|
Items in cart [0] |
Questions? Call 888-624-8373 |
| Copyright © 2009. National Academy of Sciences. All rights reserved. Terms of Use and Privacy Statement |
Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter.
Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 165
6
Furthering Coordination for Data
Collection, Research, and Modeling
There are many pieces to the retirement-income-security policy puzzle, and many
research, data collection, policy, and program agencies are involved in one or
more aspects of the topic. There are strengths to having a variety of perspectives;
however, we believe that mechanisms for coordinating agency efforts and setting
priorities within a broader perspective are needed.
There is also a need for greater involvement of the private sector and aca-
demic research community in retirement-income-related data collection, analy-
sis, and modeling. Such involvement expands the community of people who can
contribute to improved databases and behavioral analyses that are most relevant
for policy and to the development and evaluation of projection models that appro-
priately use data and research results.
ORGANIZATIONAL ISSUES
As noted in Chapter 1, responsibility for the major policies that affect retirement
income security, such as employer pension regulation, Social Security, tax policy,
and health care programs, is spread among several agencies in different depart-
ments. The responsibility for relevant data collection and research is spread
among several policy, statistical, and research agencies as well. There are ben-
efits of having many players in this complex area. Different agencies can de-
velop expertise in particular aspects of data collection, analysis, and policy for-
mulation and can bring perspectives to the table that might otherwise be
overlooked. Also, it would not make sense to combine such functions as private
pension regulation with support of basic behavioral research or broad-based data
165
OCR for page 166
166
ASSESSING POLICIES FOR RETIREMENT INCOME
collection in the same agency the concerns and skills required to be effective in
these different capacities are very different.
At the same time, there are clearly drawbacks to having the responsibilities
for retirement-income-related policies, data, and research dispersed across so
many agencies, each with its own perspective. For data on individuals, for
example, existing panel surveys, which are sponsored by different agencies with
different agendas, do not complement each other as well as they might. Even
more problematic are data on employers. As just one example, the U.S. Depart-
ment of Health and Human Services (HHS) is undertaking a large new survey of
employer health insurance plans and costs, without reference to the need for data
on pensions and other benefits. Such a course makes sense from the department' s
orientation toward health policy issues, but it does not necessarily make sense
from the government' s broader interest with issues of retirement income security
policy.
More generally, having a large number of agencies involved in retirement-
income-related data collection and analysis increases the likelihood that there
will be both unnecessary overlap and duplication of effort and significant gaps in
needed information. We conclude that more attention should be given to inter-
agency coordination to improve the quality and utility of data and research knowl-
edge, to identify priority areas in which more work is needed, and, conversely, to
identify areas in which overlap could be reduced. Highly constrained budgets for
data and research make it all the more urgent to achieve effective coordination of
agencies' agendas and plans.
The challenge is to devise appropriate coordination mechanisms that facili-
tate making hard choices among agencies' priorities, but do not stifle initiative
and innovation or introduce bureaucratic layers of review and approval. At-
tempts to develop a "unified" budget for retirement-income-related data collec-
tion or research, for example, would undoubtedly require a very heavy hand from
the U.S. Office of Management and Budget (OMB) and be largely futile, in any
case, in part because the budget authority for the various agencies that would be
affected is split among a number of congressional committees. In contrast,
interagency committees and forums, which are often very effective as a means of
sharing information, are typically much less effective as a means by which to set
priorities or to induce agencies to alter their agendas in any significant manner.
Indeed, a former official in the Department of Education once said that "coordi-
nation in the government is an unnatural act." The characteristics of existing
interagency groups that are or could be relevant to the retirement area only
underscore this point.
One such group is the Federal Committee on Statistical Methodology. It has
been very hard-working throughout its 20-year existence and has a string of
outstanding publications to its credit, several of which treat topics that are impor-
tant for improving retirement-income-related data (see, e.g., Federal Committee
on Statistical Methodology, 1988, 1990~. The committee had a strong chair, the
OCR for page 167
FURTHERING COORDINATION
167
late Maria Gonzalez of the Statistical Policy Division of OMB, and has been able
to enlist substantial contributions from staff of many member agencies. But this
committee has not sought to determine data gaps and overlaps or set priorities; its
major tasks have been to review data quality issues and describe best statistical
practice.
