Trend Estimation Method Based On Balanced Rotation Sample Survey And It's Application Research

Traditional trend estimation methods generally don't consider the source of data,and regard model error as white noise or even neglect it.However,many important time series data are generated through some sampling surveys,which will lead to overlap and correlation of survey samples,lead to variability of data,and even reduce the effect of trend estimation.In order to improve the effect of trend estimation,this paper considers that the data come from continuous sampling survey.Based on Hodrick-Prescott filtering and considering the form of sampling error in rotation survey,we establish a general sampling error model,trend estimation model and loss function model,and use maximum likelihood method to estimate parameters,fidelity constructed by root mean square error and smoothing roughness are used to evaluate the validity of the model.Through numerical simulation,we compare and analyze the results of traditional trend estimation model and new model,and analyze the tracking degree,curve smoothness,statistical properties of sampling error,mean and variance of fluctuation component and trend component,which shows the effectiveness of new method.Before 2013,the urban household survey in China conforms to the 6-0 rotation sample model.This paper uses the quarterly data of urban household cash consumption expenditure from 2005 to 2013 to verify the trend estimation method based on the balanced rotation sample survey.This paper compares and analyses the traditional trend estimation model and the trend estimation model based on 6-0 rotation sample survey.The empirical results show that,considering the trend estimation of sampling errors,the oscillation amplitude of fluctuation factors converges obviously,and fluctuates slightly around the mean value 0,while trend curves are superior to traditional trend estimation models in fidelity and smoothness.In conclusion,the trend estimation method based on balanced rotation mode has two advantages: first,the sampling error reasonably explain the irregular changes of volatility factors;second,the trend tracking degree and curve smoothness degree are better than the traditional trend estimation model.