11
Research and Data Needs for Understanding and Ameliorating Intergenerational Poverty
Based on its review of the literature, the committee outlined the scope of intergenerational poverty in the United States. It found that many children growing up in low-income households experience some degree of upward mobility. At least two-thirds are not living in low-income households in adulthood, and some enjoyed standards of living well above any definition of poverty. But one in three children are still low-income when they become adults, and the chance of having low income in adulthood is much higher for a child raised in a low-income than more affluent household.
Regardless of the data source or the definition of poverty, the committee found that racial/ethnic disparities are an enduring feature of the intergenerational trajectories of children, with Black and Native American children experiencing much less upward mobility than White children growing up in the same economic circumstances. The size and consistency of these gaps underscore the importance of understanding the causes of racial/ethnic disparities (Chapter 3), as well as developing and implementing large-scale, effective policies and programs to ameliorate intergenerational poverty.
The committee reviewed research on potentially important drivers of intergenerational poverty in the following domains: children’s education and the educational system; child health and the health care system; family income and wealth and parental earnings and employment; family structure; housing, residential mobility, and neighborhood conditions; neighborhood safety and the criminal justice system; and child maltreatment and the child welfare system (see Chapters 4–10 for the committee’s assessment of the evidence and Appendix C: Chapter 11 for the conclusions regarding
these drivers). It found that abundant correlational evidence established the potential importance of each of these domains for perpetuating or alleviating intergenerational poverty persistence. But definitive causal evidence quantifying the relative importance of each of these domains was often lacking, which pointed to a number of steps that could strengthen the evidence base.
The committee’s primary assignment was to identify evidence-based policies and programs directed at children living in poverty today that would reduce those children’s chances of being poor when they become adults. In operationalizing the definition of “evidence-based,” the committee decided to highlight incremental policy and program ideas whose effectiveness is supported by direct intergenerational evidence—studies that offered rigorous long-run causal evidence (see Table 11-1 for the committee’s list of policy and program ideas that met this standard of evidence).
The committee found a lack of high-quality evidence on the intergenerational impacts of many other promising programs. It emphasizes that this should not be taken to mean that most of those other programs are ineffective—only that their intergenerational impacts have not been assessed. This fact is sobering but not surprising, given the expense and difficulty of scaling up promising interventions identified in controlled experiments, the length of time required to see the effects of interventions on intergenerational poverty, the difficulties of assembling data for historical, retrospective analysis, and the costs of obtaining an adequate sample size for populations most at risk of intergenerational poverty, especially Native Americans.
Another noteworthy consequence of the committee’s high standard of evidence is that because studies that met it focused on individual policies, the committee was unable to identify evidence-based combinations of federal policy investments that could reduce intergenerational poverty persistence. The report notes a number of instances where combining programs might generate more benefits than the individual programs taken by themselves, but here too the evidence does not support confident conclusions.
In this chapter, the committee offers general principles to guide private and public funding organizations in supporting needed research on intergenerational poverty. It then addresses the need for a federal infrastructure of census, survey, and administrative records data that can be linked, going forward and backward in time, and accessed for research and policy analysis purposes. These purposes would include analyzing trends in intergenerational poverty, overall and for population subgroups, and estimating the likely effects of policies and programs intended to foster intergenerational mobility for all children. As with all research data, it is vital to ensure that administrative and survey data be made available to the research and policy evaluation community in ways that respect respondents and protect the confidentiality of their data. At present, substantial barriers impede access
| Driver | Program or policy idea (* indicates that the supporting evidence was particularly strong) |
|---|---|
| Education | |
|
None identified in recent research |
|
Increase K-12 school spending in the poorest districts* Increase teacher workforce diversity* Reduce exclusionary school discipline* Increase access to Ethnic Studies courses |
|
Expand effective financial aid programs for low-income students* Increase campus supports (such as tutoring and case management)* |
|
