大型追踪调查样本流失规律及模式——以中国家庭追踪调查为例  

The Patterns and Trends of Sample Attrition in Large-scale Panel Surveys——A Case Study of China Family Panel Studies(CFPS)

作　　者：孙妍 王赫 潘修明 张春泥[2] Sun Yan;Wang He;Pan Xiuming;Zhang Chunni(Institute of Social Science Survey,Peking University;Department of Sociology,Peking University)

机构地区：[1]北京大学中国社会科学调查中心 [2]北京大学社会学系

出　　处：《调研世界》2025年第3期63-76,共14页The World of Survey and Research

基　　金：教育部人文社会科学研究一般项目“追踪调查样本流失模式及维护策略探索性研究”(18YJC840034);教育部重大课题攻关项目“新中国成立以来社会调查发展进程、经验与展望研究”(20JZD032)的资助。

摘　　要：本文基于中国家庭追踪调查2010—2022年七轮数据,分析家庭样本流失模式及变化趋势。研究结果表明,以截面视角的流失率来衡量,家庭样本随着追踪期数的增加,流失规模逐渐扩大,但流失速度呈现前快后缓的特征。从追踪视角来看,最常见的流失类型是样本家庭在不同轮次间处于失访与完访状态间来回切换的摇摆流失,这意味着将曾经失访的样本保留在追踪库中有助于减少样本的损耗。社会外部环境的变化导致的访问模式转换,也会引发样本快速流失。失联和拒访是导致样本流失的主要因素,对于因不同原因失访的家庭,后续挽回的难度也有差异。对于失联流失的样本,挽回的难点在于获取有效的联系方式;而对于拒访流失的样本,挽回难度较大。家庭规模较大、居住流动性较低、农村和低收入家庭更倾向于全程参与调查,受访者对往期轮次调查的态度也对后续流失有一定影响。在实践中,综合运用往期追踪状态划分的流失组别能够较好地预测下一轮的完访率,能够为样本维护及实地执行提供有效参考信息。To better understand the characteristics and trends of sample attrition in household panel surveys in China,this paper analyzes data from the China Family Panel Studies(CFPS)spanning from 2010 to 2022.The results indicate that the scale of cross-sectional sample attrition gradually increases with the number of follow-up waves,and the attrition rate of CFPS household samples initially experiences a rapid increase and then stabilizes.Among the non-fully tracked sample households,the predominant type is swing attrition,where households toggle between non-response and response status across waves.This implies that retaining non-responsive samples can help mitigate sample loss.Changes in interview modes,driven by shifts in the external social environment,can also contribute to rapid sample attrition.Loss of contact and refusal to participate are the primary factors leading to sample attrition.The difficulty of conducting follow-up interviews in subsequent waves depends on the reason a household did not receive a previous interview.For sample attrition due to loss of contact,the challenge lies in obtaining effective contact information,while for samples lost due to refusal to be interviewed,the difficulty is even greater.Households with larger sizes,lower levels of residential mobility,and those from rural and low-income backgrounds demonstrate a greater propensity to participate in the panel survey.Additionally,respondents’attitudes toward the surveys in previous waves also influence subsequent attrition.Practically,utilizing attrition groups classified by previous response status can yield more accurate predictions of follow-up rates in the next wave,thereby providing valuable and effective reference information for sample maintenance and on-site implementation.

关 键 词：追踪调查 样本流失 序列聚类分析 中国家庭追踪调查 

分 类 号：C31[社会学]