Page 204 - Contributed Paper Session (CPS) - Volume 2

CPS1824 Sanggi L.

Cross-sectional weighting methodology for the
Korean population census longitudinal study
Sanggi Lee
Statistics Korea, Daejeon, Korea

Abstract
With introducing of registered census in 2015 we are providing a three-year
Korean Population Census Longitudinal Study(KPCLS) from 2015 to 2017. In
the general household panel surveys, the method of assigning the initial
weight to the household members newly entering the household after wave 1
as the wave progresses is problematic. This paper shows several approaches
to solve this issue and a methodology that can be applied to the KPCLS
considering the characteristics of being based on administrative data.

Keywords
panel; cross-sectional weight; longitudinal study; shared weight

1. Introduction
There was a paradigm shift in the 2015 Korea Population and Housing
Census unlike before. Shortform was performed by using about twenty four
administrative data and long-form was performed by field survey. Short-form
has basic items of population, household and housing as sex, age, type of
household and type of house, etc. Long-form was performed by face-to-face
interview including the internet survey. As the census is produced every year
based on the administrative data, the census longitudinal data can be
constructed every year. We are currently providing a three-year Korean
Population Census Longitudinal Study(KPCLS) from 2015 to 2017, and plan to
link it every year in order to increase the utilization of the census data.
In section 2, we briefly explain the sample design and following rules of
the KPCLS. In section 3, we explore the theoretical discussions on the
weighting method of new household entrants in the panel survey and discuss
ways we can apply to the KPCLS.

2. Summary of the KPCLS

2.1 Sample Design
The KPCLS was selected from the general households in 2015 census by
about 2%, and it is produced and managed according to the following rules
every year. It was stratified by 252 regions and households was selected
directly by strata by balanced sampling method(cube method). The variables

193 | I S I W S C 2 0 1 9