Research On Some Problems For Household Surveys

Household survey is the general term for all kinds of surveys organized to collect various social-economic statistics,and is the source of the implementation and evaluation of the policy of "improving the quality of employment and people's income" reported by the Party's 19th Communist Party of China National Congress.Under the background of comprehensively promoting the integration of urban and rural household surveys,it is of great significance for the formulation and implementation of national macroeconomic policies to carry out research on household survey methods and improve the sampling design system of existing household surveys.At present,almost all countries are actively carrying out household surveys.China has conducted household surveys since the 1950s,quarterly and monthly labor force surveys since November 2005,and integrated urban and rural household surveys nationwide since 2013.The theory and practice of China's household surveys have made great progress,but compared with the actual needs of China's social development,it still needs to be improved on the following aspects:(?)the construction of parameter estimators needs more auxiliary information to improve the accuracy of estimators,meet the needs of building a digital economy with data as the key element,and improve the level of national modernization governance.The estimation of parameters is mainly determined by the variable observation value and the weight of the variable observation value.The accuracy of the weight has great influence on the final estimation result.Using auxiliary information to adjust the weight has become the mainstream.(?)Variance estimation of estimators is the standard to describe the accuracy of estimators.How to better combine the specific variance estimation method with the specific scheme of China's household survey needs further demonstration.Variance estimation of estimators often adopts variance estimation methods of complex sampling,such as Successive Difference Replication method,Bootstrap method,Jackknife method,etc.,which is determined by the complexity of sampling method and estimator construction process.The sampling method usually adopts a combination of stratified sampling,multistage sampling,the probability proportion to size(PPS)method and random systematic sampling.The forms of estimators such as calibration estimators,post-stratification adjustment estimators,regression estimators and ratio estimators are complicated.How to apply the variance estimation method of complex sampling needs further analysis.(?)The solution to non-sampling error needs further improvement,especially for non-response error.In the process of investigation,the occurrence of non-response tends to increase bias of an estimator and reduce the credibility of statistical analysis results.Weighted adjustment method and imputation method are the main methods to deal with non-response questions.Constantly improving administrative record data provides more auxiliary information for non-response of household surveys,and new theoretical models need to be found to give more reasonable imputation values for non-response.How to effectively reduce the impact of non-response error on the final estimation is an urgent problem to be solved in China's household surveys.(IV)Under the background of continuously improving the scientization of government decision-making and the precision of social govermance,the determination of sample size needs more in-depth research.According to the stratified multistage sampling method currently used in China's household survey,it is of practical significance to explore the method of determining sample size in consideration of cost and accuracy requirements.Based on the above,the research contents of this paper are as follows:(?)Giving the construction method of calibration estimator,including the determination of basic weight,unanswered weight adjustment,calibration weight adjustment,etc.combining with the sampling scheme of China household surveys.The integrated weighting method used in the adjustment of the calibration weight can realize the simultaneous calibration of the weights of the primary sampling unit and the secondary sample unit.The research results show that the mean square error of the estimator is small and the estimation accuracy is high.(?)Comparing and analyzing the applicable conditions of each method from the perspective of the theory and application of variance estimation methods for complex sampling.Combined with the specific situation of China's household survey,the variance of calibration estimators is estimated by Successive Difference Replication and Bootstrap.The results show that the estimation of Successive Difference Replication is more robust than that of Bootstrap.Because the Successive Difference Replication can effectively use the sequence information on systematic sampling adopted by China's household surveys.(?)For the item non-response in non-sampling error,this paper proposes the Response Propensity Score Matching Imputation method by basing on the defect of the Propensity Score Matching Imputation and discusses its statistical properties through simulation research.Compared with other imputation methods,the Response Propensity Score Matching Imputation method has better estimation effect.(IV)For the problem of determining the sample size of stratified multistage sampling,giving the functional expressions of determining the sample size of stratified two-stage and stratified multistage sampling under the condition of certain cost or accuracy,which can be conveniently applied to sampling survey process of the futureChina is in a critical period in comprehensively building a well-off society,the development is facing complicated and severe situations at home and abroad that seldom occur in these years and the economy is under new downward pressure.These put forward more and more demands on the government statistical investigation.From the perspective of household surveys,this paper focuses on the construction of calibration estimators,variance estimation of calibration estimators,non-response and sample size determination.When more and more auxiliary information is obtained,the calibration estimator and its variance estimation can effectively improve the accuracy,and the Response Propensity Score Matching Imputation method can give more reasonable imputation values for non-response and the determination of sample size will be more accurate.The research work in this paper can promote the improvement of China's household surveys system and provide more accurate data for the country to maintain sustained and healthy economic development and implement stable policies for the overall social situation.