Research On Application Of Network News Data In Macroeconomic Cycle Identification

Identification of macroeconomic cycle is an important issue in Macroeconomic research field.Macroeconomic usually shows a cyclical change of “prosperity,recession,depression and recovery”.Since the reform and opening-up,China's macroeconomic has experienced a period of rapid expansion of 10% a year,with the continuous expansion of volume,the economic growth rate has remained around 7% a year,which named the “New Normal” stage.Since Burns and Mitchell(1964)put forward the widely accepted definition of macroeconomic cycle,scholars both at home and abroad have made a multiangle empirical study of economic cycle,such as the measurement and prediction of business cycle,the fluctuation characteristics of business cycle,the international business cycle and the influencing factors of economic cycle,etc.In those related research literatures,researchers usually use the data of macroeconomic variables,which are released with a certain lag.With the development of the Internet,the network data has gradually become an important real-time data source used by scholars for theoretical research and practical applications.In view of the timeliness requirement in identifying economic cycle,network news data are used in this paper.Obtain the financial news about 126,656 from January 1,2002 to September 30,2017,which published in the website of Hexun,Sina Finance and Sohu Finance as the original corpus.Using the Latent Dirichlet Allocation Model(LDA model)to extract the 30 topics reflected in the corpus,and calculate the emotional values for each topic.After Screening,7 topics are left as the Target topic,and the common factor can be extracted by the dynamic factor model(DFM),named WEB-INDEX.By A comparative analysis of WEB-INDEX and the quarterly year-on-year growth rates of real GDP,one can see that the WEB-INDEX is coincident with the GDP in the most time,but will have a leading trend in some certain points,therefor WEB-INDEX can be used for the study of economic cycle recognition.Based on the WEB-INDEX and quarterly year-onyear growth rates of real GDP,the MS-VAR model is applied to identify the macroeconomic performance of China from 2002 to 2017,we can get that in this period the macroeconomic has experienced two expansions,one is the soft economic expansion from the third quarter of 2002 to the fourth quarter of 2007,and the other is the economy recovery from the second quarter of 2009 to the second quarter of 2011.At the same time,China's macroeconomic has also gone through two contractions,the period from the first quarter of 2008 to the first quarter of 2009,which is caused by the financial crisis,the period from the third quarter of 2011 to the third quarter of 2017,which is caused by the European debt crisis and the economic structure adjustment.AS a benchmark model,the single variable Markov switching model is used to identify the macroeconomic situation of China.By comparing the result of two methods,one can find that after adding WEB-INDEX,the model can identify the economic turning point more timely,when the economy shifts from expansion state to contraction state,it leads 3 quarters,when the economy shifts from contraction to expansion,it leads 1 quarters,which indicates that it is feasible to apply the network data to the macroeconomic cycle identification.At the same time,the result also reflects that the self-transition probability in the expansion period is 0.9351,the average duration is 15.41 quarters,and the self-transition probability in the contraction state is 0.9620,the average duration is 26.32 quarters,which can prove the asymmetry features of the macroeconomic.