| July 2006
Volume 7, Issue 3
Although national attention may have temporarily shifted away from last year’s intense focus on Social Security reform, the underlying need for change remains undiminished. The summary of the 2006 Annual Report of the Social Security Trustees, for example, writes:
The fundamentals of the financial status of Social Security and Medicare remain problematic under the intermediate economic and demographic assumptions. Social Security’s current annual surpluses of tax income over expenditures will soon begin to decline, and will be followed by deficits that begin to grow rapidly toward the end of the next decade as the baby—boom generation retires. (www.ssa.gov/OACT/TRSUM/trsummary.html)
And, as always, the Michigan Retirement Research Center maintains its
high enthusiasm for contributing to the platform of scientific knowledge
on which public policy can be built. This year’s Retirement Research Consortium
Annual Conference will feature current research by consortium scholars
which demonstrates the RRC’s strong commitment to policy-oriented research.
The NBER Retirement Research Center takes the lead this year in organizing
the Conference, and the agenda (featured in this issue of the newsletter)
promises a stimulating conference. The theme this year is Pathways to
a Secure Retirement. With its reputation for interesting presentations
and thought-provoking discussion, the Conference is always well attended
by members of the policy and academic communities, the media and the public.
We’ll look forward to seeing you there.
John Laitner, Director
In this issue Michael Hurd and Susann Rohwedder discuss their MRRC-supported research.
Your research focuses on consumption. Would you explain how this differs from studying income or wealth?
The advantage of looking at consumption, or spending on goods and services, is that it is a direct measure of material well-being of individuals or of households in a given period. For example, households can consume less than their income, and in that case, income would be overstating the well-being of that household. On the other hand, households can consume more than their income if they have savings. In that case if we looked at income, we would get a different impression of the well-being of the household. With data on actual consumption, we are also able to explicitly model the choice to consume or to save. In models that only have income or wealth, this choice is only implied and not explicitly observed.
Finally, prior studies that have used income as a measure of well-being, have used pre-tax, or gross income. Gross income does not really reflect what is available to the household for consumption. So when we model economic behavior, consumption is the variable of choice. Until recently, however, good data on consumption have not been available.
Where do you get the data you need for these studies?
In the late 1990s, I (Hurd) applied for and received funding from the National Institute on Aging to send out a mail survey to a random subset of respondents in the Health and Retirement Study (HRS) to gather detailed information on consumption. The project is known as the Consumptions and Activities Mail Survey (CAMS). That first wave of data collection was in 2001. The Social Security Administration (SSA) provided funding for us to go back and survey the same respondents two years later in 2003. SSA again provided funding to collect a third wave of data in 2005 with an additional 800 new respondents to account for new enrollments into the HRS cohort in 2004. Susann joined in 2002, and we worked as team in the design of the 2003 and 2005 waves and have since collaborated in research using these data.
With continued funding, we are planning future waves of CAMS data collection. An extremely important part of the data collection has been the continued support of the SSA. It is also important to note that these data are unique worldwide. There is no other data set with several waves of data on consumption embedded within a larger panel study. As such, data from the CAMS can be linked with the rich data available in the HRS on income and wealth as well as many other variables.
Would you describe the retirement-consumption puzzle and the findings from your MRRC projects exploring this topic?
Prior studies found that households’ expenditures drop at or around retirement. In trying to explain the drop, there was the suggestion that at least some of it could be due to work-related expenses. But the size of the drop was larger than what could be explained by these expenses. Some researchers suggested that the drop in consumption at retirement was evidence that some people are surprised by their lack of economic resources and are forced to reduce consumption when income from earnings is no longer available. Thus, they enter retirement in economic distress. The dominant model in studies of economic behavior, the life-cycle model, posits that households are forward-looking and seek to smooth consumption or, more precisely, the marginal utility of consumption, over time. In this framework, individuals save during their work lives in order to have savings to spend during retirement. Sudden drops in spending are not easily explained by the life-cycle model, and the drop in consumption at retirement has come to be known as the retirement consumption puzzle.