Expand high-quality career and technical education programs in high school* Expand sectoral training programs for adults and youth* |
| Child and Maternal Health | |
|
Increase funding for Title X family planning programs* Ensure that Medicaid beneficiaries have access to family planning services* |
|
Expand access to Medicaid with continuous 12-month eligibility and 12-month post-partum coverage* Expand access to Indian Health Services for all eligible mothers and children |
|
Support the EPA to work with local partners to adopt and expand efficient methods of monitoring outdoor and—especially in schools—indoor air quality |
|
Remove the 5-year waiting period of Supplemental Nutrition Assistance Program (SNAP) eligibility for legal permanent-resident parents* Eliminate the proration of SNAP benefits for citizen children with undocumented parents |
| Family Income, Wealth, and Employment | |
|
Expand the Earned Income Tax Credit by increasing payments along some or all portions of the schedule and possibly by providing a credit to families with no earnings* |
| Family Structure | |
|
None identified by research to date |
|
| Driver | Program or policy idea (* indicates that the supporting evidence was particularly strong) |
|---|---|
| Housing and Neighborhoods | |
|
Expand coverage of the Housing Choice Voucher program and couple it with customized counseling and case management services to facilitate moves to low-poverty neighborhoods |
| Neighborhood Crime and the Criminal Justice System | |
|
Use juvenile confinement only for youth who pose a serious and immediate threat to public safety* |
|
Improve school quality and reduce lead exposure in ways identified in the education and health categories* Scale up evidence-based therapeutic interventions such as the Becoming a Man program |
|
Scale up programs that abate vacant lots and abandoned homes* Increase grants to community-based organizations* |
|
Expand funding for policing in high-crime neighborhoods* Expand use of effective strategies like community policing* |
|
Improve gun safety in ways that pass constitutional review* Promote child access prevention laws and restrictions on right-to-carry laws, limit access to guns by domestic abusers* Promote sentencing add-ons for violence involving firearms* |
| Child Maltreatment | |
|
None identified by research to date |
|
| Racial Disparities | |
|
A number of the policies and programs listed above have been shown to be effective for Black children and families (See Table C-3-1)* |
|
NOTES: “*” indicates that the program’s or policy’s impact on intergenerational poverty is supported by random-assignment evaluation evidence that has been replicated across several sites or by compelling quasi-experimental evidence based on national or multi-state data or a scaled-up program. Table entries without an “*” represent programs or policies for which the evidence has not been replicated or the policy has not been scaled up.
to federal data for evidence building use. The chapter ends with the committee’s conclusions and recommendations to meet research and data needs.
PRIORITIES FOR FUTURE RESEARCH
This report has prioritized drawing lessons from strong causal research on drivers of intergenerational poverty and on policies and programs that might reduce it. The most obvious examples of strong causal research designs are studies modeled after clinical trials, which either randomly assign children or families to a policy or program or take advantage of some kind of lottery process that randomly offers opportunities to participate in the program or policy. The Head Start Impact Study is an example of explicit random assignment (Puma et al., 2012). However, most clinical trial–type studies are almost never conducted or followed up at the scale needed to establish that an intervention will be effective in a variety of settings and for all population groups.
Most of the available causal research on intergenerational poverty is “quasi-experimental,” examining in retrospect the consequences for children of naturally occurring policy changes that were rolled out over time and across well-defined geographic areas. When matched to long-run administrative or survey data, these quasi-experimental data enable researchers to compare the longer-run outcomes of children to identify those who benefited from the program. A prominent example is Hoynes et al. (2016), which took advantage of the slow county-by-county roll-out of the Food Stamp Program in the 1960s and 1970s. Using longitudinal survey data from the Panel Study of Income Dynamics, the researchers found that the adult cardiovascular health of children born in counties already offering the Food Stamp Program was much better relative to children who were born at the same time but grew up in counties that did not offer food stamps until they had entered school.
Bailey et al. (2020) also analyzed the county-level roll-out of the Head Start Program, comparing long-run outcomes for children residing in Head Start counties and those in other counties using census and Social Security data. Aizer and Doyle (2015) took advantage of the fact that court cases are randomly assigned to judges who differ in their tendency to detain or incarcerate youth defendants. They matched defendants to administrative data on their adult labor-market and criminal behavior to study the impacts of juvenile detention and incarceration on adult crime and labor market success.
