What’s New in Version L of the RAND HRS?

Version L incorporates the Early Release for 2010 and the most recent versions of the cross wave Tracker and Region and Mobility files. It also adds new variables and makes adjustments and corrections. The current versions of the core and cross wave data used in Version L are:

  • 1992 Final V1.01
  • 1993 Final V2.1
  • 1994 Final V1.0[1]
  • 1995 Final V2.0
  • 1996 Final V4.0
  • 1998 Final V2.3
  • 2000 Final V1.0[2]
  • 2002 Final V2.0
  • 2004 Final V1.0 (October 2006)
  • 2006 Final Release V2.0 (September 2010)
  • 2008 Final Release V1.0 (November 2010)
  • 2010 Early Release V1.0 (July 2011)
  • Tracker 2010 Final V1.0 (September 2011)
  • Cross-Wave Region and Mobility File V3.0 (/june 2010)
  • Master ID File V5 (December 2009)
  • Cross-Wave Imputation of Cognitive Functioning Measures 1992-2008 V2.0 (November 2011)

We have added the following to the file:

  • New Memory Questions – Alzemhiers(RwALZHE) and Dementia (RwDEMEN): Starting in 2010 (Wave 10) HRS replaced the memory question with two new questions relating to Alzheimers and Dementia. The first question asks 'Has a doctor ever told you that you have Alzheimer's Disease?' and the second asks 'Has a doctor ever told you that you have dementia, senility or any other serious memory impairment?' From Wave 10 we will no longer create the Memory-related variables (RwMEMRY, RwMEMRYQ, RwMEMRYE). Instead we create two new variables: R10DEMEN and R10ALZHE

We have implemented the following updates, improvements, and corrections to the data and documentation:

  • In Tracker 2010 file, HRS updated the following weights, partly due to a change in the CPS weights and partly due to a calculation error.

    1995: Household weight (DWGTHH) and respondent weight (DWGTR)
    2004: Household weight (JWGTHH) and respondent weight (JWGTR)
    2006: Household weight (KWGTHH) and respondent weight (KWGTR)
    2008: Household weight (LWGTHH) and respondent weight (LWGTR)

    According to analyses by HRS staff, these revisions do not substantially alter the weights and do not significantly affect weighted distributions of key variables. Analyses done with the earlier versions of sampling weights should not in general require revision. Some possible exceptions include analyses that present population estimates (counts or percentages) for the 1995 AHEAD cohort or for the EBB non-minority (White/other) sample. Analyses that compare the EBB non-minority cohort to other subgroups may also be affected by the new weights. The full report can be found on the HRS website in the data description for the Tracker 2010 file.

  • Housing Variables (H9AHOUS,H9FHOUS): Our program code mishandled one case where R owned an apartment but did not answer any of the value questions In the previous RAND HRS version K this case was assigned the variable values H9AHOUS=. and H9FHOUS='2.Complete Bracket'. We corrected this so that H9AHOUS has an imputed value of $54,000 and H9FHOUS indicates '5.No value/bracket' in RAND HRS version L.
  • Job Tenure Variables: In the process of creating job tenure variables, we revised our codes not to back-fill the years of job tenure for the previous wave if the answer to the relevant question was missing in the previous wave.
  • Annunity Income Recived Last Year: HRS 2010 added questions related to annuity income received last year. Among respondents receiving income from an annuity, if they reported receiving no income in the last month then they were asked if they had received any income last year. Those who responded in the affirmative were asked for the total amount received in the last year.

    For completeness we list corrections made to the prior RANDHRS Version K dataset, which we posted in April 2010. For six variables (and their spouse equivalents) RANDHRS Version K initially did not take into account relevant information from HRS 2008 in the derivation. The correction resulted in changes for a fairly small number of cases. More specifically the corrections affected the following variables:

    • Smoking variables (R9SMOKEV, S9SMOKEV): The problem affects 72 cases for the respondent variables, and 63 cases for the spouse variables. These cases were incorrectly reported as never having smoked, when in fact, they had smoked previously.
    • Current Marital Status: Without Partnership (R9MSTATH, S9MSTATH): This problem affected 20 cases for the respondent variables, and 37 cases for the spouse variables.
    • Length of current marriage (R9MCURLN,S9MCURLN): This problem affected 20 cases for the respondent variables, and 17 cases for the spouse variables. Most cases are now unmarried and so their marriage length is set to our special unmarried missing value, .U.
    • Length of longest marriage (R9MLEN,S9MLEN): This problem affected 16 cases for the respondent variables, and 13 cases for the spouse variables. Most marriage lengths have been reduced by about 2 years.
    • Number of times divorced (R9MDIV,S9MDIV): This problem affected 11 cases for the respondent variables, and 9 cases for the spouse variables. Most cases have been increased by 1.
    • Number of times widowed (R9MWID,S9MWID): This problem affected 2 cases for the respondent variables, and 1 cases for the spouse variables. Most cases have been increased by 1.This problem affected 2 cases for the respondent variables, and 1 cases for the spouse variables. Most cases have been increased by 1.
    • Miscellaneous:We updated 2010 core data for HHIDPN=033889020, MSUBHH should be 7 and not 1 as recommendend by HRS. Also this same case (HHIDPN=033889020) was not the financial respondent or the family respondent in 2010. The MFINR and MFAMR were coded to 5 as they were missing in the Tracker file.

We have revised Appendix A in this document, which lists details of special cases and effects of changes and corrections.

The RAND/HRS data project is committed to producing high quality data for analysis. To this end, we have employed many innovative programming and quality assurance techniques including paired peer programming, standardized macros, and independent review. If you do, however, notice any undocumented discrepancies or apparent problems with the data, please let us know by e-mailing us (randhrshelp@rand.org).

Though we have attempted to derive measures that are consistent across waves, the underlying HRS data do not always allow this. Some of the native inconsistencies are present in our derived measures, but should be documented in detail in this codebook.

Before using any measure comparatively across interview years, please be sure to read the variable description in this codebook carefully, particularly the sections on "How Constructed" and "Cross Wave Differences in the Original HRS Data" that are included for each variable. If there are cross wave differences that we have not documented, please let us know (randhrshelp@rand.org).

In the future we plan to implement the following changes and additions:

  • Improvements to Income and Wealth Imputations: The improvements to the income and wealth imputations as first applied to HRS 2006 have also been applied to HRS 2008 and HRS 2010. They have not yet been applied to earlier waves. We examined trends over time to assess the compatibility with imputations in earlier waves, and found that the trends (e.g., increases or decreases in income and wealth components) made sense. However, we plan to implement these and other improvements to prior waves in coming releases. Future releases of RAND HRS will also incorporate respondent reports from HRS section U in the wealth data.
  • SSI/DI variables: In future releases we plan to drop the old set of SSI/DI variables that has been superseded by the SSI/DI episode variables. All the information captured in the old set of SSI/DI variables is also recorded in the new set of SSI/DI episode variables pertaining to episode 1. The variables we will drop in future releases are:

    RADIEVER/SwDIEVER Ever applied for SSI or SS Disability (SSDI)
    RADIAPM/SwDIAPM Month and year applied for SSI or SSDI benefits
    RADIREAP/SwDIREAP Appealed or re-applied for SSI or SSDI benefits
    RADIREM/SwDIREM Month and year appealed or re-applied for SSI/SSDI benefits
    RADIGET/SwDIGET Receives approval for SSI or SSDI
    RADIGETM/SwDIGETM Month and year started receiving SSI or SSDI benefits
    RADISABF/SwDISABF Matching SSDI in Disability and Income Sections
    RADITYPE/SwDITYPE Type of disability benefit (SSI or SSDI)
    where w=1 to 10
  • Total Medical Expenditure Brackets: In Version F we added the reported total medical expenditure brackets as categorical variables for Waves 3 to 6 (RwTOTMB), and a version of this variable that imputes complete brackets when needed (RwTOTMBI). The question about total medical expenditures is not asked from wave 7 and forward. The continuous total medical expenditures are imputed in this version of the RAND HRS up to Wave 6, but we plan to drop these imputations in future releases due to concerns about the quality of the imputations based on limited information. We will continue to impute out-of-pocket expenses.

[1] Beginning in Version F, we drop respondents from the 1994 HRS publicly distributed files who are flagged as deceased on the Tracker file. 175 of these 176 dropped cases were actually exit interviews rather than core interviews. The exit interviews were flagged with INW2=2 in prior versions.

[1] We have deleted one case from the 2000 V1.0 file, who was later discovered to be a roommate rather than a partner, according to HRS (January 28, 2005 Data Alert). This case was included in the early release of 2002 but dropped in the final release. We have also changed the HHIDPN for one case from 75573041 to 75573010 according to HRS (November 21, 2005 Data Alert), and adjusted the appropriate spouse ID.

For more information, questions or comments please see our FAQ or email us at RANDHRSHelp@rand.org.