United States General Accounting Office GAO Report to the Ranking Minority Member, Subcommittee on Employer- Employee Relations, Committee on Education and the Workforce, House of Representatives March 2003 RETIREMENT INCOME DATA Improvements Could Better Support Analysis of Future Retirees’ Prospects GAO-03-337 March 2003 RETIREMENT INCOME DATA Improvements Could Better Support Highlights of GAO-03-337, a report to Analysis of Future Retirees' Prospects Ranking Minority Member, Subcommittee on Employer-Employee Relations, Committee on Education and the Workforce, House of Representatives Future demographic trends include Experts consulted by GAO cited priorities for improving retirement data that a doubling of the nation’s retiree fit into two broad categories: (1) obtaining better data from employers on population and only modest labor employee benefits and (2) obtaining better data by linking more individual force growth, leading to concerns and household surveys with administrative data (such as employer records, about retirement income adequacy and Social Security earnings history records). Information from employers, for future generations. Credible projections of the effects of policy such as documents describing the features of their pension plans, would proposals on federal spending and enable analysts to forecast future retirement income of pension holders, future retirees’ income are based on the specific features of their pension plans and the likely necessary. Because adequate data distribution of pension income and wealth for different segments of the is critical to the analysis of population. Linking individual and household surveys with administrative retirement income and wealth, data creates new information, such as the demographic characteristics of GAO was asked to identify data employees whose pensions are affected by the formulas that employers use improvements that experts say are to calculate contributions or pension payments. a priority for the study of retirement income and wealth, as Analysts attribute the shortcomings in retirement income data primarily to well as factors limiting efforts to fragmentation of the responsibility for data collection and analysis, the obtain the needed information. burden of data collection on respondents, and confidentiality considerations that restrict access to these data. Fragmentation of responsibility occurs, in their view, because no single agency has a statutory mandate to collect or to The Congress should consider analyze all the data needed to support a more comprehensive study of directing Labor to obtain from plan retirement income and wealth. With regard to respondent burden, some administrators electronic filings of information on pension plans is no longer collected, in part, out of concern SPDs and summaries of material that it was an unnecessary burden on the firms having to submit it. Finally, modifications and make them certain kinds of data needed to make projections are not widely available to publicly available. all analysts because of the confidentiality laws that authorize their In addition, GAO recommends that collection. the Secretary of Labor • direct the Bureau of Labor Statistics to prepare a plan to improve data for analyzing retirement income and wealth in coordination with other agencies and • obtain copies of summary plan descriptions in cases where analysts working on federally- conducted or sponsored research seek them for statistical purposes. GAO recommends that the Internal Revenue Service publish tabulations of information filed in IRS forms 5498, 1099R and W-2. www.gao.gov/cgi-bin/getrpt?GAO-03-337 To view the full report, including the scope and methodology, click on the link above. For more information, contact Barbara Bovbjerg (202) 512-7215 or Robert Parker (202) 512-9750. Contents Letter 1 Results in Brief 2 Background 4 Experts Cited Need for Better Data and Better Data Set Linkage 7 Many Factors Limit Needed Retirement Income and Wealth Data 15 Conclusions 21 Matter for Congressional Consideration 23 Recommendations 23 Agency Comments 23 Appendix I Scope and Methodology 27 Appendix II Status of Recommendations from the 1997 Report of the Panel on Retirement Income Modeling 29 Appendix III GAO Survey on Retirement Income Data Needs and List of Respondents 40 List of Respondents to the Survey 51 Appendix IV Views of GAO’s Expert Panel on Retirement Income Data Needs 65 Need for Better Matched Data 66 More Employer Information Needed on the Value and Provisions of Employer Provided Pensions 69 Appendix V Characteristics of Selected Surveys for Analysis of Retirement Income and Wealth 72 Appendix VI Comments from the Department of Commerce 76 Appendix VII Comments from the Department of Labor 77 Page i GAO-03-337 Retirement Income Data Appendix VIII GAO Contacts and Staff Acknowledgments 81 GAO Contacts 81 Staff Acknowledgments 81 Tables Table 1: Examples of Federal Agency Retirement Income-Related Data Collection 16 Table 2: Federal Outlays for Selected Longitudinal Studies—Fiscal Years 1997-2001 29 Table 3: Summary Table of Selected Survey Data Sources 72 Abbreviations BLS Bureau of Labor Statistics CPS Current Population Survey EBRI Employee Benefit Research Institute EBSA Employee Benefits Security Administration ERISA Employee Retirement Income Security Act of 1974 HRS Health and Retirement Study IRA Individual Retirement Account IRS Internal Revenue Service OMB Office of Management and Budget PBGC Pension Benefit Guaranty Corporation PSID Panel Study of Income Dynamics SCF Survey of Consumer Finances SIPP Survey of Income and Program Participation SPD summary plan description SOI Statistics of Income This is a work of the U.S. Government and is not subject to copyright protection in the United States. It may be reproduced and distributed in its entirety without further permission from GAO. It may contain copyrighted graphics, images or other materials. Permission from the copyright holder may be necessary should you wish to reproduce copyrighted materials separately from GAO’s product. Page ii GAO-03-337 Retirement Income Data United States General Accounting Office Washington, DC 20548 March 21, 2003 The Honorable Robert E. Andrews Ranking Minority Member Subcommittee on Employer-Employee Relations Committee on Education and the Workforce House of Representatives Dear Mr. Andrews: The nation’s retiree population will double within the next few decades, while at the same time the labor force will grow only modestly, potentially stressing the national economy. In light of these demographic trends, policymakers have been moved to consider how the future economy can support the large retiree population, and whether retirement income levels will be adequate in the future.1 To increase their understanding about issues related to retirement, policymakers need credible projections of the effects of their proposals on federal spending and on future retirees’ income. Analysts will be better able to develop accurate projections if they have relevant, reliable, and timely data on patterns of saving and actual retirement income and wealth. Because adequate data will be so important to analysis of retirement income and wealth for future retirees, including people in the “baby boom” generation and later generations, you asked us to assess the adequacy of data available for making such projections. In response to your request, as agreed, we identified (1) data improvements that experts say are a priority for the study of retirement income and wealth and (2) factors limiting efforts to obtain the needed information. To address these topics, we conducted a Web based survey of nearly 200 individuals with retirement income expertise, held a daylong meeting with a diverse group of 11 retirement income experts, and interviewed retirement income analysts and officials of the Departments of Labor (Labor), Commerce, and Treasury. We conducted our work between 1 For a discussion of standards for evaluating retirement income adequacy, see U.S. General Accounting Office, Social Security: Program’s Role in Helping Ensure Income Adequacy, GAO-02-62 (Washington, D.C.: Nov. 30, 2001). Page 1 GAO-03-337 Retirement Income Data February and December 2002 in accordance with generally accepted government auditing standards. (For more details about our scope and methodology, see app. I.) Results in Brief Acknowledging that there is a great deal of missing data related to retirement income, experts we consulted cited priorities for data improvements that fit into two broad categories: (1) obtaining better data from employers on employee benefits and (2) obtaining better data by linking more individual and household surveys with administrative data (such as employer records and Social Security earnings history records). The kinds of information from employers that analysts indicated are missing included the documents employers provide to employees describing the features of their pension plans, such as the plan’s normal retirement age and reductions for early retirement. This information would help analysts to forecast future retirement income of pension holders, based on the specific features of their pension plans and the likely distribution of pension income and wealth for different segments of the population. With regard to linking datasets, currently linkages between individual and household survey data and administrative data are limited. Linking data creates new information by matching survey data about individuals (using names, or taxpayer identification numbers) to a second set of records, such as administrative records on pension plans. There is inadequate information, for example, about which demographic groups have different types of pensions. Thus, while analysts may know the prevalence of certain formulas used to calculate employer contributions or pension payments, they generally lack reliable information about the demographic characteristics of the employees whose pensions are affected by these formulas. Analysts attribute the shortcomings in retirement income data primarily to fragmentation of the responsibility for data collection and analysis, the burden of data collection on respondents, and confidentiality considerations that restrict access to these data. Fragmentation of responsibility occurs, in their view, because no single agency has a statutory mandate to collect or to analyze all the data needed to support a more comprehensive study of retirement income and wealth. For example, while the Department of Labor’s Bureau of Labor Statistics (BLS) collects data on pensions, and the Census Bureau collects data on individuals’ and households’ income, neither agency is responsible for all of the data needed to project future retirement income and wealth. Other agencies such as the Department of Labor’s Employee Benefits Security Administration (EBSA)—formerly called the Pension Welfare and Benefits Page 2 GAO-03-337 Retirement Income Data Administration—and the Social Security Administration also collect data, but the extent to which these agencies share data is limited. With regard to respondent burden, some information on pension plans is no longer collected out of concern that it was an unnecessary burden on the firms having to submit it, as well as concern about the Department of Labor’s costs for storing the information. For example, the 1997 Taxpayer Relief Act (P.L. 105-34) ended the requirement that employers file with the Department of Labor copies of documents summarizing the features of the pension plans they offer. Finally, certain kinds of data needed to make projections are not widely available to all analysts because of the confidentiality laws that authorize their collection. The Census Bureau and others are exploring options for expanding access without compromising the confidentiality of the data. For example, the Census Bureau has established additional research data centers throughout the country where approved researchers with approved projects can work with confidential data and produce statistical summaries that meet strict disclosure requirements. We are offering a Matter for Congressional Consideration and making recommendations to the Department of Labor and the Department of the Treasury that seek cost-effective approaches to help fill some of the data needs while taking into account respondent concerns about increased reporting burdens and agency concerns about maintaining confidentiality. We provided a draft of this report to the Departments of Commerce, Labor, the Treasury and the Internal Revenue Service (IRS). We received technical comments from all four and incorporated their suggestions as appropriate. We also provided a draft of this report to the 11 members of our expert panel and modified the draft as appropriate in response to their comments. Commerce had no major comments on the report (see app. VI). Labor agreed on the need for access to accurate data but did not agree with our recommendations to the Secretary of Labor (see app. VII). Labor indicated that it did not have authority to require that summary plan descriptions (SPD) be filed electronically. Accordingly, we changed one of our draft recommendations to the Secretary of Labor into a Matter for Congressional Consideration. Labor also had concerns about the burdens our recommendations might pose. Although we acknowledge their concerns, nonetheless, we conclude that the need for improvements in retirement income data outweighs the likely costs involved. We therefore continue to address two recommendations to the Secretary of Labor and one to the IRS. Page 3 GAO-03-337 Retirement Income Data Background Forecasting future retirement income needs—and how well they will be met through current savings, pension plans, and Social Security benefits— is a challenge, in part because of the many variables involved. Although Social Security is the primary source of income for many retirees, private pensions and other sources of income serve as key supplements and help retirees receive adequate income in retirement. In order to measure a person’s current income and wealth, information is needed about many items, including the person’s wage and nonwage income, home equity, pension, and nonfinancial assets and liabilities. In addition, to project a person’s future income and wealth, researchers need information such as a person’s earnings history, whether he or she chooses to participate in the pension plan offered by an employer, and how the person might respond to changing incentives for saving and investing for retirement. Other factors include whether the person chooses to accumulate savings apart from retirement plans, how long the person remains in one job, whether the person decides to cash out his or her retirement plan when changing employment, level of indebtedness, and the availability of health insurance during retirement. In addition, individuals’ retirement funds depend on employer decisions, such as what kinds of pension plans and the availability and cost of retiree medical and long-term care insurance. To make estimates for people in different demographic groups, not just aggregate estimates for entire generations, analysts need to know how these factors vary based on individuals’ demographic characteristics. Estimates of future income adequacy also rely heavily on projections of macroeconomic factors, including estimates of future rates of inflation, and rates of return on stocks and bonds and changes in home values. Furthermore, retirement income data needs keep changing, in part, as a result of trends in the pension industry and the labor force. The pension industry provides a growing variety of pension products with different features and legal structures. Forecasting pension income becomes more complicated with, for example, firms’ converting of traditional pension plans into new pension hybrid products such as cash balance plans. These plans combine features of both defined benefit2 and defined contribution3 2 A defined benefit plan is a type of plan where the sponsor provides a guaranteed benefit generally expressed as monthly benefit based on a formula that generally combines salary and years of service to the company. 3 A defined contribution plan is a type of pension that establishes individual accounts for employees to which the employer, participants, or both make periodic contributions. The benefits are based on employer and participant contributions to and investment returns (gains and losses) on the individual accounts. Page 4 GAO-03-337 Retirement Income Data plans, which adds to their complexity.4 Changes in employment patterns, such as the decrease in the length of time employees spend in a single job, will also have an effect on pension income. The federal government collects a great deal of data pertinent to the analysis of retirement income and wealth. Surveys of individuals and households collected by or sponsored by the Census Bureau, the Federal Reserve Board, BLS, and the National Institute on Aging are important sources of information. Similarly, surveys of businesses by BLS provide information about the pension and health care benefits firms offer their employees. A great deal more information pertinent to the analysis of retirement income and wealth is contained in the administrative documents that businesses and individuals must provide to state and federal agencies that administer governmental programs and enforce regulations. For example, many private employers must file, on a Form 5500, annual reports concerning their employee benefit plans for the IRS, Labor’s EBSA, and the Pension Benefit Guaranty Corporation (PBGC).5 Many pieces of information concerning pensions and retirement savings are also provided on forms that employers and financial institutions file with the IRS. In addition to the federal government’s collection of retirement income related data, several private entities also conduct surveys that provide useful information concerning retirement income. For example, the Health and Retirement Study (HRS) conducted by the University of Michigan’s Institute for Social Research, is a longitudinal survey of adults over the age of 50. The information collected includes such topics as respondents’ physical and mental health, insurance coverage, financial status, family support systems, labor market status, and retirement planning. The HRS is supported primarily by the National Institute on Aging, with additional funding over the years from the Social Security Administration, Labor’s EBSA, the Department of Health and Human Services Assistant Secretary for Planning and Evaluation, and the state of Florida. Also, private benefit consulting firms and organizations, such as the Employee Benefit 4 U.S. General Accounting Office, Cash Balance Plans: Implications for Retirement Income, GAO/HEHS-00-207 (Washington, D.C.: Sept. 29, 2000) and Private Pensions: Implications of Conversions to Cash Balance Plans, GAO/HEHS-00-185 (Washington, D.C.: Sept. 29, 2000). 