人口抽样数据偏误研究:测度、成因及修正

Study on the Bias of Population Sampling Data:Measurement,Cause and Correction

采用抽样年均增长率、普查年均增长率之差测度人口抽样数据偏误,该指标适用于大量观测值的跨城市、跨年份比较。在此基础上,构造回归模型研究人口抽样数据偏误的成因。发现城市的行政等级显著影响人口抽样数据偏误,主要表现为行政等级越高则低估偏误越强。经济发展、产业结构、政府干预、公共服务也显著影响人口抽样数据偏误,普查年人口对人口抽样数据偏误的影响则不显著。进一步地,研究提出一种人口抽样数据偏误的“事前”修正策略。通过回归分析考察普查年均增长率的决定,由此得到的普查年均增长率的拟合值可用来计算此后若干非普查年的人口。研究具有重要的方法论价值,有助于提高非普查年人口数据的质量。

This paper uses the difference between the average annual growth rate of sampling and the average annual growth rate of a census to measure the bias of population sampling data,which is suitable for cross-city and cross-year comparisons of a large number of observations.On this basis,a regression model is constructed to study the causes of the bias of population sampling data.It is found that the administrative hierarchy of a city has a significant effect on the bias of population sampling data,which is mainly reflected as the higher the administrative hierarchy,the stronger the underestimation bias.Economic development,industrial structure,government intervention,and public services also significantly affect the bias of population sampling data,while the impact of population in the census year on the bias of population sampling data is not significant.Furthermore,an“ex-ante”correction strategy for the bias of population sampling data is proposed.The determination of the average annual growth rate of the census is examined by regression analysis,and the fitted value obtained can be used to calculate the population of subsequent non-census years.Research in this paper shows the important methodological value and helps to improve the quality of non-census-year population data.