Another such group is the Interagency Forum on Aging-Related Statistics,
established in 1986 and led by the Census Bureau, the National Center for Health
Statistics (NCHS), and the National Institute on Aging (NIA). It has served as a
forum for information exchange in twice-yearly meetings and has tried at times to
go beyond that function. Its working groups have had some success in such areas
as identifying core sets of data items that aging-related surveys should include.
However, the forum has found it difficult to sustain efforts that could affect
agencies' survey plans.
Recently, a combination of circumstances led agencies within HHS to de-
velop an integrated design for health-related data collection. The circumstances
included the glaring weaknesses in relevant data and research knowledge that
became apparent in the 1993-1994 health care reform debate, constrained bud-
gets, and the pressures for departmental restructuring and agency cooperation
from the "reinventing government" initiative (Hunter and Arnett, 1996~. The
result has been a plan that integrates several surveys, which were previously
independently conducted by different agencies, in a way that will, it is hoped,
serve a variety of health status and health care policy analysis needs in a cost-
effective manner. HHS has also established a senior-level Data Council, with
overall responsibility for addressing data issues across the department. The
challenge for work on retirement is that coordination mechanisms for surveys and
administrative record systems must involve agencies both within and across de-
partments.
COORDINATION MECHANISMS
Data Collection
We believe there are opportunities to work toward integration of retirement-
income-related data collection for employers and individuals. What will be
required is a decision to avoid the example of health care reform, in which
embarrassing deficiencies led to integration after the fact, and, instead, try to
make priority investments in anticipation of policy needs.
Employer Data Task Force
We recommend that the U.S. Department of Labor (DOL) establish a task force
on retirement-income-related employer data that is led by the Bureau of Labor
Statistics (BLS) and the Pension and Welfare Benefits Administration (PWBA)
OCR for page 168
168
ASSESSING POLICIES FOR RETIREMENT INCOME
and involves other relevant agencies (see Chapter 4~. We urge the task force to
obtain input from benefit consultants, representatives of different types of em-
ployers, and academic researchers.
A priority for the task force should be to review the major government
employer data collection systems and to analyze the implications of alternative
ways of building on their strengths in terms of both costs and benefits. For
example, what are the prospects and what would be the costs of making the DOL/
PWBA Form 5500 data series more useful for research purposes and linking it to
the BLS Employee Benefits Survey (EBS) (and perhaps the Employment Cost
Trends Survey as well)? Alternatively, is it possible to expand the HHS National
Employer Health Insurance Survey to cover benefits more widely and at what
cost? If such an expansion is possible, would it be possible or desirable to reduce
the scope of the EBS or Form 5500 data series and with what potential cost
savings?
The task force should look for other ways to leverage existing data series and
to identify unnecessary overlap so funds would be available for needed new data
collection. For example, what are the possibilities of incorporating selected
private data sets into an integrated employer database? Should the periodic
pension coverage supplement in the Current Population Survey (CPS) be elimi-
nated in favor of continuing and improving the annual pension information col-
lected in the Survey of Income and Program Participation (SIPP)? In fact, PWBA
is now actively exploring this last possibility with the Census Bureau. There is
evidence that the SIPP pension data are of equal quality (lams, 1995), and SIPP
has the advantage of obtaining considerably more information than the CPS that
would be useful to analyze together with pension data.
Finally, the task force should consider the costs and benefits of alternative
models for obtaining panel data on employers and their workers. One model is to
build on the Health and Retirement Survey (FIRS). Another model is to launch a
brand-new survey.
While the task force will not be able to compel agencies to alter their agen-
das, its cost-benefit analyses, if well done, would provide considerable insight
into the advantages and disadvantages of alternative plans. Its analyses could
facilitate collaborative decisions by agencies to streamline existing data systems
where feasible and thereby make resources available for important unmet data
needs.
Panel Survey Review Group
We recommend that NIA facilitate collaborative efforts among sponsors and
principal investigators to regularly review the questionnaire content and data
collection practices of the major retirement-income-related panel surveys of indi-
viduals: HRS and AHEAD (Health and Retirement Survey and Asset and Health
Dynamics Among the Oldest Old survey), the various cohorts in the National
OCR for page 169
FURTHERING COORDINATION
169
Longitudinal Surveys (NLS), and the Panel Study of Income Dynamics (PSID).