5 The IRS administers and enforces tax code provisions concerning private pension plans. EBSA enforces Employee Retirement Income Security Act (ERISA) pension requirements, and the PBGC insures the benefits of participants in defined benefit pension plans that are eligible for preferential tax treatment. Page 5 GAO-03-337 Retirement Income Data Research Institute (EBRI), conduct surveys of employers concerning pension plans. Retirement income analysts currently use available retirement income and wealth data to project future retirement income needs. For example, the Social Security Administration has developed a forecasting model, Modeling Income in the Near Term (MINT), to project demographic changes, retirement income, and Social Security benefits generally for persons born between 1931 and 1960. The Congressional Budget Office has also developed the Congressional Budget Office Long-Term model to study the range of possible outcomes for the balance of the Social Security trust fund.6 In a 1997 report, the National Research Council’s Panel on Retirement Income Modeling reviewed the available sources of data on retirement income and recommended several improvements.7 While the individual recommendations covered various topics, many involved the collection of additional information, as well as the establishment of an interagency task force for coordination purposes. While some of the panel’s recommendations have been adopted, others have not. For example, as the panel recommended, the federal government has continued to support longitudinal studies. However, Labor has not acted on the panel’s recommendation that it establish an interagency task force on employer data to plan collection of retirement income related data. In some respects, less data are available than was the case when the report was prepared. (See app. II for details.) 6 For additional information on the Congressional Budget Office’s model, see Congress of the United States, Congressional Budget Office, Uncertainty in Social Security’s Long-Term Finances: A Stochastic Analysis (Washington, D.C.: Dec. 2001). The Social Security Administration and others such as EBRI have supported the development of other simulation models, such as the Social Security Policy Simulation Model (SSASIM) to study the effect of changes in the Social Security program and pension law. More recently the Social Security Administration and others have supported the development of the GEMINI model, a policy microsimulation model developed by the Policy Simulation Group. EBRI has developed its own model, the EBRI Retirement Income Projection Model. 7 Constance F. Citro and Eric A. Hanushek, eds., Assessing Policies for Retirement Income: Needs for Data, Research and Models (Washington, D.C.: National Academy Press, 1997). The panel was sponsored by Labor’s EBSA, the National Institute on Aging, PBGC, the Social Security Administration, and the Teachers Insurance and Annuity Association-College Retirement Equities Fund (TIAA-CREF) Institute. Page 6 GAO-03-337 Retirement Income Data Much of the data needed to assess retirement income and wealth are subject to federal laws protecting the confidentiality of information reported to the federal government. Laws limit the access to this information. Without access, it is not possible to pull together data from different sources to provide more complete information about individuals or organizations. The success of data gathering efforts by federal agencies and others relies on widespread trust that personal data will be kept confidential, protected from disclosure, and used only for specified purposes.8 Experts Cited Need for Experts we consulted cited priorities for improving retirement data that fit into two broad categories: (1) obtaining better data from employers on Better Data and Better employee benefits and (2) obtaining better data from individual and Data Set Linkage household surveys by linking them with administrative data. The kinds of information concerning employers and employer-sponsored benefit plans that analysts sought included the features of their pension plans, such as minimum and maximum allowable contributions, or formulas for calculating benefits from defined benefit pension plans. Retirement income experts believed these data would allow them to more accurately measure or project retirement income and wealth now and in the future and estimate effects of potential retirement policy changes. Linkage of data from different sources creates new information by providing data about individuals matched to other data about the individuals from a second set of records, such as administrative records on pension plans. Analysts are able to use general data on some employer-sponsored pension plans and data on households and individuals. However, without linkages between these types of data it is difficult to obtain information about retirement offerings for specific households. As a result, analysis of pension offerings by demographic groups is limited. Experts Place Priority on Most of the experts responding to our survey on retirement income data Improvement in Data from needs assigned better retirement income-related data from employers to Employers the high or highest priority category. (See app. III for results of the survey.) In addition, the experts asserted that employers, rather than employees, could provide more accurate information about pension plans. 8 Among other things, the E-Government Act of 2002 (P.L. 103-347) permits the sharing of information concerning businesses among designated statistical agencies and provides for additional safeguards to protect the confidentiality of statistical data collected by all agencies. Page 7 GAO-03-337 Retirement Income Data Participants on our expert panel expressed interest in improving access to the mandatory pension plan disclosure reports, such as summary plan descriptions (SPD). (See app. IV for a summary of the expert panel’s discussion.) Employers must provide these documents to employees who participate in a plan (such as a pension plan) in order to provide them with an understandably written description of their rights and benefits under the plan. To improve projections of future retirement income and the effects of policy options, panelists also expressed interest in data provided to the IRS on forms such as 1099-R and 5498.9 Employer Documents Are an Analysts responding to our survey, reported as their highest or high Important Source of priority, the kind of information reported on some employer-submitted Information on Retirement forms. Because pension plans vary widely, panel experts said they especially needed details of employee pension plans, such as the type of Plans pension plan (defined benefit, defined contribution, or other), eligibility for participation and benefits, the plan’s early retirement age, sources of contributions to the plan, and the method by which the amount of the contributions and benefits are calculated. Panelists believed this information would help analysts project future private pension benefits and the effects of proposed policy changes. Panel members also recommended that such pension data be obtained directly from employers, citing the need for accuracy as an important factor. Inconsistencies have been found between employee and employer provided data. One study, for example, compared employees’ reports concerning the employer-sponsored pension plans they were participating in, or that were available to them, with the information about those pension plans obtained directly from the employers.10 It found significant discrepancies between the two sets of data, large enough for the authors to urge a great deal of caution in using the household survey data for reliable information about the actual characteristics of employer- sponsored pension plans. For example, among employees whose 9 The IRS form titled “Distributions from Pensions, Annuities, Retirement or Profit-Sharing Plans, IRAs, Insurance Contracts, etc.” is identified as IRS Form 1099-R. The IRS form titled “IRA and Coverdell ESA Contribution Information” is identified as IRS Form 5498. 10 Alan L. Gustman and Thomas L. Steinmeier, What People Don’t Know About Their Pensions and Social Security: An Analysis Using Linked Data from the Health and Retirement Study (Cambridge, Mass.: National Bureau of Economic Research, 1999). This paper discusses in detail some of the problems associated with having respondents provide details of their own pensions, as opposed to employers directly providing information. Page 8 GAO-03-337 Retirement Income Data employers reported that they were covered by a defined benefit plan, only 56 percent of the employees thought that they had such a plan. Likewise, there seemed to be high levels of error in other basic details about pension plans, such as the eligibility date for early retirement. Using employer-submitted forms as a source of information was suggested as one way to increase the accuracy of pension data. Retirement income experts agreed that the Form 5500 is an important source of pension information available in government administrative records.11 Sponsors or administrators of employee benefit plans subject to ERISA12 must file this form annually. The Form 5500 was intended, in part, to measure employers’ compliance with both the fiduciary and funding provisions laid out in ERISA legislation. The Form 5500 filed for pension plans contains useful aggregate information about plans. It provides information about the financial condition of the plan, annual amounts contributed by participants, and the plan’s income on investments. The form also provides information on plan characteristics, such as plan type (defined benefit or defined contribution), method of funding, and numbers of employees, participants, and employees who are excluded from the plan for various reasons. While the Form 5500 provides aggregate data about plans, it does not contain information useful for calculating individual’s contributions or benefits. For example, it does not provide information on formulas for calculating employer contributions to plans or for calculating retirees’ benefits.13 In addition, it does not provide any data for pension plans outside the reporting requirements of ERISA, such as those for governmental employers, certain nonqualified plans for highly 11 Form 5500 is a disclosure form that private employers with qualified pension plans are required to file with the IRS, Labor’s EBSA, and the PBGC. Schedule SSA to the Form 5500, which is not publicly disclosable, identifies individuals who leave employment with deferred vested benefits. The Social Security Administration uses this information. 12 The ERISA of 1974 is a federal law that set minimum standards for pension plans sponsored by private employers. These standards govern the management, operation, and funding of the plan. Labor’s EBSA enforces these ERISA provisions. 13 U.S. Department of Labor. Pension and Welfare Benefits Administration, Private Pension Plan Bulletin: Abstract of 1998 Form 5500 Annual Reports, No. 11, Winter 2001-02. Page 9 GAO-03-337 Retirement Income Data compensated people,14 simplified employee pension (SEP) 15 plans, or savings incentive match plans for employees (SIMPLE).16 Reporting and processing requirements also mean that data from Form 5500 reports may not be available to researchers for up to 3 years after the plan year. Finally, Labor officials find frequent errors in information submitted on the forms.17 Partly because of the limitations surrounding the Form 5500s, retirement income experts are increasingly interested in access to the SPDs, which are summaries of employers’ pension offerings. The requirements of ERISA call for SPDs to include specific details about employee pension plans, including the type of pension plan, eligibility requirements, normal retirement age, vesting18 and disqualification rules, sources of contributions to the plan, and method by which the amount of the contribution is calculated. ERISA required employers to file SPDs and documents, called “summaries of material modifications,” describing changes to the plans with Labor. These were made publicly available at Labor’s public disclosure room in Washington, D.C. The SPDs served many purposes: (1) they were a source of information to employees about the offerings included in their own pension plans, (2) they were a means of informing Labor about what types of plans a company was offering so Labor could perform monitoring and enforcement, (3) they also provided researchers with a high level of detail on pension offerings. EBSA officials noted, however, that the SPDs received by Labor were often out of date and that it was costly to store them. In 1993 we agreed that Labor should 14 A qualified pension plan is an employer pension plan that receives preferential tax treatment in exchange for satisfying certain requirements established in the Internal Revenue Code of 1986. Employers or employees receive tax benefits for contributions they make to qualified plans within certain limits. A nonqualified pension plan is an employer-sponsored pension plan that does not meet these requirements. 15 A SEP (402(h) plan is a deferred compensation type retirement plan that allows employers and employees to make deductible contributions toward an employee’s retirement fund. There are specific rules about contribution and deduction limits, which make the plan easier for a smaller employer to administer, but less attractive for a larger employer. 16 A SIMPLE plan (401(k)(11)) is a deferred compensation type retirement plan that certain small employers (including self-employed individuals) can set up for the benefit of their employees. 17 U.S. General Accounting Office, Pension and Welfare Benefits Administration: Opportunities Exist for Improving Management of the Enforcement Program, GAO-02-232 (Washington, D.C.: Mar. 3, 2002). 18 Vesting refers to when a plan participant has earned a right to a benefit that cannot be taken away (i.e., a nonforfeitable right to the participant’s accrued benefit). Page 10 GAO-03-337 Retirement Income Data stop collecting paper copies of SPDs, and instead provide access to electronic versions.19 Labor no longer collects SPDs and public access to them has become much more limited in the last 5 years. The Taxpayer Relief Act of 1997 (P.L. 105-34) amended ERISA so that employers no longer need to file SPDs with Labor. Instead, Labor was authorized to request SPDs from employers as needed, and uses this authority primarily to assist plan participants and beneficiaries in obtaining copies, though it has authority to request them for other purposes. However, since the law, Labor no longer requires that SPDs be filed, and SPDs prepared after the Taxpayer Relief Act are not publicly accessible, either for the general public or for researchers looking to model pension behavior. Tax Information Returns Are Our expert panelists noted that some of the pension details they need can a Source of Information on be found in the administrative data provided by employers and others to Retirement Plans the IRS. In addition to its collection of income tax returns, the IRS also collects “information” returns, such as the W-2, which contain details of employee information that provides valuable details relevant to future retirement income, such as wages, tax-deferred retirement contributions, lump sum distributions, rollovers, and retirement asset balances. In discussing pension data that the IRS has access to from its tax forms, experts from our panel reported that information from the Forms W-2, 5498, or 1099-R could provide important pension details. These forms provide detail on the extent of investments in retirement plans, such as the amount of contributions made to an Individual Retirement Account (IRA) or the amount of money contributed by the employee to deferred compensation plans.20 Form W-2 is a valuable source of pension data because it provides information on whether the employer provides some form of qualified retirement plan. The W-2 also includes amounts deducted from wages for contributions to pension plans, as well as codes that provide more detail on the different kinds of plans to which the contribution was made, such as whether the plan is a SIMPLE, SEP, or some other kind of deferred 19 U.S. General Accounting Office, Management Reform: GAO’s Comments on the National Performance Review’s Recommendations, GAO/OCG-94-1 (Washington, D.C.: Dec. 3, 1993). 