| September 2003
Volume 4, Issue 3
This newsletter features articles about two MRRC data enhancement projects. One describes a version of the HRS that researchers at RAND have created that makes the data more easily accessible to a variety of users. SSA has supported this endeavor directly as well as through the MRRC. The second article describes the development of the Pension Calculator Program, a computer program that uses data on pension plans from the HRS – data both from respondents and from their employers – to generate information on respondents’ pension assets. Because of the overwhelming complexity of the private pension system in this country, correct information on pension stocks and flows is very difficult to obtain but is critical to have in order to model accurately retirement decisions and related topics.
Social Security solvency and changes that will lead to long-term solvency
remain, of course, paramount research topics. The MRRC has consistently
argued that scientific research, relying on the most modern data sources,
will provide the best foundation for developing and assessing Social Security
reforms. The Rand/HRS data accessibility project and the Pension Calculator
Program project exemplify several of our recent efforts to strengthen
the basis for such research in the future.
Social Security’s Latest Fast Facts & Figures Publication Now Available Did you know that: more than 50 million people receive some kind of benefit check from Social Security Social Security provided at least half the income for 64% of aged persons the number of people aged 65 or older receiving Supplemental Security Income has declined from 61% to 30% since 1974 These facts, and much more information about Social Security, can be found in the 2003 edition of Fast Facts & Figures About Social Security. This annual chartbook, produced by the Social Security Administration, is a handy resource for a reporter needing information, a student working on a term paper, or anyone interested in learning more about Social Security programs. Visit www.socialsecurity.gov/policy to find the latest edition of Fast Facts & Figures About Social Security. Paper copies of the chartbook may be ordered by e-mail at firstname.lastname@example.org or by phoning (202) 358-6274.
back to the top.
Abstract: This paper explores how earnings variability is related to retirement wealth. Past research has demonstrated that the average American household on the verge of retirement would need to save substantially more, in order to preserve consumption flows in old age. While several socioeconomic factors have been examined that might explain such problems, prior studies have not assessed the role of earnings variability over the lifetime as a potential explanation for poor retirement prospects. Thus two workers having identical levels of average lifetime earnings might have had very different patterns of earnings variability over their lifetimes. Such differences could translate into quite different retirement wealth outcomes. This paper evaluates the effect of earnings variability on retirement wealth using information supplied by respondents to the Health and Retirement Study (HRS). This is a rich and nationally representative dataset on Americans on the verge of retirement, with responses linked to administrative records from the Social Security Administration. Our research illuminates the key links between lifetime earnings variability and retirement wealth.
The Annuity Puzzle Revisited, Hugo Benítez-Silva
Abstract: There is a pressing need for a better understanding of how access to various types of financial products can impact retirement behavior, especially if this access comes from a change in the incentive scheme through a reform of the current Social Security system. This is especially important if we are to provide useful policy recommendations regarding reform to the current social insurance system. In this paper I focus on the “annuity puzzle,” the question as to why the annuity market is so narrow. I present a model that endogenizes the annuity decision along with the consumption/saving and labor supply decisions. This research enhances our understanding of how annuities work in a life cycle model with more realistic characterizations of the choices and incentives that individuals face. My results show that the low rates of annuitization can be the product of optimal decision making by individuals in a life cycle model which endogenizes the labor/leisure decision and accounts for Social Security. The government should pay particular attention to the rules regarding withdrawal of benefits through annuities or lump-sums when introducing individual retirement accounts or other privatization schemes, given the interaction between retirement incentives and the attractiveness of annuities.
Abstract: Economists’ most basic model for studying Social Security policy issues is the so—called life—cycle model of saving behavior. This paper sets up a life—cycle model in which a household simultaneously chooses its lifetime consumption profile and retirement age. The paper calibrates parameters from Consumer Expenditure Survey data, making special use of observations of changes in household consumption immediately following retirement. The paper’s last section provides illustrative simulations of the effect of current Social Security provisions on average retirement ages, finding evidence of a several year reduction in working life in some cases.
Social Security, Retirement and Wealth: Theory and Implications, Miles S. Kimball and Matthew D. Shapiro
Abstract: The effect of Social Security rules on the age people choose to retire can be critical in evaluating proposed changes to those rules. This research derives a theory of retirement that views retirement as a special type of labor supply decision. This decision is driven by wealth and substitution effects on labor supply, interacting with a fixed cost of working that makes low hours of work unattractive. The theory is tractable analytically, and therefore well-suited for analyzing proposals that affect Social Security. This research examines how retirement age varies with generosity of Social Security benefits. A ten-percent reduction in the value of benefits would lead individuals to postpone retirement by between one-tenth and one-half a year. Individuals who are relatively buffered from the change—because they are wealthier or because they are younger and therefore can more easily increase saving to offset the cut in benefits— will have smaller changes in their retirement ages.
