Explaining the HRS in Pictures
Rich Jones
2023-11-10
The analyses in this report are all based in the 2020 tracker file
(trk2020tr_r.dta). The first figure is lifted from the HRS
web site, the rest are new.
Survey Design
This graphic depicts the longitudinal cohort sample design of HRS. The initial 1992 HRS cohort consisted of persons born 1931 to 1941, who were then aged 51 to 61, and their spouses of any age. Members of this first HRS cohort, now in their 80s and 90s, have been interviewed every two years since 1992. A second study was added in 1993, the Asset and Health Dynamics Among the Oldest Old, or AHEAD, which captured those born before 1924, who were 70 and older at the time. Then in 1998, the HRS and AHEAD cohorts were merged, and two new cohorts were enrolled to bridge the study age gaps for Americans 50 and older. These birth cohorts are the Children of the Depression, or CODA, for those born 1924 to 1930, and the War Babies, for those born 1942 to 1947. HRS now employs a steady-state design, replenishing the sample every six years with younger cohorts. In 2004 HRS added the Early Baby Boomers (EBB), born 1948 to 1953, and in 2010 added the Mid Baby Boomers (MBB), born 1954 to 1959. The Late Baby Boomers (LBB), born 1960 to 1965, were added in 2016. Early Generation X (EGENX), born 1966-71, will be enrolled in 2022. For all cohorts, both members of a couple are included in the sample.
Follow-up
The following graph is a line chart. There are 40,172 lines drawn, where 40,172 is the number of unique sampled persons in the 2020 tracker file. My definition of a sampled person is a person with a sampling weight and valid value for age at interview. The lines are sorted according to the year of first interview and then year of last interview, so you can see how long a person remained in the HRS study. I have truncated follow-up at 2018 because the 2020 weights were not available in the version of the tracker file I used.
Because of how people were enrolled and the way the data are sorted for this plot, we can see the size of the relative cohorts and the distribution of how long people remain in the HRS. The chunking of people refers to the enrollment cohorts (HRS, AHEAD, CODA, War Babies, etc.) If it looks like there is no line (like right above y-axis point labeled 17500) there really is a line, but it is actually also a point that indicates people assessed only in 1998 (no follow-up).
Age Distribution Across Survey Year
Two ways of looking at the age distribution across survey year.
Strip plots
Histograms
Non-sampled persons
The HRS also interviewed spouses. Sometimes spouses are sampled persons, sometimes not. The way I tell is if a person has a weight of 0 or a missing weight, they are a non-sampled person. Sometimes, non-sampled persons become sampled persons (e.g., they age into age-eligibility or a new cohort is opened and they are age eligible for that cohort). The graphs below are for non-sampled persons. The limited duration of follow-up for non-sampled persons could be truncated because the non-sampled person became a sampled person.
First interviews and follow-up
Age distribution at first interview
