Study Name: Health and Retirement Study (HRS)

 

Study Director:Robert J. Willis

 

Principal Investigators: David Weir, Charlie Brown,Alan Gustman (Dartmouth College), John Henretta (University of Florida), A. Regula Herzog, Daniel Hill, Michael Hurd (RAND), F. Thomas Juster (Emeritus), Mark McClellan (Stanford University), Olivia Mitchell (University of Pennsylvania), Willard Rodgers, Thomas Steinmeier (Texas Tech University), Beth Soldo (University of Pennsylvania), Robert Wallace (University of Iowa)

 

Host Organization: Survey Research Center

Institute for Social Research

University of Michigan

 

Year Initiated: 1990

 

Governance:

 

Funding sources.The National Institute on Aging (one of the National Institutes of Health) sponsors the HRS. The Social Security Administration, the US Department of Health and Human Services Assistant Secretary for Planning and Evaluation, and the US Department of Labor Pension and Welfare Benefits Administration provide supplemental funding support for the HRS.

 

Host organization.The Institute for Social Research at the University of Michigan conducts the study.

 

Governing body and role of external research advisors.While the Principal Investigator together with the co-investigator group is responsible for the overall direction of the study, the HRS is a Cooperative Agreement and receives much input from the NIA Program Officer.The co-investigator group is in near constant contact via email technology and weekly conference calls.Additionally, the HRS receives advice and oversight by a Steering Committee and a Data Monitoring Committee.

 

The Steering Committee is charged with providing timely advice to the PI about all relevant matters of concern to HRS. This includes laying out tasks to be performed by the expert working groups, monitoring progress of the study, and making recommendations about sample design, survey content and operational procedures. This committee formally meets once or twice a year with the Principal Investigator, Co-Investigators, and Working Group Chairs and Coordinators.

 

The NIA Data Monitoring Committee is an advisory group comprised of independent members of the academic research community as well as representatives of agencies interested in the study. The structure of the Data Monitoring Committee allows and encourages the coordination of federal statistical operations, which increases the value of the study to both federal and research communities. One result of the unique configuration of the Data Monitoring Committee has been much needed support in terms of assistance with the administrative linkages efforts of the study, and the provision of substantial additional funding. This committee provides advice and recommendations to the NIA Program Officer.

 

Sample Design:

 

Sample selection.The HRS originated in 1992 with a national sample of 12,654 individuals (approximately 7,700 households) age 51-61 and their spouses, generated from a national screen of over 70,000 households.Blacks and Hispanics were oversampled at a rate of 2:1.We also oversampled residents of the State of Florida.

 

In 1993, the HRS companion study of Asset and Health Dynamics Among the Oldest Old, or AHEAD, was initiated to study those aged 70+ in 1992.The AHEAD was partially based on the HRS screening effort, and partially on a list sample from HCFA for those over the age of 80.As this study was based on the HRS sample, the same groups were oversampled.

 

In 1998, we combined the HRS and AHEAD into a single panel with a single instrument, andadded about 6,000 new households representing the 1924-1930 and 1942-1947 birth cohorts, making the HRS a single panel representative of the US population over the age of50.

 

Our long-term study design calls for the enrollment of a new six-year birth cohort every six years.The next new cohort will be brought into the study in 2004, representing those born between 1948-1953.

 

Follow rules.Respondents are followed through death.We also conduct exit-proxy interviews to obtain information about cause of death, end of life and estate settlement.

 

Weights and attrition bias.Sampling weights are constructed at the household and respondent levels based on the initial selection probabilities of the units with several additional correction factors for subsampling of secondary sampling units in special circumstances.The sampling weights are then adjusted to correct for initial nonresponse and subsequent panel attrition by post-stratification to the Current Population Survey each wave.The post-stratification factors employed reflect basic sample design and include age of respondent and spouse and race/ethnicity.

 

Sample “refreshing”.Each six years, a new cohort consisting of the next six birth year cohortswill be enrolled into the HRS.The next cohort enrollment is currently scheduled for 2004 and will consist of non institutionalized members of the 1948 – 1953 birth cohorts and their spouses.

 

 

Content:

 

Driving policy needs.

The predictable rapid aging of the U.S. population, starting in the early part of the 21^st century and continuing through the next several decades, creates a need to understand what will happen to labor force participation rates, asset accumulation patterns, demands on the health care system, and intergenerational transfers, all of which will influence theconsumption path of aging households.Although there will be much disagreement about how to organize public and private transfers so as to mitigate economic hardship, there would be no disagreement with the proposition that the growth rate of real output will be a critical determinant of how successfully the U.S. can accommodate population aging, and how output growth rates will in turn be related to saving behavior and labor force participation.

 

Research objectives (maintaining coherence of content domains).

The main research objectives are to create the capacity to model critical dependent variables such as labor force participation rates, savings rates, and demands on the health care system.To do that effectively requires comprehensive measurements of labor force participation, economic status (income and wealth), health status including both disease conditions and functioning as well as cognitive problems, and intergenerational transfers.

 

Benefits of international comparability.