The Social Security Reform Process in Italy: Where Do We Stand, Agar Brugiavini and Vincenzo Galasso
Abstract: A reform process is under way in Italy. Achieving financial sustainability of the social security system has been the first objective characterizing the reforms of 1990s, but these have also introduced rules which aim at a more actuarially fair system. Indeed the social security system prevailing in Italy, financed on a PAYG basis, was, at the end of the 1980s, clearly unsustainable and also extremely unfair to some group of workers, enacting a form of perverse redistribution which is typical of “final salary” defined benefit systems. It was also a system characterized by strong incentives to retire early. In this paper we briefly describe the different regimes of the Italian pension system in its recent history and focus on some aspects of the reform process taking place during the 1990s. Since economists and policy makers are still struggling to assess the results and the long-term effects of these reforms we provide both a survey of this debate and some fresh evidence on the evaluation of the policy changes. We carry out this analysis with a particular emphasis on two aspects, which are relevant in the debate. On the one hand we stress the role of economic incentives and the overall fiscal implications of changing the systems as well as these incentives. On the other hand we emphasize the intergenerational considerations and the political implications of the ageing process of the Italian population. From our description it emerges that the overall design of the Italian reform is probably a good one, and yet some more steps need to be taken to speed up some of the positive effects of the reform process that, due the adverse demographic trends affecting PAYGO systems as well as the political arena, could easily evaporate.
The German Public Pension System: How it Was, How it Will Be, Axel H. Börsch-Supan and Christina B. Wilke
Abstract: Germany still has a very generous public pay-as-you-go pension system. It is characterized by early effective retirement ages and very high effective replacement rates. Most workers receive virtually all of their retirement income from this public retirement insurance. Costs are almost 12% of GDP, more than 2.5 times as much as the U.S. Social Security System. The pressures exerted by population aging on this monolithic system, amplified by negative incentive effects, have induced a reform process that began in 1992 and is still ongoing. This paper has two parts. Part A describes the German pension system as it has shaped the labor market from 1972 until today. Part B describes the reform process, which will convert the exemplary and monolithic Bismarckian public insurance system to a complex multipillar system. We provide a survey of the main features of the future German retirement system introduced by the so called “Riester Reform” in 2001 and an assessment in how far this last reform step will solve the pressing problems of the German system of old age provision.
Abstract: This paper extends earlier health and work studies by examining how mental health affects transitions out of paid work in the years prior to the traditional Social Security retirement ages. Given recent changes in the labor market, optimal mental health may be as important a prerequisite for continuing employment as good physical health. This study uses data from the Health and Retirement Study to examine how mental health is linked to transitions to early retirement or other unemployment in 1996 for middle-aged adults who were currently working in 1992 and whether physical health, job, or sociodemographic factors affect those links. The study results indicated that mental health plays a strong and significant role in the move from paid work to other unemployment in three ways, net of other documented health, job, and sociodemographic correlates of work status. First, higher baseline CES-D depressive symptoms predicted the transition to retiree in male workers. Second, increased CES-D depressive symptoms between 1992 and 1994 (net of baseline symptoms) predicted exits from paid employment and into other unemployment by 1996. Finally, low job autonomy did not have the hypothesized moderating effect on the mental health-work status link. The results also indicated that mental health may be an even more important predictor of transitions out of paid work among middle-aged workers than are physical health and functioning and that patterns of labor force exit differ for men and women.
Microsimulations in the Presence of Heterogeneity, Constantijn W.A. Panis
Abstract: This paper develops a method that improves researchers’ ability to account for behavioral responses to policy change in microsimulation models. Current microsimulation models are relatively simple, in part because of the technical difficulty of accounting for unobserved heterogeneity. This is all the more problematic because data constraints typically force researchers to limit their forecasting models to relatively few, mostly time-invariant explanatory covariates, so that much of the variation across individuals is unobserved. Furthermore, failure to account for unobservables often leads to biased estimates of structural parameters, which are critically important for measuring behavioral responses. This paper develops a theoretical approach to incorporate (univariate and multivariate) unobserved heterogeneity into microsimulation models; illustrates computer algorithms to efficiently implement heterogeneity in continuous and limited dependent models; and evaluates the importance of unobserved heterogeneity by conducting Monte Carlo simulations.