These examples illustrate the importance of facilitating several kinds of research by providing the necessary data—these examples illustrate the importance of facilitating policy research by supporting long-run followups of random-assignment studies as well as providing administrative and
survey data that can be used to study the impacts of policies and programs that are implemented in ways that create “natural experiments” for policy researchers. Supporting more of these kinds of exemplary studies and expanding their capacity to inform policy, however, poses a variety of challenges for researchers and their funders:
- The costs and difficulties of scaling up promising interventions identified in controlled experiments;
- The length of time required to see the effects of interventions on future intergenerational poverty, which means that researchers could have to wait 20 or more years for results;
- The difficulty of anticipating what policies and interventions will be of interest decades in the future;
- The costs of adequate sample size for population groups of interest (e.g., most surveys and evaluation studies are not able to support reliable analyses for Native Americans); and
- The difficulties of and barriers to accessing and linking the most useful data, which often come from federal, state, and local administrative records, for evaluating the impacts of past program and policy changes.
A central goal for research policy should be to reduce barriers to developing better evidence about drivers and policies impacting intergenerational poverty. This chapter discusses several ways to do this. To help target resources for research that could lead to effective policies, we list in Box 11-1 important research priorities for each of the domains we cover.
RESEARCH FUNDING PRINCIPLES AND GUIDANCE
Principles
In an effort to identify proven programs for boosting every child’s chance to succeed, the committee proposes three broad principles for research funding. These principles apply both to evaluations of the effectiveness of past policy and program changes and to prospective research—that is, research where previous experimentation has not been done or has failed to identify promising programs and where retrospective research is impossible.
The committee also hopes that funding organizations, public and private, will support the construction and use of linked administrative and survey data that could support more extensive retrospective and prospective research (see next section). Moreover, the committee hopes that funding organizations will work together to provide the level of support that is
necessary for effective research and policy assessment aimed at reducing intergenerational poverty, particularly among high-risk population groups.
The committee’s research funding principles apply to evaluations of the effectiveness of past policy and program changes (Principle 1); prospective research (Principle 2); and focusing on populations in need (Principle 3).
- Principle 1: Prioritize strong research designs that provide causal estimates of program impacts. This report has prioritized drawing lessons from strong causal research on drivers of intergenerational poverty and on policies and programs that might reduce it. This is because methodological research has repeatedly shown that correlational techniques such as propensity score matching can often (but not always) provide badly biased estimates of program impacts (Cook et al., 2008; LaLonde, 1986).1
- Principle 2: For prospective research, set aside funding for rigorous, small-scale experiments, but also for replications, scaled-up evaluations, and long-term follow-ups of promising programs. Research portfolios on ameliorating intergenerational poverty in one or more domains should, as a matter of course, contain three broad funding components. The first would fund careful, smaller-scale, and shorter-term experimental evaluations of potentially promising program models. The second component would provide funding to replicate those programs that show promise in the experimental phase across diverse populations and at scale. The third component would fund the investigation of longer-run impacts of as many programs as possible that have shown promise in the experimental and replication arms of the research, using a combination of administrative data and surveys for follow-up purposes. So that outcomes can subsequently be tracked using administrative and survey data, researchers conducting shorter-term evaluations should secure and maintain the necessary permissions enabling them to track participants using administrative or survey data later on. In all these steps, protecting the identity of research subjects is of the utmost importance.
- Principle 3: Fund research arms for specific communities. Research portfolios should focus on population groups and communities at highest risk of intergenerational poverty, not only to target scarce research dollars as effectively as possible, but also to help people
___________________
1 Heckman et al. (1999) established some conditions under which nonexperimental evaluations of job training programs can come close to generating similar results to experimental evaluations. However, in many other kinds of evaluations, no such conditions have been established.
BOX 11-1
Research Priorities to Ameliorate Intergenerational Poverty and Facilitate Socioeconomic Mobility, by Domain
Racial Disparities
- Conduct research to quantify structural racism in such domains as neighborhood, workplace, health care, and others, using standardized measures.
