Dan Ma

This study investigates the impact of the COVID-19 pandemic on economic uncertainty and its spillover network based on social network analysis. The study constructs the inter-provincial Chinese economic uncertainty spillover network using a mixed frequency dataset and the provincial social network using microblog user data. Furthermore, the temporal exponential random graph model is used to analyze the impact of COVID-19 and social network during three periods. The results show that the COVID-19 pandemic significantly affects China’s provincial economic uncertainty and social network significantly hinders economic uncertainty spillover networks. The inhibitory effect of social networks on uncertainty spillover network has regional heterogeneity, which is more significant in provinces severely affected by the pandemic and strictly controlled.

This paper further extends the existing GARCH‐MIDAS model to deal with the effect of microstructure noise in mixed frequency data. This paper has two highlights. First, according to the estimation of the long‐term volatility components of the GARCH‐MIDAS model, rAVGRV is adopted to substitute for the RV estimator. rAVGRV uses the rich data sources in tick‐by‐tick data and significantly corrects the impact of the microstructure noise on volatility estimation. Second, besides introducing macroeconomic variables (i.e., macroeconomic consistency index (MCI), deposits in financial institutions (DFI), industrial value‐added (IVA), and M2), Chinese Economic Policy Uncertainty (CEPU) index and Infectious Disease Equity Market Volatility Tracker (EMV) are introduced in the long‐run volatility component of the GARCH‐MIDAS model. As indicated by the results of this paper, the rAVGRV‐based GARCH‐MIDAS is slightly better than the RV model‐based GARCH‐MIDAS. In addition to the common macroeconomic variables significantly impacting stock market volatility, CEPU also substantially impacts stock market volatility. Nevertheless, the effect of EMV on the stock market is insignificant.

The keywords used in traditional stock price prediction are mainly based on literature and experience. This study designs a new text mining method for keywords augmentation based on natural language processing models including Bidirectional Encoder Representation from Transformers (BERT) and Neural Contextualized Representation for Chinese Language Understanding (NEZHA) natural language processing models. The BERT vectorization and the NEZHA keyword discrimination models extend the seed keywords from two dimensions of similarity and importance, respectively, thus constructing the keyword thesaurus for stock price prediction. Furthermore, the predictive ability of seed words and our generated words are compared by the LSTM model, taking the CSI 300 as an example. The result shows that, compared with seed keywords, the search indexes of extracted words have higher correlations with CSI 300 and can improve its forecasting performance. Therefore, the keywords augmentation model designed in this study is helpful to provide references for other variable expansion in financial time series forecasting.

With the deepening of economic integration, it is important to study the spillover effect of macroeconomic uncertainty among regions along with social network knowledge. Based on microblog users' network connection, this paper analyzes the network factors that affect the uncertain spillover effect between regions by using ERGM. It is found that there are three peaks of uncertainty spillover effect in China, corresponding to the financial crisis in 2008, the post financial crisis in 2010 and COVID-19 in early 2020. As an effective transmission mode and a reflection of inter provincial relations, close social relationships are crucial in the study of uncertainty spillover effects. Relatively developed provinces have higher spillover effect, while relatively underdeveloped provinces are usually the recipients of uncertainty in the network. In particular, the impact of inter provincial social network on uncertainty spillover is statistically more significant than that of GDP, population, transportation and other node attributes.