回归模型，该模型可以捕捉公共因子的异质性特征，具有可解释性，能够度量高维混频变量间的异质性关系。通过数值模拟发现，该模型在预测方面具有独特优势。

第三，针对稳增长目标，本文基于数据特征，改进传统的无条件与条件在险增长测度方法，并对中国经济增长风险进行测度及溯源。研究结果表明：中国经济增长的上行风险和下行风险呈现显著的非对称性特征；当前宏观经济运行面临显著的经济增长下行风险和上行风险，经济不确定性增加，整体经济低位运行，但具有较高的经济韧性；经济低位运行时期，消费因子是影响经济增长风险的关键因素；极端事件驱动经济下行风险增加。

第四，针对稳物价目标，本文基于数据特征，改进传统的无条件与条件在险通货膨胀测度方法，并对中国通货膨胀风险进行测度及溯源。研究结果表明：中国通货膨胀风险和通货紧缩风险具有显著的非对称性特征；当前宏观经济运行面临显著的通货紧缩风险，物价低位运行，下行压力相对较大；经济低位运行时期，货币因子是影响通货紧缩风险的关键因素；极端事件冲击下，物价波动区间变大，不确定性增加。

综上所述，中国宏观经济风险具有以下特征：宏观经济风险具有非对称性特征；当前经济低位运行，具有较高的经济增长上行风险、下行风险和通货紧缩风险；经济低位运行时期，驱动宏观经济风险的因素主要有：消费因子、金融因子、汇率因子和投资因子。基于系列分位数因子增广回归方法探究宏观经济风险，具有较高的适配性，提高了结果的准确性。

论文的创新和边际贡献如下：

一是在理论方法层面，依据宏观经济在险框架，结合中国宏观经济数据中呈现的尖峰厚尾和具有异常值的特征，将在险测度中的降维和回归两步骤均放在分位数回归框架下建模，构建不同数据情境下的系列分位数因子增广回归方法，拓宽该领域的研究视角和研究方法，包括：（1）在同频数据情境下，构建分位数因子增广分位数自回归分布滞后模型；（2）在混频数据情境下，构建分位数因子增广混频分位数回归模型。这些方法为宏观经济风险的精准测度和风险源识别提供新的思路和方法支撑，并通过与基准模型对比，验证新模型预测的准确性。

二是在实践应用层面，选择宏观经济的核心指标，从无条件分布和条件预测分布以及相对熵三个维度，结合分位数因子增广混频分位数回归模型对中国经济在险增长进行测度，并进行溯源研究；从稳物价的角度，运用分位数因子增广分位数自回归分布滞后模型对中国通货膨胀进行风险测度及归因分析。上述研究，在深化宏观经济风险测度认识方面，提供了重要的技术方法参考，同时，模型分析的结论及政策建议，也具有实际参考价值，为宏观经济调控提供重要依据。

The report of the 20th National Congress of the Communist Party of China and the Central Economic Work Conference have repeatedly emphasized the importance of maintaining stable growth, stable prices, and preventing risks, adhering to the general working principle of seeking progress while maintaining stability, and maintaining stable economic growth and overall price stability. The strategic orientation aims to achieve an optimized combination of stable growth and reasonable price recovery, highlighting the core value of stable macroeconomic operation for the modernization of national governance. Currently, under the complex situation of overlapping internal structural contradictions and external environmental uncertainties, in-depth research on macroeconomic risk is conducive to preventing and defusing systemic economic risk, promoting the strategic synergy of "new security pattern guaranteeing new development pattern", and achieving the organic unity of effective quality improvement and reasonable growth in quantity of the economy.

The ideal state of stable macroeconomic development is characterized by a dynamic equilibrium of moderate growth and mild inflation, which form a symbiotic relationship and constitute the core representation of the macroeconomic steady state and the anchor for policy regulation. Therefore, accurately measuring economic growth and inflation risks and identifying the current vulnerable points of the macroeconomy are the fundamental prerequisites for monitoring the macroeconomic steady state and optimizing the policy toolkit. Based on this, this thesis incorporates economic growth and inflation into a unified risk framework, relies on a data-driven research paradigm, improves and expands relevant research methods, enhances the adaptability of methods to application problems, builds more accurate models, scientifically and comprehensively measures macroeconomic risk, deeply analyzes the driving factors of potential risks, and based on research conclusions, proposes appropriate macroeconomic policy recommendations. Specifically, this thesis conducts research from two aspects: method improvement and expansion, and measurement and tracing of macroeconomic risk.

