关键词: 追踪调查, 更新样本, 样本整合, 新生总体, 截面权数

Abstract: In social science research, in order to reveal the patterns of change and development of a phenomenon and effectively identify the causal relationships between phenomena, it is often necessary to follow sample members for a long period, sometimes even throughout their lifetime. Additionally, estimating the overall characteristics of the cross-sectional population is also the major concern. Longitudinal survey data have distinct advantages in studying the change of social patterns and identifying the causal relationships between variables, which is why it has been used widely in social science research. However, as follow-up waves increase, longitudinal sample is threatened by the attrition due to non-response and death of sample members, as well as underrepresentation in the young age group. Moreover, cross-sectional population constantly changes from wave to wave due to immigration, emigration and the inclusion of new-birth population. Consequently, using follow-up samples to estimate the cross-sectional population may introduce potential bias, including non- response bias and coverage bias. To reduce those potential biases, most international longitudinal survey programs regularly or irregularly draw a refreshment sample from cross-sectional population or sub-population. Those refreshment samples are then integrated with on-going sample to estimate the cross-sectional population characteristics. This procedure is known as sample refreshing. Although China has made some progress in developing longitudinal survey programs, there is limited literature addressing this issue in China. This paper aims to summarize the practical experiences of sample refreshing, including the coverage scope and frequency of sample refreshing, refreshment sampling methods, and integration methods for on-going sample and refreshment sample, by reviewing the sampling design schemes and relevant documents from some well-known international longitudinal surveys. This paper also highlights potential problems with existing methods and identifies areas for further research. The goal is to provide practical experience and theoretical support for longitudinal survey practices in China.

Key words: Longitudinal Survey, Refreshment Sample, Sample Integration, New-birth Population, Cross-sectional Weights