Sampling design, weighting, and attrition management in Korean longitudinal panel surveys: a comparative...

No systematic comparison of Korean longitudinal panel surveys has evaluated sampling design, sample-size determination, weighting procedures, and attrition management within a single methodological framework. This review synthesises these features to support the design of community-based ageing panels.

A structured comparative methodological review was conducted using official statistical registries and national research institute databases. Twenty-two eligible Korean longitudinal panel surveys were identified, and nine panels-five directly relevant and four indirectly relevant to older-adult health research-were selected for in-depth analysis.

Among the six household- or population-based panels, four used stratified two-stage cluster sampling; the Korea Welfare Panel Study used stratified double sampling, and the Korean Retirement and Income Study used unequal-probability systematic sampling. Among the three registry-based panels, the Panel Survey of Employment for the Disabled and the Disability and Life Dynamics Panel used stratified probability sampling, whereas the sampling procedures of the Panel Study of Workers' Compensation Insurance could not be verified from publicly available documentation. Only the Korean Longitudinal Study of Ageing documented an explicit sample-size determination rationale, and none of the nine panels reported design effects. Weighting typically followed three sequential steps: design-weight derivation, nonresponse adjustment, and post-stratification calibration. Panel-specific refinements included weight trimming in the Korean Longitudinal Survey of Women and Families, combined weights after refreshment-sample integration in the Korea Welfare Panel Study, and separate cross-sectional and longitudinal weights in the Korea Health Panel Study. Percentage-based retention rates were available for six panels; the Korean Retirement and Income Study reported only retained household counts, and the Panel Survey of Employment for the Disabled and the Panel Study of Workers' Compensation Insurance reported no retention figures. Refreshment strategies included periodic supplementation in six panels, no refreshment in the Disability and Life Dynamics Panel, full redesign in the Panel Survey of Employment for the Disabled, and an unclassifiable strategy in the Panel Study of Workers' Compensation Insurance.

Korean longitudinal panel surveys have converged on broadly similar sampling and weighting structures, but substantial gaps remain in the reporting of design effects, sample-size rationales, and wave-by-wave retention rates. Future community-based ageing panels should prespecify intracluster correlation coefficients and design effects, provide analysis-specific weights, document attrition-bias diagnostics, and plan refreshment procedures transparently.