The Survey of Consumer Finances (SCF) and SIPP should also be included in
these reviews (see Chapter 4~. NIA is well suited for this role, given its focus on
aging research, which includes the economic well-being of the elderly, and its
sponsorship of HAS/AHEAD.
There appear to be significant opportunities to improve the overall utility and
quality of the data from panel surveys of individuals for retirement-income-
related behavioral analysis. As just one example, adding wealth and pension
modules to the National Longitudinal Survey of Youth (NLSY) would make it
possible to determine why there are such great disparities in savings and wealth
by the time people reach their 50s (the age cohort of FIRS). Because the NLSY
cohort is now well past the age when most schooling is obtained, it is possible to
make room for retirement-income-related modules by scaling back previously
important modules on education and training.
We do not make detailed recommendations about what could or should be
done to each survey (see our general recommendations in Chapter 4~; rather we
stress the importance of periodically bringing together the sponsor agencies and
principal investigators to conduct broad oversight reviews. These reviews should
seek to identify the best data collection practices that should be universally
adopted, to consider core question content that would facilitate cross-survey and
pooled data analyses, and to consider modules to add and delete that could make
the surveys more useful for analysis of retirement income security while not
compromising their other goals.
Forum Working Group on Data Quality
It would be helpful for the employer data task force and the panel survey review
group to share their analyses and conclusions periodically with the Interagency
Forum on Aging-Related Statistics. In addition, we suggest that the forum,
perhaps working with the Federal Committee on Statistical Methodology, estab-
lish a working group to conduct important cross-survey studies of the quality of
retirement-income-related data. We expect that each survey or administrative
records data system regularly conducts its own quality reviews; however, we
believe it would be useful for an interagency working group to undertake and
publish cross-cutting studies on key issues (see the list in Chapter 4 for ex-
amples). The working group could also review completed validation studies,
whether performed by the group or by individual data system sponsors, to iden-
tify priority areas for data quality improvement.
Projection Modeling
We suggest that analysts who are working on special-purpose projection models
for retirement-income-related policy analysis in different agencies get together
OCR for page 170
170
ASSESSING POLICIES FOR RETIREMENT INCOME
on a regular basis to share information and learn from one another (see Chapter
5~. Meetings of the group could be the source of ideas for the design and,
ultimately, implementation of new employer-based models and a new individual-
level microsimulation model, after new data and research results become avail-
able. Members of the group should have input to the employer data collection
task force and the panel survey review group regarding the content of specific
surveys, such as HAS/AHEAD, to ensure that policy-relevant variables are in-
cluded in formats that are tractable for new projection models.
Such a users' group should include academic and private sector analysts with
relevant interest and experience in projection modeling, as well as agency staff.
In general, it is important for agencies to involve outside researchers and private
sector representatives in retirement-income-related data collection and modeling.
Actively reaching out to the private sector and academia can have benefits rang-
ing from a better climate for collection of relevant employer data to a larger
community of researchers who are actively engaged in the development and use
of policy-relevant behavioral and projection models.
Recommendation
19. Relevant agencies should establish coordination mechanisms to help
improve the quality and utility of retirement-income-related data, reduce
unnecessary duplication of effort, and identify priorities. Coordination
mechanisms should recognize the need for flexibility and experimentation
and not impose added bureaucratic requirements. Responsibilities for re-
tirement-income-related policy analysis, research, and data collection are
widely dispersed across government agencies, yet much work in this area
requires an integrated, cross-agency perspective, as well as the involvement
of academic researchers and private sector representatives.
INVOLVING THE PRIVATE SECTOR AND ACADEMIA
Working out ways to increase the involvement of academic and private sector
analysts in retirement-income-related data collection and projection modeling, in
addition to relevant behavioral research, is important for obtaining a full return
on agencies' limited investment dollars. There are precedents for such involve-
ment in the area of data collection but not, to date, in the area of development of
projection models.