20 IRAs authorized by ERISA allow workers to make tax-deductible and nondeductible contributions to an individual account for retirement savings. Page 11 GAO-03-337 Retirement Income Data compensation plan. Besides giving more detailed information on deferred compensation, the W-2 has another advantage in terms of the pension information it provides: employers are required to submit one for every employee. This requirement covers all workers for whom federal income tax or Social Security tax is withheld, including those who do not earn enough to be required to file individual income tax returns and those who are not covered by any pension plan. Form 5498 is a form that financial institutions are required to file for all participants to report all contributions and the fair market value of their IRA accounts. It includes valuable pension information, including IRA contributions; rollover contributions;21 and SEP, SIMPLE, and Roth IRA contributions.22 Experts from the panel stated that the pension information from Form 5498 could be important in tracking an individual’s retirement income balance with a specific custodial financial institution because it provides information on the fair market value of those assets held by the individual. While it does not give information on whether an employee who has separated from an employer converts the IRA into another tax protected IRA or pension account, this information should be reflected in the employee’s federal income tax return. Form 1099-R is another source of information for pension experts, which could provide additional information on pension resources. Form 1099-R is a statement filed by trustees concerning distributions to individuals from pensions, annuities, retirement or Profit-Sharing Plans, and IRAs. However, in many cases the form does not indicate whether the distributions are lump-sum distributions or rollovers into IRAs or other qualified plans. The IRS makes some tax information publicly available through its Statistics of Income (SOI) program, which provides numerous tabulations and articles from its analysis of tax return data. From a sample of Form 1040s, the SOI currently provides aggregate tabulations of information, including taxable IRA distributions, deductible payments to an IRA, payments to a self-employed retirement (Keogh) plan, and taxable 21 A rollover contribution is a direct transfer of pension benefits received as a lump-sum payment to another tax-qualified retirement plan or an IRA free of taxes. In many cases, however, the IRS forms do not indicate whether or not distributions were rolled over. 22 A Roth IRA is a type of individual retirement plan that is similar to a traditional IRA except that contributions are not tax deductible, and that qualified distributions are tax free. Page 12 GAO-03-337 Retirement Income Data pensions and annuities.23 This information is provided as a sum total of amounts accrued for the entire U.S. population that filed income tax returns, and is also broken out in detail by income level. For example, for the year 2000, the SOI provides taxable IRA distributions for about 9 million returns, with distributions totaling about $100 billion. While SOI breaks down these totals by income brackets, the tables do not provide other useful pension information such as pension accrual amounts by race or ethnicity. These demographic characteristics could be added to the data if individual tax return records were linked to the Census Bureau’s detailed household survey records. The SOI staff are preparing an article on retirement related data available from the IRS and, in doing so, will make some retirement related aggregate data, including information from Forms 5498 and W-2, publicly available for the first time. They told us that they are considering making this information available as a part of regular SOI releases, but they currently have no formal plans to do so. The SOI tabulations being prepared include the fair market value of pension plans, broken down by 10-year cohorts. Our expert panelists said that these kinds of aggregate data from IRS forms could help them ensure that their analyses reflect accurate information about retirement assets, such as the fair market value of IRA accounts. Experts Place Priority on Given different possible options for improving retirement income data, Linking Data on Individuals retirement experts showed the greatest interest in increasing the to Administrative Data availability of, and expanding researchers’ access to, data sets on individuals or households linked to administrative data sets. (For characteristics of selected survey data sources, see table 3.) Eighty-one percent of respondents assigned this as a high or highest priority. Data linkage creates new information by matching data about individuals (using names or taxpayer identification numbers) to a second set of records, such as administrative records on pension plans, which provide additional information. The linked data are then preserved as a new data set, with personal identification information removed, that can generate new, fuller information on the population. Linking existing sources of data can provide detailed information with no additional respondent burden and at a much lower cost than is associated with collecting new survey data. 23 Keogh plans are retirement plans for self-employed workers, authorized by the Self-Employed Individuals Retirement Plan Act of 1962 (P.L. 87-792). Page 13 GAO-03-337 Retirement Income Data Our expert panelists made several suggestions for linking individual and household survey data with administrative data sources to help improve the analysis of retirement income and projections of the effects of policy changes on future retirement income. For example, they suggested expanding linkage between the Survey of Income and Program Participation (SIPP) and administrative data sources.24 SIPP is a survey of households conducted by the Census Bureau providing wide-ranging demographic information, including different age cohorts, which makes it an attractive source of information when linked to administrative records. It provides, for example, information about benefits households receive from government programs. One component of this survey deals with retirement and pension plan coverage, in which several pension and retirement related questions are asked. However, it lacks definitive information on features of respondents’ pension plans. Linking SIPP information to administrative data sets is a powerful way to expand information about how different groups will be covered in retirement. For example, linked data sets can indicate the extent to which the availability of various pension features differ for people in different age and demographic groups. Such linkage can also indicate how much knowledge respondents have about their pension plans and their retirement savings options. Panelists said that linkage to SPDs, Form 5500 data, and Social Security earnings and benefit histories would be especially valuable for projecting retirement income for different demographic groups. Survey data on individuals have already been linked to administrative data sources in order to improve retirement income data. The Census Bureau is already linking SIPP records with administrative data related to retirement income, including Social Security earnings and benefit records. However, many potential linkages are hampered by lack of access to needed administrative data. For example, the University of Michigan Institute for Social Research has linked information gathered from HRS survey participants to SPDs gathered either from the employers or by Labor prior to 1997. Unfortunately, this process was made more difficult, and the information less satisfactory, because the HRS researchers could only obtain about 50 percent of the SPDs they were seeking, in part, because employers chose not to provide the SPDs. HRS researchers have also experimented with linking respondent surveys to Social Security earning 24 The SIPP data have, for example, been linked with Social Security earnings records except in cases where respondents were unwilling to provide their Social Security numbers to the Census Bureau. Page 14 GAO-03-337 Retirement Income Data histories for two-thirds of the respondents who permitted SSA to provide the records. Many Factors Limit Experts we consulted believed that data needed for the study of retirement income are not collected or made available because of factors Needed Retirement that include the fragmentation of data collection responsibility, the burden Income and Wealth Data of data collection on respondents, and confidentiality considerations that restrict access. Fragmentation stems from no single agency having a statutory mandate to collect and analyze all the data needed to support a more comprehensive study of retirement income and wealth. Some information is no longer collected out of concern that it was an unnecessary burden on the firms having to submit it and because it was only being used to a limited extent by the government. Finally, certain kinds of data needed to make retirement income projections are not made available because they contain information that must by law be carefully protected against unauthorized disclosure and misuse. Fragmented Retirement Data Although many federal agencies are involved in collecting and analyzing Responsibilities Contributes retirement income and wealth data for a variety of different purposes, no to Data Shortcomings single agency is responsible for these activities. Instead, some agencies— including the Bureau of the Census, Labor, Federal Reserve Board, IRS, and SSA—collect data needed for their specific purposes. For example, the Census Bureau collects retirement income and wealth information on individuals and households using the Current Population Survey (CPS) and SIPP. Labor collects data on pensions for both ERISA enforcement purposes and to track pensions for statistical purposes. The IRS collects data from individuals and firms for tax enforcement purposes. Table 1 shows the major agencies involved in retirement income and wealth data collection, the purpose of their data collection, and a brief description of the information collected. None of these agencies is charged with coordinating all retirement income data collection efforts or planning improvements needed in data collection and analysis. Page 15 GAO-03-337 Retirement Income Data Table 1: Examples of Federal Agency Retirement Income-Related Data Collection Data collection Agency program Purpose of data collection Examples of data collected Census Bureau Survey of Income and Provide information about income -Work experience (Department of Commerce) Program Participation and government program -Earnings (SIPP) participation -Program participation -Benefits received -Property Income -Demographic characteristics Bureau of Labor Statistics National Compensation Provide information on wages -Occupational earnings (Department of Labor) Survey salaries and benefits -Compensation trends -Benefits offered -Benefit participation -Detailed plan provisions Employee Benefits Security Form 5500 Enforce ERISA pension -Type of plan Administration requirements -Number of employees (Department of Labor) -Number of participants -Plan amendments -Plan financial position -Actuarial assumptions -Employer & employee contributions Social Security Administration SSA earnings and Administer the Social Security -Earnings histories benefit records benefit programs -Social Security benefit histories -Supplemental Security Income benefit histories Internal Revenue Service IRS tax records Administer/enforce the tax code -IRS individual 1040 returns (Department of the Treasury) -IRS Information returns -Account balances -Withdrawals Source: GAO’s analysis of data from the U .S. Census Bureau, the U.S. Department of Labor, Social Security Administration, and the U.S. Treasury Department. Agencies involved in the analysis of retirement income and wealth data often limit their analysis to a portion of the retirement income and wealth information. Panel members noted that many of the agencies place little or no priority on a comprehensive analysis of retirement income and wealth data. Instead, agencies focus their data collection and analysis on data needed to address their mission. For example, Labor’s EBSA collects information on compliance with ERISA regulations, including Form 5500 submissions. Although EBSA’s strategic plan includes retirement income data analysis efforts, its principal focus is on enforcement of ERISA. Similarly, the Bureau of the Census collects retirement related information in the SIPP and other surveys but its analysis of this information is primarily in the context of its income and poverty measurement mission. The SSA’s MINT model uses an estimate of workers’ pensions based on SIPP data from the Census Bureau, which generally relies on survey responses, has been criticized for using inaccurate estimates of Page 16 GAO-03-337 Retirement Income Data nonearnings income, including retirement income. Because the Census Bureau data provided to the SSA for the MINT was not gathered with it in mind or coordinated for its use, the MINT model has had to use simplifications and assumptions of these data which makes its final model less useful for policy makers.25 Despite this fragmentation, some agencies have attempted to increase communication between federal agencies concerning data collection efforts related to retirement income. For example, the BLS’ strategic goals include improving the accuracy, efficiency, and relevancy of U.S. economic statistics and enhancing coordination with other agencies. The Census Bureau’s strategic goals likewise include an emphasis on providing accurate, timely, and accessible information on the U.S. population and economy, and to maintain relationships with agencies compiling administrative record data. Both the Census Bureau and the BLS are members of the Federal Interagency Forum on Aging Related Statistics,26 which has a goal of improving both the quality and the use of aging related data. In addition, BLS is one of three agencies that share responsibility for leading the Inter-Departmental Committee on Employment-Related Health Insurance Surveys. This committee of a dozen members was created in 1998 to improve coordination and reduce respondent burden by reducing redundant requests for information. Respondent Burden Members of our expert panel reported that respondent burden, as well as Considerations Limit requirements set up to limit respondent burden, could hinder agencies’ Available Data efforts to obtain retirement-related information. Panel members noted that answering survey questions about retirement income and wealth could be a substantial burden on respondents. They acknowledged that asking for too much information could result in partial responses, erroneous responses, or, in some cases, a reduction in the overall response rate. 25 SSA statistically matched defined benefit pension plan characteristics from the Pension Benefit Guarantee Corporation to the survey responses. 26 The Federal Interagency Forum on Aging Related Statistics was established in 1986, with the goal of bringing together federal agencies that share a common interest in improving aging related data. Member agencies include: National Institute of Aging, National Center for Health Statistics, Census Bureau, Administration on Aging, Agency for Healthcare Research and Quality, BLS, Centers for Medicare and Medicaid Services, Department of Veterans Affairs, Office of Management and Budget (OMB), Office of the Assistant Secretary for Planning and Evaluation in the Health and Human Services Administration, and SSA. Page 17 GAO-03-337 Retirement Income Data However, efforts to reduce respondent burden may also limit the collection of retirement information. Legislation requires OMB to review surveys before they are used to collect data. The Paperwork Reduction Act of 1995 (P.L. 104-13) and similar previous legislation27 is designed to minimize the paperwork burden on the public while at the same time recognize the importance of information to the successful completion of agency missions. The act requires OMB to approve all existing and new collections of information by federal agencies. In approving agency collection efforts, OMB must weigh the burden to the public against the practical utility of the information to the agency. Panel members noted that agencies were reluctant to propose additional data collection unless they could clearly establish that the benefit outweighed the perceived burden. Panel members noted that efforts to reduce existing data collection requirements sometimes result in a loss of information. For example, the 1997 elimination of the requirement that firms routinely submit SPDs was connected to an effort to reduce the respondent burden on employers. Privacy and Confidentiality While restrictions on the collection and use of retirement data are critical Concerns Limit the for the protection of personal information, some panel members noted Collection and Use of that these restrictions also limit the availability of such information. Several laws exist to protect individuals’ rights to privacy and protect the Retirement Data confidentiality of personal and proprietary information held by federal agencies. For example, laws set strict requirements to protect data collected by the Census Bureau and to limit the use of taxpayer data.28 Implementing such laws requires federal agencies to restrict access to data they collect. For example, the Census Bureau’s data set that links the SIPP with earnings and benefit records from SSA is not available to the public. Protecting the confidentiality of such linked data sets is particularly crucial because linked data sets may be more detailed or more sensitive than the component data sets were before they were linked. Agency officials must remove personal identifiers such as Social Security numbers, names, and addresses. Even without these personal identifiers, 27 Previous legislation includes the Paperwork Reduction Act of 1980 (P.L. 96-511) and the Paperwork Reduction Reauthorization Act of 1986 (P.L. 99-500). 