Structural Time Series Modelling And Its Application Of Successive Chinese Labor Force Survey

Labor force survey is a process in which a continuous sample survey method is used to conduct a sample survey of households on a regular basis and estimate the target variable of the population.The labor force survey can comprehensively measure the employment status of a country,and is an important source of labor force information such as urban and rural labor resources,total amount,structure,and distribution of the unemployed and employed population in China.This paper constructs a structural time series model of China’s labor force survey according to the changing characteristics of the population and samples of the labor force survey,which can more accurately estimate the target variables of the labor force survey at the provincial level and below,and can provide a reference for the government to accurately judge the social employment situation and optimize macroeconomic policies in a timely manner.First of all,this paper makes a systematic review of the domestic and foreign labor force survey systems and studies on estimation methods in the literature,and reviews the literatures on sample rotation methods and estimation methods in labor force surveys.On the basis of previous studies,this paper constructs a set of structural time series models applied to China’s labor force survey to estimate target variables.This model represents the series by decomposing the time series data into three parts:the population parameter,the rotation bias and the survey error,so that the estimation of the population parameter is not affected by the rotation bias and the survey error,and improved the estimation accuracy.Secondly,this paper designs a bivariate structural time series model of the labor force survey along with auxiliary information.The auxiliary information such as population census data and administrative records are characterized as high accuracy and authoritative,and is the most widely used auxiliary information source for the current labor force survey.On the basis of the structural time series model,introducing census data and administrative records as auxiliary information,and using of the correlation between the slope of auxiliary variables and the target variable,can improve the estimation accuracy of the structural time series model.In this paper,we established a bivariate structure time series model of labor force survey along with supplementary information.Compared with the univariate structure time series model,the estimation results has higher accuracy under limited samples so that the labor force survey can be conducted in a smaller and more detailed domain,and can provide a richer level of labor information,Thirdly,in order to introducing auxiliary information sources with stronger real-time performance,wider coverage and lower cost,this paper establishes a bivariate structural time series model which use Baidu search index as auxiliary information.Emerging information technologies such as Internet data collection and other big data processing are reshaping public life,in the meanwhile,make it possible to collect and analyze network digital footprints.Massive search behavior data in cyberspace has become a feasible way to effectively measure the public’s attention to labor market’s supply and demand information.Compared with auxiliary information such as census and administrative records data which require more manpower and time to collect and process,Internet data is more timely,covered a wider range,and cost less.Taking Baidu Search Index as auxiliary information,this paper establishes a bivariate structural time series model of labor force survey,which can help government departments to grasp the structural changes and trends of China’s labor market in a more timely and accurate manner.Finally,this paper construct a univariate structure time series model and a bivariate structure time series model along with Baidu search index information to empirically test the national monthly unemployment rate and Beijing quarterly unemployment rate.Both model has high estimation accuracy and good application effect in labor estimation practice.