Background: Since its inception, the Health and Retirement Study (HRS) project has been interested in estimating pension wealth and predicting future income from pensions, and has periodically collected data to obtain pension Summary Plan Descriptions (or SPDs) from the employers of the study respondents. These documents are then coded, and the plan description data, along with some specific data from the respondents, are fed into a stand-alone program, the Pension Calculator, that estimates the pension income based on the program’s algorithms and user-defined parameters. The program is design to estimate the pension entitlements held by respondents of the HRS and the Panel Study of Income Dynamics (PSID), based on the plan formulas and benefit provisions obtained from the linked sample of pension providers. The Pension Calculator uses systems of equations to represent each of the pension plans, including all benefit formulas and payment provisions. These equations use, as input, work and income histories of the respondents. When combined with assumptions needed for the estimations of the present value of future benefit flow, the program calculates the appropriate pension entitlements, and prepares output data files for subsequent analysis.
History (Old Pension Calculator Program): The Pension Calculator program, first written in the 1980s, has been continuously improved over the years. The decision to rewrite the program was made for several reasons. The Pension Calculator program had been written in Borland PASCAL. Borland PASCAL has become obsolete. It is no longer supported by Borland and there are very few developers who still use it. This lack of support makes any large-scale changes to the old system risky. Should a developer encounter a compiler bug or an intentional design limitation, there is no support network available. The procedural style of programming combined with poorly named variables made the old Pension Calculator program difficult to decipher. In addition, the old Pension Estimation program lacked the ability to output intermediate results. Although the old Pension Estimation program provided present values of pension benefits, it did not output actual cash flow. Because of the old program's design, this change would have been very difficult to implement in the old system. Pension Calculator Program Rewrite The goals of this project are to enhance overall design, programming logic, and data representation, and to provide a graphical user interface. In addition, the new application will provide further capabilities and wide-ranging control over the input data and estimation parameters. The new application is implemented with advanced design and the object-oriented capabilities of the newly released Microsoft Visual Studio.NET.
New Pension Calculator (Alpha Release): General Information. An Alpha version of the new Pension Calculator is now available for testing purposes. This version is distributed to gather comments only, and it is intended for researchers who have previously used the PASCAL version of the program. This version is not feature-complete, but is designed to convey the look and feel of the new application. Researchers unfamiliar with its PASCAL predecessor may find it difficult to use. This initial release is distributed with 1992 HRS data only. Because the alpha version is not feature-complete, comparisons with the old PASCAL system are inappropriate at this time. In preparation for a final release, comparisons between the systems will be run and discrepancies resolved. Supported platforms include Windows NT, Windows 2000, and Windows XP.
New architecture: The Visual Studio application is designed as an object-oriented system. The main idea of this system is that the programming solution resembles the problem. Hence, a defined benefit plan has a (some) normal retirement formula(s). The programming architecture resembles that. There is a DefinedBenefitPlan object that encapsulates (has a) NormalFormulas object. Defined benefit plan has definitions for service years, final pay, social security, and many other objects. Each of these objects is written as a programming class, which means its behavior is independent of the other objects. The object-oriented approach will also result in a program that is easier to maintain and modify according to the individual research interests.
New features: User Interface: a visually intuitive graphical user interface to the parameters file allows users to change parameters, set new program defaults or restore original default settings.
Cash Flows option: this major enhancement allows users to output a vector of actual cash flows paid to a respondent. Additionally, combination plans are missing from the alpha version. Only 1992 Health and Retirement Study data is used in Alpha release. Several run options are not implemented in this alpha release, including multiple runs, different payout options for cash flows, the second Social Security wage base option, maximum retirement age and IRS limits for defined contribution plans. Although intermediate results are available, more detail will be provided in subsequent versions, particularly for defined benefit plans. Subsequent versions will also provide support for simulated respondents. The alpha release only supports the system default small fonts. Large fonts will be supported in future releases.
How to obtain a copy of New Pension Calculator: If you are a currently registered user of the previous version of the Pension Estimation Program, submit an electronic mail request to Cathy Liebowitz (email@example.com) stating that you wish to receive a copy of the new version. A CD-ROM containing an encrypted version of the software installation package will be sent via courier. The decryption pass-phrase will be sent separately. If you are not a currently registered user of the Pension Estimation Program, you may obtain access to the new version by submitting an application for restricted data (see http://hrsonline.isr.umich.edu/rda for details) or by applying for access to the new version through the MiCDA Data Enclave (http://micda.psc.isr.umich.edu/enclave).