28 See 13 U.S.C. 9 and 26 U.S.C. 6103. An exception in 26 U.S.C. 6103(j) authorizes the furnishing of return information to Census “for the purpose, but only to the extent necessary in the structuring, of censuses and…conducting related statistical activities authorized by law.” Page 18 GAO-03-337 Retirement Income Data as more and more information is linked, the risks that individuals could be identified increases. The computer files for these nonpublic data sets are available only at a limited number of secure research data centers to approved analysts working on approved projects.29 Outside analysts working with these data sets must be sworn Census Bureau officers and their work must serve, at least in part, to support the Census Bureau’s mission. If analysts permitted to use these data combine any other data with the restricted data, the combined data are subject to the Title 13 protections. Analysts are not allowed to make copies of the data or remove data from the secure data center.30 Before taking any of the results of their work from the center, Census Bureau staff must review the results to ensure that they meet the agency’s requirements for confidentiality protection. Thus, the results that can be taken out of the center are limited to statistical results that do not disclose data for specific individuals. Also, the external researchers must agree that the results of their work will be available in the public domain and not maintained as proprietary information. However, despite the government’s efforts to protect the information they collect from misuse, surveys of the public and Census Bureau interviewers indicate that people are apprehensive about the government’s use of personal information. In a survey conducted both before and after (or just prior to) the 2000 Census of Population and Housing, about half of the respondents (51 percent before and 50 percent during) indicated they thought the Census Bureau’s promise of confidentiality could be trusted, down from about 79 percent in 1990. A smaller, but still substantial proportion of the workers conducting census interviews and providing those promises to respondents also indicated a lack of trust in the Census Bureau’s assurances. A 1998 study indicated that 16.7 percent of Census Bureau interviewers and 32.2 percent of non-Census Bureau interviewers believed that the Bureau would give individual survey data to government agencies such as the Federal Bureau of Investigation, the Central Intelligence Agency, the Immigration and Naturalization Service, and the IRS. Federal law prohibits the Census Bureau from sharing information about individuals with these agencies. Census Bureau information about 29 The Census Bureau’s research data centers are located in Washington, D.C.; Boston, Massachusetts; Pittsburgh, Pennsylvania; Los Angeles, California; Berkeley, California; Durham, North Carolina; Ann Arbor, Michigan; and Chicago, Illinois. 30 Individuals with access are subject to penalties, including fine and imprisonment if they disclose any confidential information. Page 19 GAO-03-337 Retirement Income Data particular individuals or businesses is only available for statistical purposes, not for law enforcement purposes. Public distrust of federal agencies’ use of their personal information can undermine people’s willingness to participate in federal surveys, potentially making the information collected less reliable.31 An article in the Journal of Official Statistics noted that in the 1990s the rate at which people refused to participate had risen for six surveys conducted by the Census Bureau.32 For questions about types of income in the March 1999 CPS survey, for example, the percentage of respondents providing data for particular items ranged from 33 percent to 78 percent. Low response rates can undermine analysts’ statistical projections if the individuals who choose not to respond differ in important respects from those who provide data. If the remaining respondents are dissimilar to the population being surveyed, conclusions about the population may not be reliable. For example, if those who chose not to respond have higher incomes, estimates of the populations’ income may not be reliable. Statisticians make adjustments that can mitigate this problem, but the lower response rates are, the more uncertainties remain. Federal agencies and researchers continue to explore options to maximize data usefulness without compromising respondent privacy and confidentiality.33 For example, the Census Bureau has received permission from the IRS to link the survey records to additional items from IRS and SSA records.34 In addition, the Census Bureau has recently increased the 31 See Eleanor Singer’s study “Public Perceptions of Confidentiality and Attitudes Toward Data Sharing By Federal Agencies” in Confidentiality, Disclosure, and Data Access: Theory and Practical Applications for Statistical Agencies, Pat Doyle, et al., (Amsterdam: Elsevier Science B.V., 2001). In some cases after controlling for other factors associated with response rates, such as respondents’ age, the length of the form, race, and education, privacy concerns were not significant predictors of response rates. 32 B.K. Atrostic et al., “Nonresponse in U.S. Government Household Surveys: Consistent Measures, Recent Trends, and New Insights,” Journal of Official Statistics, vol. 17 no. 2, 2001, 209-226. 33 We have discussed options for protecting privacy and confidentiality while conducting record linkage in U.S. General Accounting Office, Record Linkage and Privacy: Issues in Creating New Federal Research and Statistical Information, GAO-01-126SP (Washington, D.C.: April 2001). 34 26 CFR Part 301, Federal Register vol. 68, no. 13, January 21, 2003, p. 2691.The IRS shares responsibility with SSA for protecting the confidentiality of Social Security earnings records compiled from W-2 forms submitted to the IRS. Page 20 GAO-03-337 Retirement Income Data number of its research data centers where approved analysts working on approved projects can access confidential data. Researchers continue to explore statistical techniques for providing more information from survey data sets while reducing the risks that confidential information will be compromised. One well-known technique for doing this, “top coding,” involves reporting in data files only that an individual respondent’s income, for example, exceeds the “top code” amount, not the actual value. In this way, so many individuals are included in the high-income group that their identities cannot be determined. More sophisticated techniques include the use of methods to substitute artificial records containing estimated values based on knowledge of the real data. These simulated subjects are assigned number values selected to ensure that relationships between important variables are preserved. This allows people to remain anonymous to the researchers using the data. However, pension experts state that this method can only accommodate certain kinds of variables and, therefore, complex relationships between variables may not be maintained. 35 Recent legislation may enable the Census Bureau and BLS to link their data on businesses for statistical purposes. In December 2002, the Congress enacted the Confidential Information Protection and Statistical Efficiency Act of 2002 as part of the E-Government Act of 2002 (P.L. 107- 347). This act permits designated statistical agencies to share information concerning businesses for statistical purposes, but not information concerning individuals or households. It authorizes three agencies—the Census Bureau, the Bureau of Economic Analysis, and BLS—to share data on businesses with one another for statistical purposes. Conclusions With the proportion of retirees to workers expected to increase dramatically over the next couple of decades, important decisions lie ahead. Access to retirement income data needed to inform those decisions has actually decreased in some respects, despite the recommendations of 35 Critics of these efforts say that these techniques cannot preserve all the relationships between variables in a data set. Moreover, the techniques are not workable for variables, such as the age of retirement, that do not conform to a simple mathematical pattern. The frequency at which people retire, for example, is spiked at certain ages, such as 60, 62, and 65 years of age, which makes it difficult to summarize the data using a statistical formula. One of our panelists noted, however, that if only one variable has such characteristics, the actual data for that variable could be left as long as other variables were masked. Page 21 GAO-03-337 Retirement Income Data the 1997 National Research Council’s Panel on Retirement Income Modeling. Although many sources of useful retirement income data remain, retirement analysts we consulted cited shortcomings. They noted, for example, that data shortcomings persist when no one federal agency is responsible for coordinating efforts to fill retirement income data needs. Indeed, several such data needs could be met with information the federal government already possesses or to which it already has access. For example, Labor has the authority to obtain existing documents describing the features of private pension plans. However, though it has this authority, it gives employers in its National Compensation Survey a choice about whether to provide them to support Labor’s BLS statistical studies of pension plans. To encourage voluntary participation in the survey, Labor does not make them available to other agencies, outside analysts or the general public. The Census Bureau gathers or collects information about some general features of private pension plans through surveys of households and individuals but has not yet had the opportunity to corroborate and supplement this data using information from respondents’ employers available through Labor. While information on pension and individual retirement accounts is gathered through forms collected by the IRS, the data are not regularly tabulated or linked to survey data, and thus, are not available for the study of pensions. While improvements in data are essential for more reliable forecasts of retirement income, protecting respondents’ information and minimizing the burden data collection efforts impose on firms and individuals are also crucial. To sustain programs for compiling statistics about firms and individuals, respondents must be able to trust that their personal information will not be misused. Finding an appropriate balance between providing wider access to data to support policy analysis and keeping data secure is a persistent and evolving challenge that policymakers must continually address. In addition, federal agencies need to consider both the federal cost of these efforts and the financial and nonfinancial costs imposed on respondents in comparison with the benefits expected from improvements in data. While taking into account these cost considerations, respondent concerns about increased reporting burden, and agency concerns about maintaining confidentiality, the Congress, and the Departments of Labor and the Treasury could take the next steps to help fill some of the data needs. Page 22 GAO-03-337 Retirement Income Data Matter for To facilitate plan participants’, beneficiaries’, and analysts’ timely access to information about employer-provided pension plans, the Congress Congressional should consider directing the Secretary of Labor to obtain from plan Consideration administrators electronic filings of all SPDs and summaries of material modifications required by ERISA and make them publicly available in electronic form. Recommendations To help provide the data needed to inform important policy decisions concerning retirement programs, the Secretary of Labor should direct the Bureau of Labor Statistics to prepare a plan to improve data for analyzing retirement income and wealth in coordination with OMB, the Federal Reserve Board, IRS, and the agencies represented in the Federal Interagency Forum on Aging Related Statistics, including the Census Bureau, SSA, and the National Institute on Aging. The plan should include cost-effective strategies to • make better use of existing statistics by linking survey and administrative data, • improve access to linked data consistent with privacy and confidentiality legislation, and • improve data collected from employers related to retirement income and wealth. For plans in place before these new filing requirements go into effect, and where it is cost-effective, the Secretary of Labor should use existing authority to obtain copies of summary plan descriptions and summaries of material modifications in cases where analysts working on federally conducted or federally sponsored research seek SPDs for statistical purposes. This should assist analysts of retirement income in obtaining information about the features of employer-sponsored benefit plans that are not electronically available. To help analysts improve analyses of retirement plan finances, the Internal Revenue Service should publish on an annual basis aggregate tabulations, such as those prepared in the Statistics of Income Bulletin, of information filed on IRS Forms 5498, 1099-R, and W-2. Agency Comments We provided a draft of this report to the Departments of Commerce, Labor, the Treasury, and the IRS and received technical comments from all four. In response we modified the draft as appropriate. We also provided a draft of this report to the 11 members of our expert panel and modified the Page 23 GAO-03-337 Retirement Income Data draft as appropriate in response to their comments. Commerce had no major comments on the report (see app. VI). Labor agreed on the need for access to accurate data but did not agree with our recommendations to the Secretary of Labor (see app. VII). Regarding our recommendation to collect electronic copies of SPDs, Labor concluded that this is at odds with the Taxpayer Relief Act of 1997, which eliminated requirements that SPDs be regularly filed with Labor. Although Labor still has authority to request SPDs, it had indicated in final regulations concerning SPDs that it generally intended to limit the exercise of its authority to requesting SPDs on behalf of participants and beneficiaries. Labor also indicated that there was little public interest in the SPDs. Although there was not substantial public demand for paper copies of SPD’s in Labor’s Public Disclosure Room in Washington, D.C., in the last decade there has been a great increase in policymakers’ demand for better data on retirement income in light of the rapid future increase in the retiree population. Although in both 1993 and 1995 we supported elimination of paper filing requirements, we also endorsed electronic access to SPDs. We continue to believe that it is time to phase in a requirement that SPDs be filed electronically. The costs involved should be considerably less than those incurred filing and storing paper copies of the documents. The burden imposed on plan administrators would not be unreasonable. Labor estimated in 1998, for example, that it would cost an average of $1.55 to provide SPDs for health benefit plans to each plan participant, of which $1.00 was the estimated cost of materials and postage.36 Cognizant of the shortcomings of many SPDs, experts we consulted nonetheless indicated that better access to SPDs is a top priority for improving retirement income data. Accordingly, we changed our recommendation to a Matter for Consideration for the Congress to direct the Secretary of Labor to obtain from plan administrators electronic filings of all SPDs and summaries of material modifications required by ERISA and make them publicly available in electronic form. Regarding our recommendation that BLS prepare a plan to improve data for analyzing retirement income and wealth, Labor indicated that past efforts to coordinate improvements in retirement income data have not been successful due to privacy concerns, and other issues. They stated that the type of planning and coordination we envisioned usually is the 36 U.S. Department of Labor, “Proposed Amendments to Summary Plan Description Regulations,” Federal Register, vol. 63, no. 174, September 9, 1998, p. 48384. Page 24 GAO-03-337 Retirement Income Data purview of OMB’s Office of Statistical Policy, and that BLS could participate in that coordination. Furthermore, the extra demands placed on staff would not be negligible. In our view, the need for improvements in retirement income data warrants renewed efforts to address the priorities identified by the experts we consulted. The recommendations of these experts focused primarily on improved use of existing data to support policy analysis, and not on additional data collection. We recognize OMB’s role overseeing information collection and developing policies to improve government statistics. However, this does not preclude efforts by other agencies to take the lead in developing plans for improvements focused on data within specific subject areas such as retirement income. Because OMB does not have the retirement income focus needed to coordinate in this way, we have retained our original recommendation. That recommendation specifically identifies OMB as one of the agencies that should be involved in the development of a plan, and the plan should be developed in a manner consistent with OMB’s policy. With respect to our recommendation regarding provision of SPDs before new electronic filing requirements go into effect, Labor stated that the need to protect the confidentiality of survey data may hamper wider access to SPDs. BLS, for example, assures