- Conduct quasi-experimental policy and intervention studies that prioritize data and research designs that provide separate impact estimates by race and ethnicity.
- Develop and expand interventions and evaluations aimed at decreasing intergenerational poverty among Black and Native American people, while prioritizing community control and input.
Education
- Carry out longer-run follow-ups of promising interventions, such as the highest-quality state pre-K programs and high-quality tutoring programs.
- Examine home-visiting programs to determine which ones best promote short- and longer-term developmental outcomes for different groups children.
- Expand the definition of “high-quality” early care and education to include teaching practices recommended by expert researchers and teachers, such as limiting whole group instruction and promoting multi-turn teacher-child conversations.
- Assess whether early childhood education programs with the recommended practices described above generate positive medium- and longer-term impacts.
- Evaluate, at scale, promising scholarship and support programs that seek to improve low-income students’ access to higher education and to increase their success after enrollment.
- Evaluate sector-based training programs at scale, especially at community colleges.
- Design and evaluate the long-term impacts of interventions to reduce harsh school discipline and racial disparities in harsh school discipline.
Child and Maternal Health
- Conduct rigorous evaluations of the longer-run impacts of programs based in schools, clinics, community care settings, and primary care providers addressing children and youth’s mental health needs.
- Conduct rigorous research around the most effective gun safety measures to reduce youth firearm-related injury and death, the leading cause of death among U.S. children.
- Conduct rigorous research on the prevalence and effects of air pollution—indoors and communitywide—on child health and wellbeing.
- Conduct long-term evaluations of nutritional and other child health interventions.
Family Income, Wealth, and Parental Employment
- Evaluate the long-run effects of federal assistance programs (e.g., SNAP, NSLP, LIHEAP, SSI), including strategies to reduce administrative burden to increase take-up.
- Evaluate promising programs at scale that attempt to promote high-wage jobs and ensure that low-income people have access to them (e.g., Good Jobs Challenge, Good Jobs Initiative).
- Evaluate the medium- and long-term effects of regular unconditional payments for children (e.g., the expanded refundable Child Tax Credit that was in effect in 2021).
Family Structure
- Evaluate potentially promising strategies for promoting two-parent family structures.
Housing and Neighborhood Environments
- Conduct long-term follow-ups (15–20 years) of broad-based housing vouchers and the Low Income Housing Tax Credit.
- Include small grants for remediation of toxic environments in programs to improve housing quality and monitor long-term child outcomes of remediation.
Neighborhood Safety and the Criminal Justice System
- Evaluate promising crime-prevention programs at scale, such as Becoming a Man, community policing, and others.
- Evaluate long-term impacts of supportive alternatives to juvenile arrest and detention.
Child Welfare System
- Conduct rigorous evaluations of policies and programs to reduce stigma and administrative burden and provide integrated services to families and children in the system to identify the most promising policies and practices for subsequent scaling up and long-term assessment.
__________________
NOTES: LIHEAP = Low Income Home Energy Assistance Program; NSLP = National School Lunch Program; SNAP = Supplemental Nutrition Assistance Program; SSI = Supplemental Security Income.
most in need. Evidence presented in Chapters 2 and 3 shows that intergenerational poverty is much more prevalent in some communities and population groups than others, in particular among Black and Native American children and youth. But high-risk subgroups can also be found among Asian, Latino, and White communities, as well as in other groups. Community-based strategies for developing tailored interventions are important (see below), but so are replication studies of programs that have proved successful in the context of other communities. Funding for these priorities should be responsive to Executive Order 13985 (January 20, 2021) on “Advancing Racial Equity and Support for Underserved Communities Through the Federal Government.”2
Other Guidance
Experiments and quasi-experiments have the potential to deliver convincing estimates of policy and program impacts for the populations studied, in the context in which the interventions and their sequelae unfold. As valuable as this is, there are a number of ways to enhance these designs:
- Supplement the “black box” information emerging from experimental studies with information about the active ingredients driving the effects. Information on drivers can come from observational studies of program features and implementation quality, and from mixed-method designs involving open-ended interviews with participating families describing how they experienced the policies and programs (Weisner, 2005).
