基于微博大数据的经济不确定性测度及其对宏观经济的影响研究

经济不确定性是一个重要的经济参数,它反映了经济的运行状态与经济参与者的评估与预期不一致的程度,与宏观经济运行和政策之间存在密切联系。经济不确定性的增加可能会对经济运行产生负面影响,因此研究经济不确定性对理解经济运行中的波动、制定宏观调控政策、激发市场经济活力等具有重要的意义,量化测度经济不确定性就是其中最关键的一步。文章提出了一种基于微博大数据的经济不确定性指数的量化构建方法,主要思想是利用人们对经济形势的主观认知态度的不一致程度来衡量经济的不确定性。该方法能考虑到更多经济参与者对经济的主观看法,从新的角度量化了我国经济不确定性的程度,且数据量大,覆盖面广,更能实时地体现经济不确定性程度的变化。首先,利用长短期记忆（LSTM）模型对人们实时发布的微博中对经济运行的主观认知态度的情感倾向进行分类,效果良好,节约大量人工判别成本,有利于经济不确定性指数的实时计算和预测。然后,通过计算公众对经济持乐观态度和悲观态度的比例差异来计算经济不确定性指数EUI （Economic Uncertainty Index）。EUI反映的是经济参与者对经济运行状态的情感不一致程度,其不一致程度越大,经济不确定性程度越大,反之则越小。实证分析结果显示：微博数据量与经济热点事件有直接联系,时效性强,微博数据对重大经济事件的发生具有良好的反映;基于微博大数据构建的EUI与经济发展趋势具有一致性,对重大事件发生都有即时的反映;EUI与其他文献中常用的指数反映的经济不确定性程度和变化具有较高的一致性,例如EUI与Baker等编制的中国经济政策不确定性（EPU）指数变化趋势基本相似,且EUI的时效性更强;EUI与市场波动性指数和股票收益率方差具有正相关性,与消费者信心指数和基金经理信心指数具有负相关性,符合实际经济意义。更进一步将EUI应用于对我国宏观经济的影响分析中,结果表明,经济不确定性对进出口和消费有显著的负向影响。

Measuring economic uncertainty of China based on Weibo data and its impact on macroeconomy

Economic uncertainty is an important economic index, which measures the degree of inconsistency between the economic conditions and the expectations of economic actors, and is closely related to the macroeconomy performance and policy making. The increase of economic uncertainty may have negative effects on economic activities. Therefore, it is important to study economic uncertainty for understanding fluctuations in the economy, formulating macro-control policies, and stimulating market vitality. Since the economic uncertainty is not directly observable, it is necessary to develop a method to quantify the uncertainty and assess its macroeconomic effects. We propose to construct an economic uncertainty index based on the Weibo data, which involves more people''s views regarding the economy. This method takes the public opinions into account and can show the timely changes of the economic uncertainty. First, we use a long and short-term memory (LSTM) model to classify the sentiments of the comments posted on Weibo on the economic issues. This method saves a lot of time and cost to classify the sentiments of the Weibo comments manually and is helpful for the real-time measurement of the economic uncertainty index. Then, we construct the Economic Uncertainty Index(EUI) by measuring the inconsistency of the public opinions. The empirical analysis results show that:1) Weibo data has a good reflection on the important economic issues, and EUI is consistent with the economic development trends and shows spikes around the major events; 2) EUI is consistent with other economic uncertainty indices which are commonly used in literatures. For example, EUI shows similar patterns with EPU proposed by Baker et al. (2016) and provides timely information for important economic issues. Furthermore, EUI is positively correlated with the market volatility index and the stock return variance and negatively correlated with the Consumer Confidence Index and Fund Manager Confidence Index; 3) Using the vector autoregressive (VAR) model to assess the impact of EUI on the macroeconomy, the results show that EUI has a significant negative impact on imports, exports and consumption.