The first matter we sought to address was to what extent households were actually surprised by the drop in spending at retirement. In CAMS we ask workers whether they anticipate reducing spending when they retire. We found that a large portion of the population anticipates the drop in spending at retirement. This indicates that the drop in spending is not a surprise to households. We also examined the magnitude of the anticipated drop and compared it to realizations. Controlling for age, anticipations and realizations are very close, on average. We also examined the distribution of recollected changes in spending at retirement. We found that about 37% of the population reported no change in spending at retirement. For about 11 percent, spending increased, for 20 percent, it drops by 20% or less, and for about 30 percent, spending drops more than 20 percent. Prior studies have generally reported average drops in spending, and this last group would contribute significantly to the average. But large drops in spending at retirement are by no means a population-wide phenomenon.
For the group with a modest decrease in spending at retirement, it may be that the substantial increase in available time could be used for smart shopping or home production, i.e. doing things yourself like gardening, home repairs, and cooking, some of which may have been hired before. There is also a different composition of spending—for some there is an increase in expenses for leisure activities. We also looked a little more carefully at the group with substantial declines and find that they are more likely to have experienced health shocks—that an important reason for retirement was poor health. So it may be less that they were not prepared for retirement and more that they were forced to retire early and with lower resources due to lost years of earnings.
The policy implications are important: if you were to find that there’s a population-wide problem, that would lead to different policy conclusions than if you were to find that there is not a population-wide problem but that there are subgroups where there’s a problem. You would certainly devise very different policies for those two situations.
How does your current project to study alternative replacement rates relate to your research agenda in this area?
This is part of larger research agenda that looks at retirement behavior and financial preparation for retirement. Replacement rates are an important ingredient in determining how much people need to save for retirement. We are very interested in developing a comprehensive summary measure—that is easily communicated to policymakers—of how well people are prepared for retirement. Existing measures like income replacement rates are only partial measures and are not very informative for some households. In general, income replacement rates are more meaningful for those whose retirement resources come in the form of an income stream such as Social Security or defined benefit pension plans. But with defined benefit plans on the decline, more and more people fund their retirement out of a mix of resources including Social Security, savings, IRA’s and defined contribution plans. We are trying to come up with an alternative measure which brings in all of the sources of retirement funding that people have, not just income, and also takes into account taxation.
Also, when people talk about replacement rates, they are really thinking in terms of a single person. For example, imagine a 65 year-old male reaching retirement and he’s earning $40,000 a year and now his income after retirement is going to be $25,000, and you can make some judgment about whether that is adequate. But most people reach retirement as married people. The question is, what are spending needs of the household in retirement given that the probability that both will be alive in 11 years is below 50%. So that has to be taken into account.
Another important consideration is the time path of consumption over the full period of retirement. Consumption patterns in advanced old age will not be the same as immediately after retirement. We are able to look at this now with the CAMS panel data because we have change in spending during retirement at the household level. So we see among couples how their spending evolves as they age, we see among singles how their spending evolves and we can compare that with spending change that is predicted by our model of economic behavior.
What is the importance of this topic?
Using consumption as a measure of well-being provides a more accurate picture of the well-being of the household. This point is underscored in other work we have done for MRRC which found that a consumption-based measure of poverty would actually decrease the percentage of people we categorize into poverty, and this is especially true in advanced old age as people spend increasingly out of savings. If policy is concerned about the economic well-being of various subpopulations in the overall population, then we need to be able to measure accurately their economic status before we start to devise policies to alter that relative economic status. In our view, the profession has somewhat mismeasured economic status, particularly among the very old who primarily are widows because we have not adequately taken into account the ability to consume out of wealth.
Where do you see this line of research going?
The long-term objective of this line of research is to understand consumption and saving behavior of people as they approach retirement and pass through retirement and into advanced old-age. A related goal is to develop a comprehensive measure of preparation for retirement that is well-founded in theory and in data. The work we have done so far has included important steps towards this overall objective. Our funding from NIA and SSA through MRRC have worked synergistically over the years to develop the necessary data and to foster the development of theory in this area.
We still have a number of important steps. We still do not have a comprehensive measure of returns-to-scale for households at older ages. We are also working on a comprehensive measure of pension entitlements and other retirement resources. In the long run, we feel we will be able to not only inform economic science but also policymakers about economic preparation for retirement.
If we are to address a possible shortfall in Social Security, we need to understand the consequences that might come from restructuring Social Security. In order to understand that, we need to have reliable measures of economic resources among the retired population and a standard by which to judge whether these resources are adequate. It is very important to have a good working model of economic behavior of households that is well-structured and well-founded in the data that will allow us to model and simulate the effects of potential reforms.
