The RAND HRS Longitudinal File 2018 (V1) is needed to replicate the results for the do files as well as the paper.
The datafile is downloadable at: https://hrsdata.isr.umich.edu/data-products/rand. The dataset is publicly available but creating an account with HRS and honoring the conditions of use are required to download the dataset.
The analytic sample included 57,648 person-wave observations.
The following variables were used in the analysis. The description of these variables are available at the RAND HRS Longitudinal File Codebook (PDF).
hhidpn: The unit of observation is an individual. Each individual is uniquely identified by a household ID (HHID) and a person number (PN). These variables were combined into a single ID variable, HHIDPN. 
shhidpn:  Spouses’ ID variable. 
rabyear: Respondent’s reported birth year. 
agey_m: Respondent’s age. 
ragender: Respondent’s gender.
rahispan: Indicates whether the Respondent is Hispanic.
raracem: Provides the race of the Respondent. 
raedyrs: Respondent’s years of education.
shlt: Respondent’s self-reported of health. 
iwstat: This variable gives the response and mortality status of the Respondent at each wave.
racohbyr: Identifies the cohort a Respondent fits into based on birth year.
raehsamp: Stratum Half-Sample Code.
wtcrnh: Combined Respondent weight and Nursing home Resident weight.
raestrat: Std Error Stratum.