respondents to its National Compensation Survey that their identities will be kept confidential. The Secretary’s authority to request SPDs is delegated to EBSA. If BLS were to obtain SPDs from EBSA it would have to reveal the identity of its respondents and therefore would have to obtain their consent. We conclude that the need for improvements in retirement income data warrants Labor’s use of its existing authority to obtain SPDs for analysts engaged in federally conducted or federally sponsored research. Arrangements could be developed through which BLS and other statistical agencies could both obtain SPDs and protect the identity of respondents. They could, for example, request SPDs from a larger number of employers without identifying which employers were being surveyed. This is the kind of improved access to data that we envisioned BLS could take the lead in identifying in coordination with other agencies. We are sending copies of this report to the Secretary of Labor, the Secretary of the Treasury, the Secretary of Commerce, and the Commissioner of the Internal Revenue Service. We will also make copies available to others on request. In addition, the report will be available at no charge on GAO’s Web site at http://www.gao.gov/. Page 25 GAO-03-337 Retirement Income Data If you have any questions concerning this report, please contact Barbara Bovbjerg at (202) 512-7215, Robert Parker at (202) 512-9750. See appendix VIII for other contacts and staff acknowledgments. Sincerely yours, Barbara D. Bovbjerg Director, Education, Workforce and Income Security Issues Robert P. Parker Chief Statistician Page 26 GAO-03-337 Retirement Income Data Appendix I: Scope and Methodology Appendix I: Scope and Methodology To identify information that experts say is a priority for improving data for the study of retirement income and wealth, we conducted a Web based survey of experts in the field and convened an 11 member panel of experts to discuss opportunities for improving these data. We used the Web based survey instrument to survey 276 experts in retirement income data.1 Before implementing the survey, we contacted respondents via email and asked them to participate. Out of our initial list of 326 experts, 27 declined to participate, with the majority citing busy schedules or lack of sufficient expertise as their reasons. In addition, we concluded that we had inaccurate or out-of-date E-mail addresses for 23 of the experts. We studied available research and interviewed experts in order to develop a questionnaire of options to improve retirement income. The questionnaire asked respondents to indicate the priority (from highest to lowest) they would place on 22 actions to improve retirement income data. Respondents were asked to rate each action independently, without making comparisons between the actions. We pretested the content and format of the questionnaire with 6 experts in the area of retirement income. The questionnaire was then refined and posted on our Web site and an E-mail message informed participants of its availability. This E-mail message also contained a unique user name and password that allowed each respondent to log on and fill out his or her own questionnaire. As of December 12, 2002, 190 of the experts responded to the survey (a 69% response rate). Eighteen percent of respondents indicated that they were affiliated with federal agencies, about half were affiliated with colleges and universities, 24 percent were affiliated with other nonprofit organizations, and the remaining 9 percent were affiliated with for profit or other organizations. Our preliminary results of the survey identified two areas for improvement that respondents most often cited as having the highest priority: (1) matching survey and administrative data and (2) employer data. We used these areas to serve as the principal topics at a 1-day expert panel meeting at our headquarters on September 10, 2002. The 11 panelists included 5 federal officials with responsibilities related to retirement income data, 3 university based analysts, and 3 from private not-for-profit agencies. 1 We identified these people through literature searches on topics related to retirement income data and by asking members of our Retirement Advisory Panel for suggested names and in turn asking them for additional names. Page 27 GAO-03-337 Retirement Income Data Appendix I: Scope and Methodology Barbara Bovbjerg, Director, Education, Workforce, and Income Security Issues, and Robert Parker, Chief Statistician, moderated the discussion. (For a summary of the panel’s discussion and a list of panelists, see app. IV.) To identify factors limiting the availability of information needed for the study of retirement income and wealth, we reviewed documents obtained from and interviewed officials at the Department of Labor’s (Labor) Employee Benefit Security Administration (EBSA), and Bureau of Labor Statistics (BLS), the Census Bureau, the Treasury Department’s Office of Tax Analysis, the Internal Revenue Service’s (IRS) Statistics of Income Division, the National Institute of Health’s National Institute on Aging, and the Social Security Administration’s (SSA) Office of Research, Evaluation, and Statistics. We also attended conferences related to retirement income analysis sponsored by the Retirement Research Consortium and the Society of Actuaries and interviewed analysts at the Urban Institute, The Brookings Institution, the Employee Benefit Research Institute, and the National Research Council. The scope of our work did not include an evaluation of estimated costs and benefits of specific proposals for improving retirement income data. We did not independently verify the federal funding figures provided to us by longitudinal survey administrators or agencies sponsoring the surveys. Page 28 GAO-03-337 Retirement Income Data Appendix II: Status of Recommendations From Appendix II: Status of Recommendations From the 1997 Report of the Panel on Retirement Income Modeling the 1997 Report of the Panel on Retirement Income Modeling Below are recommendations concerning retirement income data needs excerpted from the 1997 report of the National Research Council’s Panel on Retirement Income Modeling followed by summaries on the status of each as of December 2002.1 1. Continue support of longitudinal studies Recommendation: Existing panel surveys of middle-aged and older people should receive continued government support. Longitudinal data from these surveys are essential to analyze retirement and savings decisions and determine behavioral responses to changes in public and private sector policies. Such analyses in turn are essential to develop better models for forecasting the likely effects of alternative policy proposals on retirement income security. In particular, the Health and Retirement Study (HRS) and Asset and Home Dynamics Among the Oldest Old (AHEAD) surveys should receive continued support. These surveys should be refreshed periodically with new cohorts in order to offer insight into how behavior changes over time. Status: As shown in table 2, the amounts of federal support for three major longitudinal surveys have been sustained. The HRS and AHEAD studies, which are now jointly funded, have increased after taking the effects of inflation in account. Table 2: Federal Outlays for Selected Longitudinal Studies—Fiscal Years 1997-2001 a Millions of constant fiscal year 2001 dollars 1997 1998 1999 2000 2001 2002 HRS and AHEAD $6.8 $7.0 $7.4 $9.0 $9.4 $10.1 b National Longitudinal Studies 14.0 11.8 13.2 12.8 12.8 12.4 c Panel Study of Income Dynamics 2.5 2.5 3.7 3.7 3.6 2.8 Total $23.3 $21.3 $24.3 $25.5 $25.8 $25.3 Source: GAO analysis of data from the National Institute of Aging and the University of Michigan’s Institute for Social Research. a These figures are adjusted for inflation using the Bureau of Economic Analysis’s gross domestic product price index. b NLS c PSID 1 Constance F. Citro and Eric A. Hanushek, eds., Assessing Policies for Retirement Income: Needs for Data, Research and Models (Washington, D.C.: National Academy Press, 1997). Page 29 GAO-03-337 Retirement Income Data Appendix II: Status of Recommendations From the 1997 Report of the Panel on Retirement Income Modeling The National Institute on Aging continues to fund the HRS and AHEAD studies. Supplemental funding also comes from SSA. Both the AHEAD and HRS studies have been refreshed with new cohorts. In 1998, interviews began for a cohort of people born from 1924 to 1930 and a cohort of people born from 1942 to 1947. The National Science Foundation continues to fund the PSID. Additional support comes from the National Institute on Aging, the National Institute on Child Health and Human Development, and the Departments of Health and Human Services, Agriculture, Housing and Urban Development (HUD), and Labor. Labor sponsors the National Longitudinal Survey (NLS) through BLS. The NLS is conducted under contract with researchers at Ohio State University, the University of Chicago, the Census Bureau and the University of Wisconsin. In addition to Labor funding, financial support has come from agencies including the National Institute of Child Health and Human Development, the National Institute on Aging, the National Institute on Drug Abuse, the National Institute on Alcohol Abuse and Alcoholism, and the Departments of Defense, Justice, and Education. 2. Measure expenditures Recommendation: Panel surveys of middle-aged and older people should experiment with methods to develop measures of families’ total expenditures and expenditures on housing and medical care. Such consumption measures are important for projections of economic well being in retirement. Status: The HRS and AHEAD surveys provide on a longitudinal basis respondents’ estimates of many categories of expenditures including housing expenditures, total out-of-pocket medical expenditures, and total expenditures, total expenditures relative to income, as well as information on savings preferences. Labor’s Consumer Expenditure Survey provides cross-sectional rather than longitudinal data on many types of expenditures, including housing expenditures, and medical expenditures, and total expenditures and related income. The American Housing Survey, sponsored by HUD, provides detail on housing expenditures . The proposed American Community Survey also would provide limited detail on housing expenditures. The Medical Expenditure Panel Survey (MEPS) provides extensive data on medical expenditures. Page 30 GAO-03-337 Retirement Income Data Appendix II: Status of Recommendations From the 1997 Report of the Panel on Retirement Income Modeling 3. Gather more data on younger people Recommendation: Panel surveys of younger people, such as the National Longitudinal Survey of Youth (NLSY), should include detailed questionnaire modules on pension coverage, wealth, health status, and retirement- and savings-related expectations. Such information is needed to fully understand life-cycle behavior and to track the disparities in income and wealth that are evident by middle age. Status: The NLSY asks about the total amount of retirement savings, amounts contributed, and amounts withdrawn, and pension coverage. It also provides information about assets and debt and limited information about health (height, weight, and general evaluation of health), but not retirement or savings expectations. In 1995-99 interviews, the National Longitudinal Survey of Young Women (NLSYW) asked respondents the extent to which they agreed or disagree with statements such as “Work is the most meaningful part of life” and “People who don’t retire when they can afford to are foolish.” The NLSYW also asked respondents about the availability of a retirement or pension plans in 1978 and in each round from 1978, 1983, through 1999. The 1991 and 1995-99 rounds of the survey included more detailed questions on amounts in retirement plans, formula for calculating benefits for defined benefit plans provided by current and previous employers, vesting status, and expectations about retirement, such as expected amounts of benefits. 4. Collaborate to improve data quality and utility Recommendation: Agencies and researchers involved in retirement income-related panel surveys of individuals, and other surveys as appropriate (such as the Survey of Consumer Finances (SCF) and the Survey of Income and Program Participation (SIPP)), should collaborate regularly in reviewing questionnaire content and data collection practices to identify ways to improve data quality and utility. For example, the bracketing technique used in HRS and AHEAD that has been demonstrated to reduce nonresponse to important items should be adopted in other surveys. Also, such surveys might include a common core of questions on specific topics. The National Institute on Aging should facilitate such collaborative efforts. Status: Federal agencies collaborate through entities such as (1) the Federal Interagency Forum on Aging-Related Statistics, originally established by the National Institute on Aging, National Center for Health Page 31 GAO-03-337 Retirement Income Data Appendix II: Status of Recommendations From the 1997 Report of the Panel on Retirement Income Modeling Statistics, and Census Bureau and (2) the Inter-Departmental Committee on Employment-Related Health Insurance Surveys, headed by the BLS, National Center for Health Statistics, and the Agency for Healthcare Research and Quality. The Federal Committee on Statistical Methodology and the Interagency Committee on Confidentiality and Data Access have coordinated research on techniques such as the bracketing technique are now widely used and researchers do collaborate on questionnaire design. In addition, the Census Bureau and other statistical agencies seek comments from a wide range of researchers on the content of questions before fielding survey instruments. For example, the Interagency Committee on the SIPP, which consists of representatives of interested federal agencies, and the American Statistical Association’s Survey Research Methods Section SIPP Working Group, which consists of federal and non-federal analysts, both advise the Census Bureau on that survey. 5. Establish interagency task force on employer data Recommendation: Labor should establish an interagency task force on employer data to specify an integrated plan for collecting retirement income-related information. The plan should specify short-term and long- term goals that consider user needs, resource constraints, and the problems of obtaining information from employers due to such factors as low response rates, locating the appropriate respondents, and confidentiality concerns. The task force should involve researchers, private benefit consultants, and representatives of public and private employers in its work. Status: According to officials at the Department’s EBSA, Labor has not made this a priority because its resources are limited. 6. Gather more data on benefit plan offerings Recommendation: The employer data collection plan should include short-term and long-term goals for obtaining improved information on the distribution across employers of all benefit plan offerings (including pensions, health insurance, disability insurance, retiree health insurance, life insurance). Comprehensive baseline information is a priority need, along with a plan for regular updating. Needed data elements include benefit plan characteristics and costs, employer characteristics (e.g., number of employees, financial characteristics, wage structure), and workforce characteristics (e.g., age structure) for public and private employers and the self-employed. Page 32 GAO-03-337 Retirement Income Data Appendix II: Status of Recommendations From the 1997 Report of the Panel on Retirement Income Modeling Status: Labor’s BLS has increased the amount of information gathered through its National Compensation Survey, which includes components that had been gathered through the Employee Benefits Survey, the Employment Cost Index survey, and the Occupational Compensation Survey. The Employee Benefits Survey did not provide tabulations of the estimated percentage of establishments providing specified types of benefits, but the National Compensation Survey does. The survey includes components for state and local government employers, medium and large private employers, and small private employers, but not federal employers or the self-employed. The survey provides data on number of employees, employer and employee costs, wage structure, and the characteristics of retirement benefit plans. It does provide data on the employer’s cost of providing defined benefit and defined contribution retirement programs. It does not provide information on financial characteristics of plans (such as assets and liabilities) and does not provide information on the age structure of the workforce. In addition, the MEPS, sponsored by the Agency for Healthcare Research and Quality, provides additional data on health insurance plans. 