What’s not available in Alpha version
MRRC researcher Olivia Mitchell co-organized a symposium on “Developments in Decision-Making Under Uncertainty: Implications for Pensions.” Her co-organizer was Stephen P. Utkus. The conference was supported in part by the Social Security Administration through the MRRC and was held at the Wharton School’s Pension Research Council on April 28-29, 2003. The conference examined the implications for pension plan design of recent behavioral research on decision-making under uncertainty. This research has important implications for retirement income adequacy, as well as for the ways in which pension sponsors as fiduciaries should respond. It also can inform policymakers seeking to improve the pension regulatory environment and working to increase financial literacy as a national goal.
The symposium was co-sponsored by the Wharton School’s Pension Research Council and Financial Institutions Center. Please visit their website for a full conference agenda (http://rider.wharton.upenn.edu/~prc/03conf.html).
5th Annual Retirement Research Consortium Conference held in Washington, DC
The fifth annual conference of the Social Security Retirement Research Consortium was held at the National Press Club in Washington, D.C. on May 15th and 16th, 2003. The two-day conference, entitled “Securing Retirement Income for Tomorrow’s Retirees,” was planned and arranged this year by the Center for Retirement Research at Boston College (CRR) in conjunction with the Michigan Retirement Research Center (MRRC) and the Social Security Administration (SSA). To open the conference, greetings were offered by Alicia Munnell, Director of the CRR, and John Laitner, Director of the MRRC. Introductory remarks were made James B. Lockhart III, Deputy Commissioner of the Social Security. The keynote address was delivered by Stephen Friedman who is Assistant to the President for Economic Policy and Director of the National Economic Council.
A panel discussion entitled: Social Security Personal Retirement Accounts:
the Nuts and Bolts was lead by Deputy Commissioner Lockhart. The panel
included William Shipman of Carriage-Oaks Partners, and MRRC researchers
Jeffrey Brown from the University of Illinois, Urbana-Champaign, Richard
Burkhauser from Cornell University, and Kent Smetters of the University
of Pennsylvania. Over the two days, researchers from each of the two consortium
centers presented papers on current projects, all of which are funded
by SSA. The topics covered in the five paper sessions included Participation
and Impacts of Social Security Reform, Savings Decision, the Role of Earning
in Retirement Income, Current Issue in Private Pension Plans, and the
Impact of Unexplained Events on Retirement Income. The final session of
the second day featured eleven Sandell Awardees who presented their research
findings. Papers presented at the conference are available in the MRRC
Overview: The HRS is a longitudinal survey of households with at least one member aged 51-61 at the time of the initial wave in 1992. It collects information on a wide range of issues that are potentially relevant to health and retirement. The survey is fielded every two years. At present, the 1992, 1994, 1996, 1998, and (preliminary) 2000 survey data have been released. The 2002 data are currently being collected. In 1993 and 1995, the AHEAD cohort (born in or before 1923) was added, and merged with the original HRS cohort in 1998. In 1998, two additional cohorts were added, the Children of the Depression Age (born in 1924-30) and the War Babies (born in 1942-47). The resulting sample is representative of the US population age 51 and older in 1998. All four cohorts were re-interviewed in 2000 and 2002. The Institute for Social Research (ISR) at the University of Michigan collects and distributes the HRS. Its main sponsor is the National Institute on Aging (NIA).
These data are the single richest and most valuable source of information for analyzing the work, health, and saving behavior of Americans reaching retirement age, and the survey offer uniquely valuable information for analyzing the impact of Social Security benefit rules and tax policy on behavior and economic well-being. Because of their high policy relevance, the Social Security Administration contracted with RAND in 2000 to create a set of data files that are based on the 1992-2000 Health and Retirement Study (HRS) to encourage the more widespread use of these data. The resulting data files are substantially more user-friendly than the original HRS files. These data are currently known as the RAND HRS data and available for download from the HRS website (http://www.umich.edu/~hrswww/docs/rand).
Through an MRRC grant in 2001, SSA funded RAND to publicly disseminate the non-restricted portion of the user-friendly HRS version. In its current form, this MRRC-sponsored project aims to make the RAND HRS data file publicly available to the research community, to update and maintain the file and its documentation, to promote the data, and to provide technical support to (prospective) users. A number of distinct activities meet these objectives and are described in this article. The project is directed by MRRC researcher and RAND Economist Constantijn (Stan) Panis and implemented in collaboration with Michael Hurd (MRRC and RAND), Patricia StClair and Delia Bugliari, both at RAND.