Consumption and Economic Well-Being at Older Ages: Income- and Consumption-Based Poverty Measures in the HRS by Michael Hurd and Susann Rohwedder WP 2006-110, January 2006.
Changes in Consumption and Activities in Retirement by Michael Hurd and Susann Rohwedder WP 2005-096, March 2005.
The Retirement-Consumption Puzzle: Anticipated and Actual Declines in Spending at Retirement by Michael Hurd and Susann Rohwedder WP 2004-069, January 2004.
Michael D. Hurd is Senior Economist at RAND and Director of the RAND Center for the Study of Aging and a Research Associate at the National Bureau of Economic Research. His current work includes the effects of pensions on retirement, the use of subjective information, particularly survival probabilities, to explain economic decisions such as saving and retirement, and methods of assessing uncertainty in a population. He served on the Technical Panel of Experts (1990) and the Panel of Experts (1991) to the Social Security Advisory Council, on the Advisory Committee for the World Bank Old-Age Security Study, on the Panel on Retirement Income Modeling, Committee on National Statistics, National Research Council, 1995-1996, and on the Workshop on Priorities for Data on the Aging Population, Committee on National Statistics, National Research Council, 1996. He is a Co-PI of the Health and Retirement Study.
Susann Rohwedder is an Economist at RAND. Her research focuses on the economics of aging in the areas of household consumption and saving behavior, retirement and expectation formation. She has written on the impact of pension reforms on household saving in the UK; on the retirement-consumption puzzle; on spending and saving patterns among the older population; and on individuals’ expectations about future Social Security benefits and longevity. Other papers deal with data quality and survey methods in the Health and Retirement Study. She is involved in several data collection efforts: she is jointly responsible for the design of the Consumptions and Activities Mails Survey (CAMS) and for several modules of the American Life Panel (ALP).
A new version of the RAND HRS Data File has been released, replacing all prior versions. The RAND HRS Data file is a cleaned and easy-to-use version of the Health and Retirement Study (HRS) with derived variables covering a broad range of measures and named consistently across waves. The file was developed by the RAND Center for the Study of Aging with funding from the National Institute on Aging (NIA) and Social Security Administration (SSA).
As of 2006, nine HRS waves are available for study. The RAND HRS Data file (Version F) is based on 1992, 1993, 1994, 1995, 1996, 1998, 2000, and 2002 final releases, and the 2004 early release of HRS data. The complete HRS includes five entry cohorts, as does the RAND file. The file incorporates only the core interviews. It does not include exit interview or any restricted data.
Besides incorporating the 2004 early release, this version corrects known problems and adds a few new variables, including 2nd home value from 1998 forward. The Office of Research, Evaluation, and Statistics at SSA provided important research direction in the design of this data file.
See the RAND Contributions page for further information. Questions or problems concerning the RAND HRS Data should be addressed to RANDHRShelp@rand.org at RAND.
The Health and Retirement Study (HRS) has released a new user guide for beginning HRS data users: Getting Started with the Health and Retirement Study. This guide details the HRS history, Web site, useful documentation, and includes three chapters on data analysis examples using the HRS RAND Data. Access the guide at http://hrsonline.isr.umich.edu/docs/dmgt/IntroUserGuide.pdf
New user-friendly tutorials make it easy to use and understand Panel Study of Income Dynamics (PSID) data. The new tutorials are targeted to new users, including undergraduates and advanced secondary school students. The new tutorials provide a step-by-step guide to answer specific questions. The tutorials demonstrate various features of the PSID, including cross-sectional, panel, and intergenerational analyses. They illustrate several research domains such as housework, employment, welfare, poverty, health status, food stamp use, housing, etc… All tabulations can be conducted within Excel.
Access the tutorials at http://www.psidonline.isr.umich.edu/Guide/tutorials/tutorials_home.html
Access the PSID Online Data Center at http://www.psidonline.isr.umich.edu/
The Data Center:
Supports custom designed subsetting of the data;
Provides customized codebooks;
Allows variable searching and browsing;
Supports complex PSID-CDS dataset merges;
Provides the Data in SAS, STATA, dBase, ASCII, Excel;
Allows the entire PSID dataset to be downloaded directly.
The Impact of Misperceptions About Social Security on Saving and Well-being by Susann Rohwedder and Arthur van Soest WP 2006-118
=>People who overestimate their Social Security benefits before retirement are worse off, across a number of measures of well-being, in retirement.