7. Redesign and enhance employee benefits survey. Recommendation: The employer data collection task force should give priority to redesigning and enhancing existing data collection systems on employer benefit offerings and related topics. Such systems include the Employee Benefits Survey, which currently provides information for broad categories of employe[rs] but not for employe[es], and the Form 5500 data series, which serves regulatory purposes and currently has limited research use. Consideration should be given to improving the Form 5500 series, including: • making the data more timely and accessible (e.g., on-line); • linking records over time to provide panel data; • merging the Form 5500 benefit plan information with the kind of employer financial characteristics found in the Compustat database; • working to standardize the reporting for health care and disability plans, so that they can be added to the Form 5500 database; and • finding ways to add information about benefit plan features to the database, perhaps by abstracting analytically useful information from the narrative plan descriptions that are filed with the Form 5500. Status: The Employee Benefits Survey (now part of the National Compensation Survey) continues to provide data for categories of employers, not categories of employees. It provides data by employer Page 33 GAO-03-337 Retirement Income Data Appendix II: Status of Recommendations From the 1997 Report of the Panel on Retirement Income Modeling industry group, employer employment size group, employee union status, employee occupational group, and employer geographic location. BLS has begun providing aggregate estimates for all private employers, rather than only providing data separately for small private employers, and large private employers, as it did in the past. However, the survey does not cover federal, military, agricultural, fishing, forestry, or private household employers. Data from IRS Form 5500 continues to become available well after the end of the reporting year. (Data for 1998 or the 1998-99 reporting year became available in 2002.) In part, this is because the IRS deadline for submitting the forms is 7 months after the end of the reporting year. For example, for a firm with a 1998 reporting year beginning December 15, 1998, and ending December 14, 1999, the Form 5500 was due July 31, 2000. In addition, Labor takes several months to review and edit the returns before making them publicly available. Labor makes Form 5500 data files available to researchers and policy analysts, but its Web site does not include links to the data. A private firm provides a Web site with images of the completed forms, but not compiled data sets. Labor does not link Form 5500 reports by firm to facilitate longitudinal analysis. Linking these consolidated reports of publicly held companies with Form 5500 data is difficult because these reports can cover only parts of a company, more than one company, or privately held companies. Some researchers have linked these data for selected firms. Labor no longer requires that it receives the summary plan descriptions regularly, and as a result the public no longer has access rights to new or revised versions. Labor does, however, incorporate some data from these plans in its National Compensation Survey data. This includes, for example, information on normal retirement ages and formulas for calculating employer contributions to pension plans. 8. Gather more data on labor demand for older workers Recommendation: The employer data collection plan should include short-term and long-term goals for obtaining information on labor demand for older workers and the factors that may affect that demand. Needed data elements include employment patterns of older workers, compensation and benefit costs by age, and worker productivity by age. Very little information on these topics is currently available, and some raise difficult measurement issues. A reasonable short-term goal is to sponsor case studies of employers that can help identify important variables and feasible means of collecting them on a larger scale. Page 34 GAO-03-337 Retirement Income Data Appendix II: Status of Recommendations From the 1997 Report of the Panel on Retirement Income Modeling Status: The Health and Retirement Study and other surveys provide much data on employment patterns and salary and wages of older workers. Little has been done, however, concerning compensation and benefits costs or productivity by age. 9. Conduct longitudinal survey of employers and their workers Recommendation: The employer data collection task force should consider the feasibility and cost-effectiveness of a panel survey, which is periodically refreshed that collects detailed information on employers and their workers. Such a survey should cover the full universe, including private for profit, nonprofit, and government employers, and the self- employed. Longitudinal data from an employer-based survey are needed to analyze the factors that affect employer decisions about recruitment and retention of older workers and benefit plan offerings and how these decisions, in turn, affect workers. Status: Such a task force has not been formed and no survey like the one the panel recommended has been undertaken. BLS has developed a longitudinal database of business establishments, the “LDB”, based on data from states’ unemployment insurance programs, and the Census Bureau continues to maintain a longitudinal establishment database covering all private nonagricultural establishments. Neither of these databases includes data on pensions or other retirement plans. The BLS database provides data on employees’ hours and wages. The Census Bureau database also provides data on employment and wages, and periodically includes data on employer contributions to pension plans and health insurance. To some extent the HRS links data from individuals and their employers on a longitudinal basis, but it tracks the individuals, not the firms over time. The E-Government Act of 2002 (P.L. 107-347) may facilitate collaboration between the Census Bureau and BLS and could provide for linking National Compensation Survey (NCS) data on retirement plans to the Census Bureau’s data on business establishments. 10. Gather more data from employers of HRS/AHEAD sample members Recommendation: HRS and AHEAD should develop and implement a plan for obtaining information on a continuing basis on the pension and health insurance offerings of the employers of the HRS/AHEAD sample members. Page 35 GAO-03-337 Retirement Income Data Appendix II: Status of Recommendations From the 1997 Report of the Panel on Retirement Income Modeling Status: Data from HRS sample members’ employers is available. Similar efforts have been undertaken with other surveys, such as the National Longitudinal Study of Mature Women. 11. Match panel survey responses and key administrative records Recommendation: Matched files of panel survey responses and key administrative records should be regularly produced for retirement- income-related policy analysis and projection purposes. Examples include exact matches of survey records and Social Security earnings histories and benefit records, Medicare and Medicaid records, and the National Death Index. The added information in matched files is obtainable at low marginal cost and is essential for analysis of retirement and savings decisions and the effect of medical care use and expenditures on retirement security. Status: Some matches between administrative data and panel survey data have been achieved. HRS investigators have completed matches between HRS and Social Security earnings and benefit record data, National Death Index data, and Medicare records. These are available on a restricted basis to selected researchers and policy analysts. The Census Bureau has matched SIPP data with Social Security earnings records and extracts from individuals’ IRS tax return data. The Census Bureau’s Longitudinal Employer Household Dynamics (LEHD) project is exploring options for more extensive matches between Form 5500 data from employers and Census Bureau survey data on establishments.2 Recently finalized Treasury regulations give the Census Bureau access to 2 The LEHD program, which was started in 1998, is designed to evaluate and improve the quality of data collected in the Census Bureau’s demographic and economic censuses and surveys through longitudinal analysis. The program combines longitudinal micro data from federal and state administrative data on employers and employees with these census and survey data. In addition to its use to improve its census and surveys, the Census Bureau conducts policy- relevant research on labor force and employment issues and creates new data products. Currently, the LEHD program provides quarterly workforce indicators for a number of participating states. The linkage of Form 5500 data and Census Bureau establishment data is described in Julia Lane, et al., “New Uses of Health and Pension Information: The 5500 file at the Census Bureau, LEHD Technical Paper No. TP-2002-03” (Washington, D.C.: U.S. Bureau of the Census, January 2002). Page 36 GAO-03-337 Retirement Income Data Appendix II: Status of Recommendations From the 1997 Report of the Panel on Retirement Income Modeling specified IRS tax return records to support SIPP and LEHD data linkage efforts.3 Also, researchers at the Employee Benefit Research Institute (EBRI) linked survey data to state Medicaid records in order to develop state specific projections of Medicaid expenditures. 12. Increase researchers’ access to exact match files safeguarding confidentiality Recommendation: Agencies should collaborate on the development and oversight of matched data sets for individuals and employers, with input from researchers on content. They should also vigorously explore creative solutions for providing research access to exact match files that safeguard the confidentiality of individual responses. Possible solutions include: (1) developing public use files that contain summary variables derived from the administrative records portion of the matched file (2) requiring researchers to sign nondisclosure agreements with significant penalties for violations; and (3) providing researchers with access to matched files on site at secure data centers. Status: Access to linked data sets remains quite limited. Access to HRS linked data for example, is typically made available via a rigorous application process resulting in a data use agreement with the University of Michigan. To date, none of the linked data sets are available in encrypted public use files. However, according to HRS researchers, “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.” (One such method is the further use of data centers, which provide access to restricted information, including linked data sets, for approved researchers working on data sets. For more information on data centers, see page 19). 3 Final regulation, “Disclosure of Return Information to the Bureau of the Census, Department of the Treasury, Internal Revenue Service, 26 CFR Part 301,” Federal Register vol. 68, no. 13, January 21, 2003, p. 2691. Page 37 GAO-03-337 Retirement Income Data Appendix II: Status of Recommendations From the 1997 Report of the Panel on Retirement Income Modeling 13. Fund regular evaluation of data quality Recommendation: Budgets for retirement income-related surveys should include sufficient resources for regular evaluation of data quality. Evaluation methods include reinterviewing sub-samples of respondents to measure consistency of reporting; experimentation with alternative question wording to identify possible reporting problems; and comparing survey estimates with administrative records to determine the completeness and accuracy of survey reporting, taking care to adjust for differences in definitions and other aspects of the two sources. Status: Several studies using the recommended methods have been conducted, with mixed results. The Census Bureau conducted a study comparing estimates of various types of 1990-96 incomes in SIPP and CPS to benchmark data estimated by the Census Bureau from the personal income estimates in the National Income and Product Accounts. The study found disparities between the survey based Census estimates and the administrative record based personal income estimates that could not be explained by differences in definitions.4 For 1996, the aggregate wages and salary estimate based on CPS survey data was 102 percent of the benchmark based on administrative records and the CPS social security and railroad retirement benefit payments data were 92 percent of the benchmark. In contrast, the study found more substantial disparities for several other types of income. The aggregate CPS data for property income was 71 percent of the benchmark, and CPS data for pension income was 77 percent of the benchmark. The study is being replicated with data for 1999. Other studies have noted definitional and quality differences between estimates of personal savings from the Flow of Funds Accounts and National Income and Product Accounts.5 Another study compared estimates of wealth from the SCF, PSID, and SIPP.6 4 Marc I. Roemer, Assessing the Quality of the March Current Population Survey and the Survey of Income and Program Participation Income Estimates, 1990-1996, Income Surveys Branch Housing and Household Economics Statistics Division, U.S. Census Bureau (Washington, D.C.: June 16, 2000). 5 U.S. Congressional Budget Office, The Budget and Economic Outlook: Fiscal Years 2004-13, (Washington, D.C., January 2003), p. 33. 6 See for example, R. Curtin, F.T. Juster and J. Morgan, “Survey Estimates of Wealth: An Assessment of Quality” in The Measurement of Saving, Investment and Wealth, R.E. Lipsey and H.S. Tice, editors, National Bureau of Economic Research, Studies in Income and Wealth, 52 (Chicago: University of Chicago Press, 1989). Page 38 GAO-03-337 Retirement Income Data Appendix II: Status of Recommendations From the 1997 Report of the Panel on Retirement Income Modeling The Department of Treasury’s Office of Tax Analysis has compared estimates of pension plan participation and contributions with estimates from the Census Bureau survey data. The Office of Tax Analysis linked W- 2 data with Statement of Income (SOI) data from Form 1040. The results from this data set matched results from Census Bureau surveys, except for low-income households. The tax records obviously don’t include nonfilers, but in addition, by design, the SOI sample includes relatively few low- income filers (in the order of 1 in 5,000 filers), but all filers in the highest income brackets. The Census Bureau surveys such as the SIPP over sample low income households and have much less data for the highest income households. For most income brackets, however, the data match quite well, according to Treasury officials. The Census Bureau periodically assesses the quality of CPS data by reinterviewing a subsample of respondents, but none of the reinterviewing has covered questions on income for at least the last 4 years, according to the Census Bureau quality assurance staff. The Census Bureau has also studied the accuracy of respondent data by matching income data in the March CPS with selected income detail on individual IRS income tax returns and SSA earnings and benefit records. A similar effort is underway using 1999 data. One of the components of the BLS’s National Compensation Survey is a program of re-interviews of a sub-sample of respondents to verify and clarify survey data, including data on retirement plans. HRS investigators have compared employee responses about retirement income to employer provided data and Social Security records and found wide discrepancies. It is unclear to what extent these result from respondents’ limited knowledge of their pensions or data errors. Page 39 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents Data Needs and List of Respondents Questionnaire Page 40 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents Page 41 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents Page 42 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents Page 43 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents Page 44 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents Page 45 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents Page 46 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents Survey Results 1. Data on households and individuals: Please indicate the priority you would place on taking the following actions to improve retirement income data. Highest High Medium Low Lowest No Number priority priority priority priority priority opinion of (percent) (percent) (percent) (percent) (percent) (percent) cases Increase support for longitudinal studies of individuals over 50 years of age, such as the HRS 27.6 40.0 23.8 5.9 2.2 0.5 185 Expand longitudinal studies of retirement savings of younger individuals (age 50 or below) 26.9 39.2 24.2 7.5 1.6 0.5 186 Increase support for other studies of households and individuals' retirement and wealth, such as the SIPP, and the SCF 18.2 28.3 40.6 10.7 1.1 1.1 187 Improve measurement of family and household consumption expenditures in surveys such as the CEX and in panel surveys such as HRS 15.2 33.2 32.1 12.0 6.0 1.6 184 2. Data on employers and employee benefits: Please indicate the priority you would place on taking the following actions to improve retirement income data. Highest High Medium Low Lowest No Number priority priority priority priority priority opinion of (percent) (percent) (percent) (percent) (percent) (percent) cases Improve the design and reporting of retirement income-related data from employers, such as mandatory pension plan disclosures and surveys on benefit plan offerings 33.0 31.4 25.0 8.0 1.6 1.1 188 Conduct research on labor demand for older workers and the factors that may affect that demand 13.8 28.2 31.9 19.1 6.4 0.5 188 Conduct a longitudinal survey of employers and their workers 14.0 25.3 33.9 21.5 3.2 2.2 186 Improve data from the employers of respondents in surveys such as the HRS, and the AHEAD 30.1 35.5 21.9 7.1 2.2 3.3 183 Page 47 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents 3. Other areas for improvement: Please indicate the priority you would place on taking the following actions to improve retirement income data. Highest High Medium Low Lowest No Number priority priority priority priority priority opinion of (percent) (percent) (percent) (percent) (percent) (percent) cases Improve national data on aggregate retirement and non retirement assets such as Flow of Funds, and National Income and Product Accounts 9.2 20.0 30.3 28.6 8.1 3.8 185 Improve matching of survey respondents with key administrative records, while protecting confidentiality 45.2 30.1 14.5 5.4 2.2 2.7 186 Increase researchers' access to datasets on individuals or households matched with administrative data sets 56.8 25.1 9.8 4.4 2.7 1.1 183 Improve collaboration between agencies and researchers to improve questionnaires and data collection and dissemination practices 24.1 40.1 27.3 7.0 0.5 1.1 187 4. Crosscutting actions: Please indicate the priority you would place on taking the following actions to improve retirement income data. Highest High Medium Low Lowest No Number priority priority priority priority priority opinion of (percent) (percent) (percent) (percent) (percent) (percent) cases Collect additional data in existing surveys 23.0 36.1 31.1 6.6 0.5 2.7 183 Begin new surveys 4.0 14.7 29.9 35.0 13.6 2.8 177 Improve the quality of existing surveys 25.7 44.3 22.4 2.7 2.2 2.7 183 Improve the timeliness of existing data 21.6 33.0 24.9 14.1 4.3 2.2 185 Improve researchers' access to existing administrative data, such as Social Security earnings records or Medicare records 50.8 28.6 13.5 3.8 2.2 1.1 185 Improve techniques for protecting the privacy of respondents' survey and administrative data 20.5 22.2 34.1 16.2 3.2 3.8 185 Fund research on retirement income and wealth 28.1 38.9 22.7 7.6 1.6 1.1 185 Fund expanded data collection efforts 18.4 45.3 27.4 5.0 1.7 2.2 179 Fund increased or improved matching of data 31.1 37.3 22.0 5.6 1.7 2.3 177 Fund new or improved modeling efforts 11.7 18.4 40.8 20.7 6.1 2.2 179 Page 48 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents 5. Are there other types of actions that are important for improving the availability, timeliness, or quality of retirement income and wealth data? Percent writing comments Number of cases 43.2 190 6. Which of the following describe the way that you work with retirement income and wealth data? (Check all that apply.) Percent Number checking of cases Data collection and/or data management 26.7 187 Primary data analysis 53.5 187 Secondary data analysis 58.8 187 Policy analysis or development 74.3 187 Other 9.1 187 Percent explaining Number other of cases 6a. If you checked 'other' (above), please specify the way you work with retirement income data in the text box below. 