The major advantages of collecting internationally comparable data is that it creates the opportunity to make much more precise estimates of the impact of policy changes, since policies will predictably differ among countries and those differences will provide leverage for modeling differences in outcome.If intra-national differences were all that we could observe, identifying the impact of, say, differences in pension arrangements on labor force participation rates would be much less robust than if we had cross-national data to look at, including information on countries with very different pension systems.

 

Content decisions (income, wealth, health, employment, family history and data linking).

Over the course of the last decade, the content of HRS has focused on the measurement of income and employment (aided greatly by access to Social Security earnings records), income and wealth (aided greatly by methodological enhancements in the measurement of both variables), health conditions, health status, and health insurance coverage (including a substantial effort to understand cognitive changes), intergenerational transfers, family structure, and a set of expectations expressed as probability scales.

 

Tradeoffs between continuity and the incentives for new directions.

HRS is now quite close to a survey content model in which the major content areas are measured without change in the four major domains (economic status, employment, health, and intergenerational transfers), with any additions taking the form of off-year mailouts.The major area for potential content changes is probably in the elimination of variables that seem not to have much power in explaining behavior.That would require a systematic assessment of how the current variables are used by the scientific community, which variables seem to be largely ignored, etc.

 

 

Collection:

 

Mode.The 1992 HRS baseline data collection was a face-to-face interview using paper and pencil questionnaire.Thereafter, the majority of interviews are conducted over the telephone using CAI technology. In some instances, where respondents’ circumstances call for it, face-to-face CAI is used, as well as in new baseline interviews for new cohort enrollment.

 

Instrument design (paper and pencil, CAI).The study first used computer-assisted interviewing in 1993. Now, the majority of all interviews are conducted using computer-based instruments.However, we are proposing to conduct off-year mail surveys to address additional content and content that is better suited to this format.

 

Dependent interviewing.In order to streamline the interview as well as to minimize false positive measures of change, a number of household and respondent characteristics from prior waves are preload into the CAI application and used to control the flow of the interview.Among these are employer identification, family listing information and asset ownership and value.These measure improve the quality of the data while simultaneously speeding up the interview by eliminate unnecessary questions.If the respondent still works for the same employer, for instance, we need not re-ask the respondent about employer specific measures.In the near future we plan to include an asset change reconciliation module that will flag large unexplained changes in assets from the prior wave and prompt the interview to seek clarification.

 

 

Processing:

 

HRS Post-interview Processing.Post-interview processing starts with the coding of open-ended questions while data is being collected, followed by a systematic review and selected correction of data based on respondent comments recorded by the interviewers. A database containing the metadata for the particular wave being processed is produced from the specifications of the program (currently SurveyCraft) used to collect the data.This database holds the information about masking, variable deletions, formats, variable label and question text, the analysis level of each variable, and interview flow.The codebook is created from the metadata database and variable frequency files, followed by further review and checking of data, elimination of wild codes and empty data cells.Final steps involve splitting the data into household and respondent levels, creating ASCII data files with input state-ments for SAS, SPSS, and STATA,and double checking the accuracy ofIDs and basic flow control variables.

 

HRS is in currently involved in improving and refining the process of preparing data to release to the public and is building into the Blaise specifications (the CAI program for the 2002 data collection) as much as possible the elements needed to simplify and reduce post-interview processing.

 

Production of Specialized files.In addition to the core household, respondent, and “other person” level files, the HRS produces and number of specialized files.The cross-wave tracker file is available as a public release dataset and contains one record for each individual who has every been a respondent in any of the entry cohorts or anyone who has been married to or partnered with such a respondent during the study period.In addition to basic identification variables this file contains cross-wave information on response and vital status, National Death Index year of death, and sampling weights for each wave for respondent and household level analysis.

 

Dissemination:

 

Dissemination techniques (web, restricted data diskettes, secure data facility).All HRS public use datasets and their supporting documentation are available for download from the HRS website (http://www.umich.edu/~hrswww/).In addition, the HRS makes available preliminary datasets within three months of the close of the field period.These preliminary datasets are available via private FTP after completion of a registration process, and are only supported limitedly.Although the HRS public use data have always been freely available for download by the public via the world-wide-web, recently the HRS Steering Committee strongly recommended that we implement a policy of mandatory registration for the public-use data sets.We will be moving to a web-based registration format in the very near future.Mid-term plans include expanding the web-based registration to include the preliminary release files.

 

Due to the wide array of data and the longitudinal nature of the study, HRS has restricted access to some data in order to protect the privacy and confidentiality of the respondents.These data are typically made available via a rigorous application process resulting in a data use agreement with the University of Michigan.Additional data in this restricted category come from administrative linkages and other sources, such as the Social Security Administration, the Health Care Financing Administration, and employers.The HRS, in conjunction with several other funded projects, has established a secure data facility to broaden access to the restricted datasets.We are exploring ways to eventually implement a system for encrypted online delivery of sensitive data files, as well as extend access to our restricted data.

 

Value of joint multinational analysis projects.

The HRS has become a standard for studying retirement and aging.Similar studies have been or are planned to be conducted in Amsterdam, Mexico, Australia, England, and Sweden.

 

 

 

 

 

 

 

V:\REFDOCS\Descript\HRS overview for Oct 2000 wkshp.doc