Data Collection
HRS and AHEAD offer a model of successful involvement of the research com-
munity in generating databases that appear to have extraordinary analytical prom-
ise for explaining key retirement-income-related behaviors and contributing to
the development of a new generation of projection modeling tools. The start-up
OCR for page 171
FURTHERING COORDINATION
171
phase of HRS and AHEAD obtained input on design and content issues from
literally hundreds of researchers in several disciplines. Interdisciplinary steering
committees of researchers have continued to guide the development of the two
surveys and have maintained a strong focus on data needs for behavioral analysis
and experimentation with new questions and procedures to improve policy rel-
evance and data quality. Federal agencies have had input to HAS/AHEAD
through an interagency review group coordinated by NIA.
We urge that this approach be extended to employer surveys. In this area, it
is particularly important to involve private sector representatives as well as aca-
demic researchers in issues of survey design, content, and data collection proce-
dures. Success in obtaining needed employer data is unlikely if employers them-
selves are not consulted about feasibility issues and if they are not made aware of
the value of improved data, not only for government planning and policy analy-
sis, but also for use by the private sector. As key participants in the U.S. system
of pension, health care, and disability benefits, employers have a stake in having
access to the best data and analytical models with which to make decisions about
benefit offerings and about personnel and compensation practices more broadly.
Projection Modeling
In the area of projection modeling, there are as yet no success stories of effective
collaboration of agencies and their contractors with the academic community
more broadly. Historically, the research community has been largely disassoci-
ated from the development and application of projection models for policy analy-
sis (as distinct from analytical models to study behavior). Reasons for this
disassociation include the complexities of many models (e.g., microsimulation
models have been viewed as "black boxes") and limited access to them. Also,
academic research incentives emphasize new findings and theoretical applica-
tions rather than estimations of the costs and distributional implications of pro-
posed policy changes (see discussion in Citro and Hanushek, 1991:281-289~.
We stress that the federal government should give priority to retirement-
income-related data collection and research, not to investment in large-scale
projection models. At the same time, we support some continuing discussion of
the requirements for a new generation of employer-based and individual-level
microsimulation projection models (see Chapter 5~. We suggest that modelers
provide input to the task force on employer data and the panel survey review
group and that a projection modeling users' group involve interested academics
and private sector analysts. In this way, a dialogue could begin on such issues as
the feasibility of integrating behavioral models of particular types and with par-
ticular input variables into projection models.
When investments in data collection and associated research begin to bear
fruit so that it makes sense to actively plan for the development of a new genera-
tion of projection models, it will be important to continue a dialogue between the
OCR for page 172
72
ASSESSING POLICIES FOR RETIREMENT INCOME
model builders and academic researchers. Complete integration of the work of
researchers and projection modelers is not likely a feasible goal: it seems unreal-
istic to expect that researchers working in different areas could all develop look-
alike, "plug-compatible," analytical models for use as modules in a projection
model. Research will inevitably proceed in different directions with differing
implications for the adaptation of research results to projection model needs.
However, it should be possible to involve researchers from key analytical fields-
labor supply, savings and consumption, employer behavior more closely in the
development of useful projection models than has been the case in the past.
When it becomes appropriate to establish steering groups for new models, such
groups should include researchers as well as agency and contractor staff, similar
to the approach of HAS/AHEAD. The use of discussion groups and World Wide
Web sites on the Internet can provide a low-cost way to maintain an active
dialogue among steering group members. Periodically, they could use the Internet
to obtain critical reviews from other researchers about the design elements for
new projection models (e.g., the functional form and input variables for particular
behavioral components).
From participation in such a dialogue may well come increased interest on
the part of the research community in projection modeling, including the key
issue of model validation, and an enhanced capability of researchers to make
effective contributions in this area. Ultimately, the combined efforts of people
working in the public, private, and academic sectors will be needed to develop
improved projection models built on improvements in data and research knowl-
edge that can support more informed debate about the effects of alternative
policies on the retirement income security of current and future generations of
Americans.
Recommendation
20. Collaboration between government agencies and the research com-
munity and private sector should be developed to spur improvements in
retirement-income-related data and research knowledge that can support
the development of improved projection models. The model for such col-
laboration is the involvement of large numbers of researchers from several
disciplines in the design and content of HRS and AHEAD, which has had
high payoff for the relevance and quality of the data. For employer data
collection systems, this model should also involve private sector representa-
tives as well as academic researchers in design and content issues.
Representative terms from entire chapter:
employer data