100 17 7. Which of the following describes your affiliation? Percent Number checking of cases Federal government 18.0 189 State or local government 0.0 189 University or college 48.7 189 Other not-for-profit organization 24.3 189 Private for profit organization 7.4 189 Other 1.6 189 Percent explaining Number other of cases 7a. If you checked "other" (above), please specify your affiliation in the text box below. 66.7 3 Page 49 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents 8. Which of the following data sets have you used in your work during the past 5 years? (Check all that apply.) Percent Number checking of cases AHEAD 20.1 184 CEX 35.9 184 CPS 76.6 184 Flow of funds data 30.4 184 HRS 63.6 184 IRS Form 5500 data 38.0 184 NCS (incorporating the Employee Benefits Survey 17.9 184 National Income and Product Accounts data 35.9 184 PSID 35.9 184 SSA administrative files 36.4 184 SCF 47.3 184 SIPP 51.1 184 Other 21.7 184 Percent explaining Number other of cases 8a. If you checked "other" (above), please specify the data sets in the text box below. 100 40 Additional comments: If you would like to make additional comments concerning any topic covered in this questionnaire, please enter them in the textbox below. Percent writing comments Number of cases 14.7 190 Page 50 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents List of Respondents Henry J. Aaron The Brookings Institution to the Survey Julie Agnew College of William and Mary John Ameriks TIAA-CREF Institute Joseph M. Anderson Capital Research Associates Kenneth Apfel University of Texas at Austin Katherine Baicker Dartmouth College Vickie Bajtelsmit Colorado State University Dean Baker Center for Economic and Policy Research Laurel Beedon Public Policy Institute, AARP Dan Beller Employee Benefits Security Administration Department of Labor Keith A. Bender University of Wisconsin-Milwaukee Mark C. Berger University of Kentucky B. Douglas Bernheim Stanford University Merton C. Bernstein Washington University Page 51 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents Joydeep Bhattacharya Iowa State University Andrew Biggs Cato Institute Emily Blank Howard University Robert B. Blancato Matz, Blancato & Associates, Inc. Henning Bohn University of California - Santa Barbara Christopher M. Bone Actuarial Science Associates Barry P. Bosworth The Brookings Institution Linda Smith Brothers University of Wisconsin - Madison Charlie Brown University of Michigan Jeffrey Brown Harvard University Richard V. Burkhauser Cornell University Gary Burtless The Brookings Institution John Y. Campbell Harvard University William J. Carrington Welch Consulting Page 52 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents Yung-Ping Chen University of Massachusetts Boston Constance F. Citro National Research Council Robert L. Clark North Carolina State University Denise M. Clark Feder Semo Clark & Bard, P.C. Joao Cocco London Business School Lee Cohen Social Security Administration Steven B. Cohen Agency for Healthcare Research and Quality| Department of Health and Human Services Courtney C. Coile Wellesley College Craig Copeland EBRI Christopher Cornwell University of Georgia Julia Lynn Coronado Federal Reserve Board David Cutler Harvard University Kimberly Darling SAG Corporation Angus Deaton Princeton University Page 53 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents Jeff Dominitz Carnegie Mellon University Stuart Dorsey Baker University Karen Dynan Federal Reserve Board Ryan D. Edwards Stanford University Douglas W. Elmendorf Federal Reserve Board Gary V. Engelhardt Syracuse University Eric M. Engen American Enterprise Institute William E. Even Miami University Jeff Faux Economic Policy Institute Melissa Favreault Urban Institute Karen W. Ferguson Pension Rights Center Douglas Fore TIAA-CREF Institute Jonathan Barry Forman University of Oklahoma Leora Friedberg University of Virginia Page 54 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents Robert B. Friedland Georgetown University Don Fullerton University of Texas at Austin William G. Gale The Brookings Institution Ron Gebhardtsbauer American Academy of Actuaries Thomas Glass Glass & Co. CPAs Jagdeesh Gokhale Federal Reserve Bank of Cleveland Carol Gold Internal Revenue Service Nancy M. Gordon Census Bureau, Department of Commerce Pierre-Olivier Gourinchas Princeton University Brian Graff American Capital Strategies Matthew Greenwald Matthew Greenwald & Associates, Inc. Lijia Guo University of Central Florida Alan L. Gustman Dartmouth College Steven Haider RAND Page 55 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents Eric Hanushek Stanford University Brian Headd Small Business Administration Joni Hersch Harvard University Roger Hickey Institute for America’s Future Catherine Hill National Academy of Social Insurance Richard Hinz World Bank (formerly, Department of Labor) Lorrie L. Hoffman University of Central Florida Karen C. Holden University of Wisconsin-Madison Sarah Holden Investment Company Institute Kevin M. Hollenbeck W.E. Upjohn Institute Martin Holmer Policy Simulation Group Marjorie Honig City University of New York M. Cindy Hounsell Women’s Institute for a Secure Retirement Warren Hrung Department of the Treasury Page 56 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents Michael D. Hurd RAND Edwin C. Hustead The Hay Group Robert M. Hutchens Cornell University Howard M. Iams Social Security Administration Estelle James Consultant David C. John Heritage Foundation Richard W. Johnson Urban Institute David Joulfaian Department of the Treasury F. Thomas Juster University of Michigan Arthur B. Kennickell Federal Reserve Board Surachai Khitatrakun University of Wisconsin-Madison Kilolo Kijakazi Center on Budget and Policy Priorities Geoffrey Kollmann Congressional Research Service Sophie M. Korczyk Analytical Services Page 57 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents Marvin H. Kosters American Enterprise Institute Douglas L. Kruse Rutgers University Julia Lane Urban Institute Annamaria Lasardi Dartmouth College Ronald Lee University of California, Berkeley Jules Lichtenstein AARP Jeffrey B. Liebman Harvard University Dennis Logue Price College Robin Lumsdaine Brown University David A. Macpherson Florida State University Brigitte C. Madrian University of Chicago Joyce Manchester Social Security Administration Charles F. Manski Northwestern University Ann A. McDermed North Carolina State University Page 58 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents Andrew Metrick University of Pennsylvania Daniel J. Mitchell Heritage Foundation Olivia S. Mitchell University of Pennsylvania H. Fred Mittelstaedt University of Notre Dame Catherine Phillips Montalto Ohio State University James H. Moore Social Security Administration Kathryn L. Moore University of Kentucky Brent R. Moulton Bureau of Economic Analysis, Department of Commerce James Musumeci Southern Illinois University at Carbondale Randall J. Olsen Ohio State University Van Doorn Ooms Committee for Economic Development Peter R. Orszag The Brookings Institution Beverly J. Orth William M. Mercer, Inc Michael Packard Pension Benefit Guaranty Corporation Page 59 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents Michael G. Palumbo Federal Reserve Board Constantijn W. A. Panis RAND Jonathan A. Parker Princeton University Donald O. Parsons George Washington University Christina Paxson Princeton University Cynthia D. Perry Massachusetts Institute of Technology Pamela J. Perun Urban Institute Joseph S. Piacentini Employee Benefits Security Administration Department of Labor Christopher Polk Northwestern University James Poterba Massachusetts Institute of Technology Elizabeth T. Powers University of Illinois at Urbana-Champaign Patrick J. Purcell Congressional Research Service John W. R. Phillips Social Security Administration Anna M. Rappaport William M. Mercer, Inc. Page 60 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents Robert R. Reed III University of Kentucky Cordelia W. Reimers Hunter College John C. Rother AARP John E. Sabelhaus Congressional Budget Office Dallas L. Salisbury EBRI Andrew A. Samwick Dartmouth College Thomas R. Saving Texas A & M University Patricia L. Scahill Ernst & Young LLP Sylvester J. Schieber Watson Wyatt Worldwide Robert F. Schoeni University of Michigan James H. Schultz Brandeis University (retired) John C. Scott American Benefits Council Lois B. Shaw Institute for Women’s Policy Research Stuart A. Sirkin Pension Benefit Guaranty Corporation Page 61 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents Jonathan S. Skinner Dartmouth College Timothy A. Smeeding Syracuse University Kent Smetters Wharton School Karen E. Smith Urban Institute James P. Smith RAND Paul Smith Department of the Treasury Ralph Smith Congressional Budget Office Frank P. Stafford University of Michigan Norman Stein University of Alabama Thomas L. Steinmeier Texas Tech University C. Eugene Steuerle Urban Institute Ann Huff Stevens Yale University Annika E. Sundén Boston College Richard M. Suzman National Institute on Aging Page 62 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents Kenn B. Tacchino Widener University Albert Teplin Federal Reserve Board (Retired) Lawrence H. Thompson Urban Institute Shripad Tuljapurkar Mountain View Research Cori E. Uccello Urban Institute Jack L. VanDerhei Temple University Steven F. Venti Dartmouth College Alice Wade Social Security Administration Daniel Weinberg Census Bureau, Dept. of Commerce David R. Weir University of Michigan Joseph White Case Western Reserve University William J. Wiatrowski Bureau of Labor Statistics, Department of Labor Joshua L. Wiener Oklahoma State University Ernest Wilcox Bureau of Economic Analysis, Department of Commerce Page 63 GAO-03-337 Retirement Income Data Appendix III: GAO Survey on Retirement Income Data Needs and List of Respondents Samuel H. Williamson Miami University Robert J. Willis University of Michigan Doug A. Wolf Syracuse University Aliya Wong Thelen Reid & Priest, LLP Jing Jian Xiao University of Rhode Island Paul J. Yakoboski American Council of Life Insurers Sheila R. Zedlewski Urban Institute Stephen P. Zeldes Columbia University Page 64 GAO-03-337 Retirement Income Data Appendix IV: Views of GAO’s Expert Panel on Appendix IV: Views of GAO’s Expert Panel on Retirement Income Data Needs Retirement Income Data Needs This appendix provides the summary of discussion by members of an expert panel that we convened on retirement income data needs on September 10, 2002. The panel consisted of 11 nationally recognized experts who, during a day-long meeting, discussed the issues the federal government should address in order to improve the quality of retirement income data statistics. All the ideas presented in this appendix may not represent the view of every member of the panel. Moreover, these ideas should not be considered to be our views. Members of Our Expert Panel The following individuals were members of our expert panel on retirement income data: • Eric Engen, Resident Scholar, The American Enterprise Institute • Melissa Favreault, Research Associate, The Urban Institute • Nancy Gordon, Associate Director for Demographic Programs, US Census Bureau • Alan Gustman, Professor of Economics, Dartmouth College • Richard Hinz, Chief Economist and Director of Policy and Research, EBSA, Department of Labor (now at the World Bank) • Howard Iams, Director, Division of Policy Evaluation, Office of Policy, Social Security Administration • John Sabelhaus, Unit Chief, Long-Term Modeling Group, Congressional Budget Office • Dallas Salisbury, President, Employee Benefit Research Institute • Jack VanDerhei, Associate Professor, Department of Risk, Insurance & Healthcare Management, Fox School of Business and Management, Temple University • William Wiatrowski, Chief, Compensation Data Analysis and Planning Division, Bureau of Labor Statistics, Department of Labor • Robert Willis, Professor of Economics, Institute for Social Research, University of Michigan Views of the Panelists The panel members discussed a number of issues the federal government needs to address in order to improve the quality of retirement income statistics. Specifically, panelists discussed actions and strategies the federal government could undertake related to (1) improving matching of survey and administrative data, (2) improving access to administrative data, and (3) improving the quality of employer data. Page 65 GAO-03-337 Retirement Income Data Appendix IV: Views of GAO’s Expert Panel on Retirement Income Data Needs Need for Better Panelists said that a significant amount of the needed information about American workers and pension behavior is already being collected in Matched Data government and private surveys and government administrative reports. While no one single survey or report collects all the pension information panels expressed interest in, they said that some household surveys could be linked with administrative report data from employers. Specifically, panelists discussed the following: • The Survey of Income and Program Participation (SIPP) a household survey conducted by the Bureau of the Census, has information on demographic characteristics, labor force participation, and detailed information on income, including some basic pension information. The SIPP does not gather detailed information about the characteristics of these pensions.1 • Summary Plan Descriptions (SPD), prepared by employers as required under ERISA, provide detailed information about the characteristics of the pension plans that they provide to their employees. • In the past, the Health and Retirement Study (HRS), which is primarily a household survey conducted by the University of Michigan, has matched survey data from households who permitted the Social Security Administration to provide Social Security earnings records and benefit records, the National Death Index, and to Medicare records for those individuals who are Medicare eligible. HRS records also have been matched to SPDs obtained either from Labor or from employers. • The panelists were interested in the information available from the 1979 National Longitudinal Survey, which is sponsored by the Bureau of Labor Statistics and gathers a wide range of information over a long time period. Panelists felt that this information could be potentially be useful because the participants, who initially participated as youth, are now approaching retirement age, which would give researchers access to a lifetime of data. • In addition to matching currently being done, some panelists said that they were interested in matching existing pension information to other sources of employer information and public financial data, such as information from reports filed with the SEC and available, for example, from Compustat. Limitations on Access to Data Members of our expert panel felt that there are several sources of administrative data that could give researchers valuable information, especially those that could be linked, but legal and logistic restrictions and 1 While some employee information is gathered from other surveys, the panel felt that the SIPP survey is extremely important for matching to administrative records because it gives demographic and labor force characteristics not available in most administrative record files. Page 66 GAO-03-337 Retirement Income Data Appendix IV: Views of GAO’s Expert Panel on Retirement Income Data Needs limitations prevent access to some of this pension data. Specifically, panelists discussed different types of limitations to access. Legal Restrictions • The Federal Code Title 13, Section 9 sets very specific limitations on the access to any data identifying individuals gathered by the Census Bureau and any data that are linked to Census data has the Title 13 limitation “attached” to it as a result.2 • Individual records from the IRS have some access restrictions similar to those of the Census Bureau except that legislation allows specified agencies access to certain tax return records.3 • As a result of amendments to ERISA legislation SPDs are no longer routinely collected by Labor. Fragmented Retirement Data Responsibilities Contributes to Data Shortcomings • The responsibility for gathering and analyzing pension data is fragmented between different government agencies. There is no central agency responsible for retirement data – it is fragmented between the Pension Benefit Guaranty Corporation, Labor and IRS. As a result, individual agencies do not have the incentive to provide access to information or to collect statistics for research purposes. • While Labor’s regulations specify that employers must supply SPDs if requested, it does not specify that employers must have a copy of the SPDs. As a result, researchers who request SPDs from employers are frequently told that the Plan Administrator has the documents. The Plan Administrators in turn tell researchers that they have no authority for providing them to anyone except plan participants. • Since 1980, OMB’s Office of Information and Regulatory Affairs has provided government wide leadership and oversight of efforts to reduce the burden on respondents to government information collection, including statistical surveys”. This “reduction” effort has meant that some research data previously collected in administrative reports has been eliminated. Other Suggestions for Improving Access to Data The expert panelists made many suggestions for improving access to pension data. More specifically, panelists discussed the following: 2 The newly enacted E-Government Act of 2002 (P.L. 107-347) extended Census type confidentiality restrictions to all data collected in a federal statistical survey. 