Harmonized life histories and childhood conditions from the China Health and Retirement Longitudinal Study (CHARLS)
Written by: Morten Wahrendorf and Christian Deindl
Published on: Apr 25, 2023
The China Health and Retirement Longitudinal Study (CHARLS) is a longitudinal household survey interviewing people aged 45 and over (and their partner) in China. Apart from their regular survey waves CHARLS conducted a separate life history interview in 2014, similar to its sister studies in England (ELSA), Europe (SHARE), and the United States (HRS).
Life history data is an easy way to collect information over past experiences in a respondent’s life. In contrast to “normal” data collection, where respondents are asked about their current situation and where surveys follow them over a couple of years, in life history interviews respondents are asked about events and circumstances in the past. CHARLS collected a wide arrange of life history information that in some instances exceeds the information provided by SHARE, ELSA, and HRS. CHARLS provides information about the demographic background, family information, education history, health and health care history, wealth history, and work history. For the harmonized CHARLS life histories we used information that were similar to the ones in SHARE and ELSA that is the number of children, the partnership status, the Houko category, employment status, and childhood information, when respondents were 16 years of age.
This data is available in two types. We have childhood information and we have sequential data. The sequential data gives yearly information about employment, housing, health, partnership, and children from the age of 15 till the age of 80. Childhood information contains information about the socio-economic situation and the health of CHARLS respondents when they were around 15 years old.
The original data provided by CHARLS is in a wide spell format that is rather complex. For example, in the case of employment, the data provides information on all jobs the respondents had in the past, with varying number between respondent. This includes variables that describe the year when a spell started and ended together with variables that provide details on the respective job. To facilitate the use of life history data, we have rearranged this spell-information in a more intuitive state sequence format with one state variable for each age (e.g. work state at age 20, 21, 22, 23, 24, …..). In contrast to the original, this format directly allows to describe conditions at a specific age without need of extensive data management and is the standard format for sequence analyses.
Harmonized life history data provide annual information from age 15 to age 80 in the following five domains: Children: total number of children, total number of children below age 18; Partnership: whether respondent lived with a partner or lived alone; Houko: type of Houko (Agricultural, Non-agricultural, Unified Residence, None); Work: whether in paid work or not working - distinguishing ten different work states. Figure 1 displays the changing proportions of these ten different work states by age for both men and women. This chronogram highlights that certain labor force states are more common at different points in the life course, for example agricultural employment in China is observed most often at ages younger than 50, and that some labor force states are more common for specific genders, for example non-agricultural employment is much more common for men than for women. If need be the do-files can be easily modified according to specific research questions.
CHARLS also offers detailed information about the situation during respondent’s childhood. Among the great variety of questions about childhood conditions, we picked the closest to SHARE and ELSA. Childhood conditions encompass the socio-economic background: that is the occupation of the male or female guardian before respondents were 17. In the childhood health section, we include variables about different dimension of health including vaccination when respondents were below the age of 15.
- Morten Wahrendorf is a Senior Medical Sociologist at the University Düsseldorf.
- Christian Deindl is a Sociologist at TU Dortmund University.