Data Enhancements: The RAND HRS contains variables that are derived from publicly available raw HRS variables only. Missing values of many variables have been imputed; variables are named consistently across survey waves; spousal information is added to the records of married respondents; and the files contain summary information on measures of health, wealth, and family structure. Specifically, the file contains variables in the following areas:
The use of this file requires substantially less time and effort than of the underlying HRS data releases. In addition, the files eliminate the need for common data preparation tasks, such as merging in spousal information and imputing missing values. They are increasingly being used at SSA, RAND, and the research community at large.
Current Project Activities: Expand the Data File to Include All HRS Cohorts, A major new release of the RAND HRS data, Version C, occurred in June 2003. The major change from Version B to C is the addition of the AHEAD entry cohort, completing the HRS sample. The file now incorporates the HRS, AHEAD, Children of Depression (CODA) and War Babies (WB) entry cohorts. This version adds data from AHEAD 1993 and 1995, and AHEAD respondents from HRS 1998 and 2000. It now includes data from 1992, 1993, 1994, 1995, 1996, 1998, and 2000. Where possible, measures consistent with those available from the HRS data are derived.
Prepare for Incorporation of the HRS 2002 Preliminary Release: The HRS 2002 is currently in the field. The project aims to incorporate the preliminary 2002 data as soon as possible in the RAND HRS data.
Expand the Subset of Variables on the File in Response to Requests of Users: As mentioned, the RAND HRS data contain information on demographics, health, wealth, income, health insurance, family structure, retirement, and employment. Most users will want to merge in additional variables from the standard HRS release files. For variables that are merged in by many users, it makes sense to process them centrally and make them part of the RAND HRS. Judging from technical support questions, this applies particularly to information on pensions and family structure. (The file from which the RAND HRS data are derived contained pension variables from the Employer Pension File. That file is restricted and the remaining, self-reported, pension information on the RAND HRS file is limited.) The household member count is improved, now counting spouses of household members for HRS 1996, 1998, and 2000. Some minor changes have been made to variables capturing retiree health insurance coverage of employer-provided health insurance and to Social Security and SSI disability variables.
Finally, Version C contains a single Stata Version 7 Special Edition file with all data waves. Prior RAND HRS releases did not contain a single Stata file, because the number of variables would exceed Stata’s maximum. The regular Version 7 is still bound by this maximum; Stata Version 7 Special Edition is not.
Update and Improve User Documentation: A critical part of the success of this project is well-composed, user-friendly documentation. As changes are made to the data, concomitant changes are made to the documentation. The current guide is an 898-page User’s Manual that undergoes continuous improvements.
Actively Promote the Data: A number of different promotion strategies
are utilized including: promoting the data via newsletters, electronic
list servers, and updated webpages at both the HRS site and the site of
the RAND Center for the Study of Aging (http://www.rand.org/labor/aging).
The MRRC also provides a link to the data from its link to Data Access.
The data were also promoted at the 2002 and 2003 annual meetings of the
Population Association of America with a user-initiated session and dissemination
of flyers and documentation throughout the conference.
Provide Responsive Technical Support to (Prospective) Users of the Data Files: At present, the RAND HRS support team receives about 2-4 new requests for support daily pertaining to RAND HRS data, with each request typically resulting in multiple exchanges. Almost all contact is by e-mail. The most commonly queried areas are HRS sampling weights; the timing and content of the next release of the RAND HRS data; how to merge RAND HRS data with other files; and requests for pension and family structure variables.
Requests for support are growing in proportion to increasing requests for the data. Between March 2002 and July 2003, the entire RAND HRS distribution set was downloaded 205 times. In addition, the data in Stata-format-only were downloaded over 200 times, in SAS-format-only over 100 times. The documentation was downloaded separately about 600 times. The download statistics do not permit an exact determination of the number of distinct individuals that downloaded the data, but our estimate is that there now are close to 300 users of the RAND HRS data.
Conclusion: The HRS provides a valuable source of data for researchers
and policy analysts. The complexity of the raw data files, however, is
daunting to many potential users. The RAND HRS project aims to create
more user-friendly data files and documentation with the goal of increasing
usage of this important data source. More widespread use of the data files,
coupled with responsive technical support and continued communication
with users, will lead to more and presumably better analyses of Social
Security issues by the external research community.
University of Michigan
Institute for Social Research
426 Thompson Street
Ann Arbor, MI 48106-1248