3 See 26 U.S.C. 6103. Page 67 GAO-03-337 Retirement Income Data Appendix IV: Views of GAO’s Expert Panel on Retirement Income Data Needs • It was suggested that agencies with access to data, such as Labor, take advantage of improvements in technology to require and provide electronic copies of Form 5500s and SPDs. • Panel experts suggested that some sort of license be issued for research vehicles (such as the SCF, the SIPP and the HRS) to have legal access to employer pension information, provided that they take measures to ensure confidentiality.4 • Panelists suggested creating more research data centers to match otherwise restricted data, including Census Bureau data. They suggested changing legislation so that the Census Bureau’s data would not make everything subject to Title 13. The Census Bureau, however, is concerned about possible reidentification issues.5 • Panelists suggested studying more closely the Census Bureau’s results in its experimentation with the development of “synthetic data” in the LEHD program, a technique in which many relationships between variables are maintained in a data set, but in a manner that makes it impossible to identify specific individuals.6 Panelists also cautioned, however, that in many cases these techniques are not workable and researchers will need access to the original survey data. Restricted Access Sites Have Provided Some Increased Opportunities for Access to Survey Records and Matched Files Experts discussed research data centers, operated by several Federal agencies and private survey organizations, as an effective technique to make data not available because of confidentiality restrictions more available to researchers, but with certain restrictions. While the data centers provide opportunities for conducting research with survey records or matching records between surveys or with administrative records, experts said that there are limitations to the data centers, which make data 4 Unlike the Census Bureau, some agencies, such as the National Center for Education Statistics, have authority to use a data license procedure. Under these licenses, licensees must submit detailed research plans, sign disclosure protection agreements, and agree to restrictions similar to those by the Bureau of the Census at their research data centers. For more detail, see Paul B. Massell and Laura Zayatz, “Data Licensing Agreements at U.S. Government agencies and Research Organizations,” presented at the International Conference on Establishment Surveys-II, Buffalo, N.Y., June 17-21, 2000. 5 Implementation of this suggestion would require changing not only Title 13 but other laws as well–like the new E-Government Act. 6 U.S. General Accounting Office, Record Linkage and Privacy: Issues in Creating New Federal Research and Statistical Information, GAO-01-126SP (Washington, D.C.: April 2001) p. 88 and Federal Committee on Statistical Methodology, Report On Statistical Disclosure Limitation Methodology, Statistical Policy Working Paper, 22 (Washington, D.C.: May 1994). Page 68 GAO-03-337 Retirement Income Data Appendix IV: Views of GAO’s Expert Panel on Retirement Income Data Needs much more difficult to access. Specifically, panelists discussed the following: • Restricted access sites are a useful means of allowing confidential information to be accessed by researchers, subject to certain restrictions. • A federal storage data center where a number of federal data sets could be brought together could allow agencies to share some otherwise inaccessible information. • Data centers could aid the work that researchers are doing by storing research already conducted in the data center. • The Census Bureau, which operates Research Data Centers in eight locations throughout the country, allows researchers with pre-approved projects to use confidential economic and demographic survey data, such as SIPP, for which either no public-use version is available, or the public-use version does not contain the information required by the researcher. While researchers can access these data at the centers, they are not allowed to remove individual records from the data center. This restriction prohibits researchers from matching Census Bureau data with data sets available to researchers, unless those data are imported into the data center. • BLS has a similar data center located in Washington, D.C. in which data extracted from SPDs collected as part of the National Compensation Survey are stored. As with the Census Bureau centers, there are limitations on who can access this information and restrictions on removing data from the data center. • Because confidential data cannot be removed from either the BLS or the Census Bureau data centers, it is currently not possible for researchers to match data sets from the two agencies. More Employer There was wide agreement on the panel that greater access to employer information was needed to accurately capture the value and provisions of Information Needed on employer-provided pensions. More specifically, panelists discussed the the Value and Provisions following: of Employer-Provided • Employee-provided information about pensions is not a viable source because Pensions individuals often do not have a good understanding of the value or characteristics of pensions. In addition, agencies expressed concern about the impact of additional questions on nonresponse, and it may be difficult to obtain OMB approval for adding questions to statistical surveys or to administrative reports. • Valuable information on the value and characteristics of employer pension information is already collected by the Department of Labor on the Form 5500s. • Through the LEHD program, the Census Bureau has been trying to link together employer information with data from surveys. • Agencies have concerns about maintaining the privacy and confidentiality of data. For example, there was concern that linking SIPP information about Page 69 GAO-03-337 Retirement Income Data Appendix IV: Views of GAO’s Expert Panel on Retirement Income Data Needs individuals with Form 5500 files or summary plan descriptions could facilitate the reidentification of individual data reported to the Census Bureau. As a result, those linked data would be available only within the data centers. Page 70 GAO-03-337 Retirement Income Data Appendix IV: Views of GAO’s Expert Panel on Retirement Income Data Needs Page 71 GAO-03-337 Retirement Income Data Appendix V: Characteristics of Selected Surveys Appendix V: Characteristics of Selected Surveysfor Analysis of Retirement Income and Wealth for Analysis of Retirement Income and Wealth Below is a table highlighting some of the features of selected surveys pertinent to the analysis of retirement income and projection of future retirees’ income.1 Table 3: Summary Table of Selected Survey Data Sources Asset and Health Dynamics Among the Consumer Current Panel Study of Survey of Oldest Old and Health Expenditure Population Employee Benefits Income Consumer a and Retirement Study Survey Survey Survey Dynamics Finances Unit of AHEAD: Men and women Consumer units/ Households Employer Head of Household/ observation aged 70 and above in household household primary 1993-94 and aged 69 –75 members economic in 1998-99 and spouses unit HRS: Men and women aged 51-61 in 1992 and aged 51 to 56 in 1998-99 a and spouses Cross Longitudinal Cross-sectional Cross-sectional Cross-sectional Longitudinal Cross- sectional or sectional longitudinal Cohorts or Began in 1992. Planned Civilian non- Civilian non- In even years state & All All years to include a new cohort of institutionalized institutionalized local governments & b covered people aged 51 to 56 population population beginning in 1999, every 5 years private employers surveyed annually Panel period AHEAD: 1993-present Current ongoing 1942 to present 1997 to present 1968 - present 1983-present HRS: 1992-present survey: 1980- present 1 AHEAD, HRS, and PSID are conducted by university research entities with support from federal agencies. The other surveys are conducted by federal government agencies. Page 72 GAO-03-337 Retirement Income Data Appendix V: Characteristics of Selected Surveys for Analysis of Retirement Income and Wealth National National Survey of Income National Longitudinal Longitudinal National Longitudinal National and Program Survey of Mature Survey of Older Longitudinal Survey Survey of Young Longitudinal Survey Participation Women Men of Young Women Men of Youth Household Women aged 30-44 in Men aged 45-60 in Women aged 14-24 Men aged 14-24 in Men and women members 1967 and spouses 1966 and spouses in 1968 and spouses 1966 and spouses aged 14-21 in 1979 and spouses Longitudinal Longitudinal Longitudinal Longitudinal Longitudinal Longitudinal All Women born 1922 to Men born 1906 to Women born 1943 to Men born 1941 to Men & women born 1937 1921 1953 1952 1957 to 1964 1981 - present 1967 to present 1966 to 1990 1968 to present 1966 to 1981 1979 to present New panels approx. every 3 years Page 73 GAO-03-337 Retirement Income Data Appendix V: Characteristics of Selected Surveys for Analysis of Retirement Income and Wealth Asset and Health Dynamics Among the Consumer Current Panel Study of Survey of Oldest Old and Health Expenditure Population Employee Benefits Income Consumer and Retirement Study Survey Survey Survey Dynamics Finances Frequency of Biennial The interview Monthly for a total Once in a survey Annual/ Triennial interviews survey has a of 8 months; period biennial rotating panel households are design; interviewed 4 consumer units months in a row, are interviewed rest for 8 months, every 3 months and are for 5 periods. interviewed for The diary another 4 months survey also has (4-8-4 sample a rotating panel rotation) design - consumer units keep weekly diaries. Over samples Blacks, Hispanics, and No Hispanics in the N/A Low income Wealthy Florida residents March data households collection, (includes the March supplement.) Most recent 22,000 individuals Interview: 55,000 4,500 establishments 7,406 families 4,500 sample size: 30,600 households Diary: 15,300 interviewed Administrative Medicare files, Nat’l Death None SSA and IRS N/A None None c record Index, SSA earnings and (restricted) linkages projected benefits files, Employer Pension Plan SPDs Page 74 GAO-03-337 Retirement Income Data Appendix V: Characteristics of Selected Surveys for Analysis of Retirement Income and Wealth National National Survey of Income National Longitudinal Longitudinal National Longitudinal National and Program Survey of Mature Survey of Older Longitudinal Survey Survey of Young Longitudinal Survey Participation Women Men of Young Women Men of Youth Every 4 Every 1 to 3 years Every 1 to 2 years Every 1 to 3 years Every 1 to 2 years Every year until 1994 months until 1983; then until 1981 then every 2 years 1990 beginning in 1996 Low-income Blacks Blacks Blacks Blacks Blacks, Hispanics, households for 1990, and economically and 1996 panels, and disadvantaged thereafter 35,000 households 5,000 5,000 5,000 5,000 9,000 interviewed d SSA & IRS Employer pension plan None None None None (restricted) SPDs for 1989 Source: GAO analysis of data from University of Michigan’s Institute for Social Research, the Census Bureau, the Bureau of Labor Statistics, and the Federal Reserve Board a The Employee Benefits Survey is now a component of the National Compensation Survey (NCS). It excludes federal, military, agricultural, fishing, forestry, or private household employers. b The total civilian non institutional population of the United States, as well as that portion of the institutional population living in the following group quarters are included in the CEX: Boarding houses; housing facilities for students and workers; staff units in hospitals and homes for the aged, infirm, or needy; permanent living quarters in hotels and motels; and mobile home parks. Excluded are military personnel living on military bases and nursing home residents. c Restricted refers to those persons at the SSA, BLS, and Census Bureau with authority to access these data. d Links with Social Security earnings records and extracts from individuals’ IRS tax returns. Page 75 GAO-03-337 Retirement Income Data Appendix VI: Comments from the Department of Appendix VI: Comments from the Department of Commerce Commerce Page 76 GAO-03-337 Retirement Income Data Appendix VII: Comments from the Department of Appendix VII: Comments from the Department of Labor Labor See p. 23. Page 77 GAO-03-337 Retirement Income Data Appendix VII: Comments from the Department of Labor See p. 23. Page 78 GAO-03-337 Retirement Income Data Appendix VII: Comments from the Department of Labor Page 79 GAO-03-337 Retirement Income Data Appendix VII: Comments from the Department of Labor See p. 23-24. Page 80 GAO-03-337 Retirement Income Data Appendix VIII: GAO Contacts and Staff Appendix VIII: GAO Contacts and Staff Acknowledgments Acknowledgments GAO Contacts Alicia Puente Cackley (202) 512-7022 Benjamin Pfeiffer (206) 287-4832 Staff Acknowledgments Timothy Fairbanks, Nicholas Larson, Lynn Musser, Emily Pickrell, and Roger Thomas also contributed to this report. (130117) Page 81 GAO-03-337 Retirement Income Data The General Accounting Office, the audit, evaluation and investigative arm of Congress, GAO’s Mission exists to support Congress in meeting its constitutional responsibilities and to help improve the performance and accountability of the federal government for the American people. GAO examines the use of public funds; evaluates federal programs and policies; and provides analyses, recommendations, and other assistance to help Congress make informed oversight, policy, and funding decisions. GAO’s commitment to good government is reflected in its core values of accountability, integrity, and reliability. The fastest and easiest way to obtain copies of GAO documents at no cost is through the Obtaining Copies of Internet. GAO’s Web site (www.gao.gov) contains abstracts and full-text files of current reports and testimony and an expanding archive of older products. The Web site features a GAO Reports and search engine to help you locate documents using key words and phrases. You can print Testimony these documents in their entirety, including charts and other graphics. Each day, GAO issues a list of newly released reports, testimony, and correspondence. GAO posts this list, known as “Today’s Reports,” on its Web site daily. The list contains links to the full-text document files. To have GAO e-mail this list to you every afternoon, go to www.gao.gov and select “Subscribe to daily E-mail alert for newly released products” under the GAO Reports heading. Order by Mail or Phone The first copy of each printed report is free. Additional copies are $2 each. A check or money order should be made out to the Superintendent of Documents. GAO also accepts VISA and Mastercard. Orders for 100 or more copies mailed to a single address are discounted 25 percent. Orders should be sent to: U.S. General Accounting Office 441 G Street NW, Room LM Washington, D.C. 20548 To order by Phone: Voice: (202) 512-6000 TDD: (202) 512-2537 Fax: (202) 512-6061 Contact: To Report Fraud, Waste, Web site: www.gao.gov/fraudnet/fraudnet.htm and Abuse in Federal E-mail: email@example.com Automated answering system: (800) 424-5454 or (202) 512-7470 Programs Jeff Nelligan, managing director, NelliganJ@gao.gov (202) 512-4800 Public Affairs U.S. General Accounting Office, 441 G Street NW, Room 7149 Washington, D.C. 20548
Retirement Income Data: Improvements Could Better Support Analysis of Future Retirees' Prospects
Published by the Government Accountability Office on 2003-03-21.
Below is a raw (and likely hideous) rendition of the original report. (PDF)