The main work and conclusions of this thesis are as follows:

First, for the context of high-dimensional same-frequency data with outliers and heavy-tailed characteristics, this thesis proposes a Quantile Factor-Augmented Quantile Autoregressive Distributed Lag model. This model can capture the heterogeneous characteristics of common factors. It is interpretable and can also describe the heterogeneous relationships among high-dimensional same-frequency variables. Through numerical simulation, it is found that this model has unique advantages in prediction.

Second, for the context of high-dimensional mixed-frequency data with outliers and heavy-tailed characteristics, this thesis proposes a Quantile Factor-Augmented Mixed Data Sampling Quantile Regression model. This model can capture the heterogeneous characteristics of common factors. It is interpretable and can also measure the heterogeneous relationships among high-dimensional mixed-frequency variables. Through numerical simulation, it is found that this model has unique advantages in prediction.

Third, in response to the goal of stable growth, based on data characteristics, this paper improves the traditional unconditional and conditional Growth-at-Risk measurement methods and measures and traces the growth risk of the Chinese economy. The research results show that the growth risk of the Chinese economy has significant asymmetric characteristics; the current macroeconomic operation faces significant downward growth risk and upward growth risk, economic uncertainty increases, and the overall economy operates at a low level, but has high economic resilience; during the period of low economic operation, the consumption factor is the key factor affecting growth risk; extreme events drive an increase in downward growth risk.

Fourth, in response to the goal of stable prices, based on data characteristics, this paper improves the traditional unconditional and conditional Inflation-at-Risk measurement methods and measures and traces the inflation risk in China. The research results show that the inflation risk and deflation risk in China have significant asymmetric characteristics; the current macroeconomic operation faces significant deflation risk, prices operate at a low level, and the downward pressure is relatively large; during the period of low economic operation, the monetary factor is the key factor affecting deflation risk; under the impact of extreme events, the price fluctuation range increases and uncertainty increases.

In summary, the macroeconomic risk in China have the following characteristics: macroeconomic risk have asymmetric characteristics; the current economy operates at a low level, with significant upward growth risk, downward growth risk, and deflation risk; during the period of low economic operation, the main factors driving macroeconomic risks are: consumption factor, financial factor, exchange rate factor, and investment factor. Exploring macroeconomic risk based on a series of Quantile Factor-Augmented Regression methods has high adaptability and improves the accuracy of the results.

The innovations and marginal contributions of this thesis are as follows:

First, in terms of theoretical methods, based on the macroeconomic at-risk framework and the characteristics of Leptokurtic, heavy-tailed, and outliers in Chinese macroeconomic data, both the dimension reduction and regression steps in the at-risk measurement are modeled within the quantile regression framework, constructing a series of Quantile Factor-Augmented Quantile Regression methods under different data scenarios, broadening the research perspective and methods in this field, including: (1) In the same-frequency data scenario, a Quantile Factor-Augmented Quantile Autoregressive Distributed Lag model is constructed; (2) In the mixed-frequency data scenario, a Quantile Factor-Augmented Mixed Data Sampling Quantile Regression model is constructed. These methods provide new ideas and methodological support for the precise measurement of macroeconomic risk and the identification of risk sources, and their accuracy is verified through comparisons with benchmark models.

Second, in the practical application aspect, the core indicators of the macroeconomy are selected, and Chinese Growth at-Risk is measured from three dimensions: unconditional distribution, conditional predictive distribution, and relative entropy, combined with the Quantile Factor-Augmented Mixed Data Sampling Quantile Regression model, and a traceability study is conducted. From the perspective of stabilizing prices, the Quantile Factor-Augmented Quantile Autoregressive Distributed Lag model is used to measure and attribute the risk of Chinese inflation. The above research provides important technical method references for deepening the understanding of macroeconomic risk measurement. At the same time, the conclusions and policy recommendations from the model analysis also have practical reference value and provide an important basis for macroeconomic regulation.