A study on construction of inflation index based on web search data

摘要

In recent years, with the rapid development of network technology, network technology and social behaviour have crossed and integrated deeply. Web search data contains a huge search concerns and needs, which provides the necessary database for the study of social and economic behaviour. This paper applies web search data to the construction of inflation index. To begin with, the selection of key words; then, an empirical analysis based on the use of the Baidu index of access to the keywords search data. Empirical analysis is divided into three parts:(1) key words classification;(2) construct the leading inflation index;(3) establish inflation forecasting model. The results show that the web search data exist co-integration with CPI; second, the leading inflation index constructed using principal component analysis with strong timeliness, about 4-8 months earlier than the data released by the National Bureau of Statistics; furthermore, the forecast model passes the equation of goodness-of-fit test, the test of significance and heteroscedasticity test, with a certain degree of scientific and effective.