=>This connection is stronger for those who claimed Social Security retirement benefits earlier than they expected, perhaps because of unexpected poor health.
Public Health Insurance and SSI Program Participation Among the Aged by Todd Elder and Elizabeth Powers WP 2006-117
=>SSI participation declines with the expansion of alternative routes to Medicaid (i.e., noncategorical Medicaid).
=>Availability of non-categorical Medicaid increases Medicaid participation among SSI nonparticipating eligibles.
Social Security Replacement Rates for Alternative Benchmark Earnings by Olivia S. Mitchell and John W.R. Phillips WP 2006-116
=>Workers with the median HRS profile receive benefits worth 55% of lifetime earnings versus 48% for the medium scaled profile used by SSA.
=>Replacement rate measures linking benefits to national average wages at retirement generate more modest replacement rates (about 33% for the median HRS worker).
The Inattentive Participant: Portfolio Trading Behavior in 401(k) Plans by Olivia S. Mitchell, Gary R. Mottola, Stephen P. Utkus, and Takeshi Yamaguchi WP 2006-115
=>80% of participants do not trade at all.
=>11% make a single trade over two years.
=>Very few participants trade actively and tend to be affluent older men with higher incomes and longer job tenure.
Pathways to a Secure Retirement
National Press Club – 529 14th St. NW, 13th Floor - Washington, DC 20045 August 10-11, 2006
Sponsored by the Office of Policy, Social Security Administration and the Retirement Research Consortium
Contact: Janet Stein (email@example.com)
The Health and Retirement Study (HRS) is a nationally representative longitudinal study initiated in 1992 and continuing to the present, under the prime sponsorship of the National Institute on Aging (NIA) with additional support from the SSA, the US Department of Labor, and other entities. The study is based at the University of Michigan, Institute for Social Research (ISR/UM).
A large percentage of HRS respondents provide written consent agreeing to SSA's disclosure of confidential administrative records to ISR/UM. The administrative records are SSA wage and self-employment income data (earnings) and data on OASDI and SSI benefits. Linkage of SSA administrative records to HRS data make it possible for qualifying researchers to undertake a wide range of scientific analyses and policy studies of crucial interest to retirement researchers. Linked data on earnings and benefits are now available for all waves of the HRS through 2004.
Importance of the Data
Policy analysts both in the government and in the academic community are interested in the influence of pensions, wealth, saving, health, and intra-family transfers on economic status of older Americans. Research on these topics is disadvantaged in the absence of merged data files containing both SSA and survey data for the same respondents. For example, experts seeking to understand how public and private retiree pensions interact, and whether these interactions are changing over time, require information on both types of pension systems for a wide range of cohorts. Administrative records are also keys to testing how well informed respondents are regarding benefit amounts they are entitled to, and to accurately determining how much they receive.
It is also important to link SSA information with survey responses in order to track patterns of well-being across life transitions. Administrative records contain essential information regarding the time path of application for and receipt of Social Security benefits by type (i.e., disability, retirement, and survivor benefits) and Supplement Security Income (SSI) payments. Merged SSA earnings histories are also key in understanding the work-life patterns that provide economic security to retired Americans. Many MRRC researchers utilize this important resource, and their findings are reported in MRRC working papers.
Keeping it Confidential
All users of the Restricted Data files must sign a contractual agreement with ISR pledging to abide by a number of conditions of use, which ensure the confidentiality of the data. Both SSA and ISR provide secure arrangements under which data files, extracts, and listings can be stored and retrieved only by authorized persons. In addition, top coding of the earnings data help ensure that individuals are not uniquely identified. Users also agree to only present aggregated results in analysis using the data.
In past years, MRRC has contributed funds toward the support of a secure data environment at the Institute for Social Research, (the Data Enclave), where researchers can analyze Social Security administrative files without risking a breach of confidentiality. While this is an important resource, travel to the site remains a significant obstacle for many potential users, since users must be physically present to use the enclave facilities. To deal with this problem, work is underway to develop an alternative method for access to the enclave: a remote workstation connected to an enclave server via a virtual private network (VPN) connection that emulates (in hardware and software) a dedicated data communications line.
University of Michigan
Institute for Social Research
426 Thompson Street
Ann Arbor, MI 48106-1248