- Enrich evaluation studies with multidisciplinary and diverse research and implementation teams. Economics, sociology, cultural anthropology, psychology, and subject matter expertise in the particular domain (e.g., education, health care) and program evaluation all have something to contribute, especially since the causes and effects of intergenerational poverty span so many domains. If the research and implementation team lack diversity in their perspectives and backgrounds, miscommunication with the communities being studied and failure to take account of important control variables can undermine the validity of experiments and long-term follow-up studies. See Box 11-2 for how the Moving to
___________________
2 https://www.whitehouse.gov/briefing-room/presidential-actions/2021/01/20/executive-order-advancing-racial-equity-and-support-for-underserved-communities-through-the-federal-government/
- Involve the communities under study and seek their input even before the research gets under way. Since people whose families have been poor for generations are vulnerable and perhaps also suspicious of researchers, it is imperative, in ethical and practical terms, for research on intergenerational poverty to work with the communities involved. The committee’s listening sessions with Native American families, low-income families, and others underscored this point. Communication needs to be a two-way street—ideally, it will result in the informed participation of community members and also help the research team identify changes that will make the research design most effective in the given setting. Communication also needs to continue throughout the study—whether during initial experiments or during longer-term follow-ups.
Opportunity experiment benefited from a mixed-method, multidisciplinary approach.
BOX 11-2
Mixed Methods and Interdisciplinary Teams in the Moving to Opportunity (MTO) Residential Mobility Experiment
In the mid-1990s, the U.S. Department of Housing and Urban Development (HUD) launched MTO to test whether families living in public housing projects in high-poverty neighborhoods of five large inner cities could improve their lives and the lives of their children by offering them housing vouchers that enabled them to move to lower-poverty neighborhoods. The original conception was that the program would improve the long-term housing, employment, and educational achievements of the families participating in the program.
Shortly after the program began, HUD invited proposals for researchers to conduct intensive studies at each of the five sites to learn more about the early operation of the program. Qualitative work in the Boston site found that the most salient issues motivating interest in the program were not related to employment or children’s education. Instead, concerns over neighborhood safety and the perceived effects of neighborhood crime on mental health dominated many of these conversations (Kling et al., 2001). This led the larger evaluation team to gather a great deal of information regarding the mental and physical health of participants and their families and to bring a preeminent mental health researcher onto the project leadership team. Subsequent evaluations found important impacts on mental and physical health and very few impacts on employment and educational outcomes (Ludwig et al., 2011).
CREATING A FEDERAL DATA INFRASTRUCTURE FOR RESEARCH USE
Through censuses, ongoing surveys, and forms used in program administration (e.g., income tax forms), the federal government regularly collects data on a broad set of social and economic characteristics that can support research and policy analysis relevant to intergenerational poverty and economic mobility (see Chapter 2 for examples of relevant research). State governments, too, keep useful program administrative records. The data are controlled by different agencies, however. For this and other reasons, the data are often difficult to link or to use for evidence building, whether performed in universities, private research organizations, governmental agencies that are not the data custodians, or cross-organization research teams.
Removing the barriers to data linking and putting research and policy evaluation uses on the same footing as custodial agency uses of linkable datasets would allow researchers to monitor longer-run outcomes of promising interventions and initiatives from controlled experiments and replication research. Researchers could also analyze natural experiments, monitor trends, and assess the effects of changes in government programs, such as the expanded Child Tax Credit that was provided to families in 2021 and then allowed to expire. Such research could be conducted retrospectively (as in Hoynes et al., 2016) or prospectively. Moreover, in the case of prospective research, the use of linked datasets would be much more cost-effective than original data collection.
The recently proposed student loan debt forgiveness plan provides another example of how data linkages across agencies could facilitate analysis of the impacts of policies on economic well-being, both now and in the future. Currently, there is no direct way to estimate the incidence of student loan forgiveness by students’ own income, family income, or other demographic characteristics like race, nor whether and how student debt impacts intergenerational mobility out of poverty. If data from the Department of Education’s National Student Loan Data System3 could be linked to Internal Revenue Service (IRS) income data and Census Bureau demographic data, it would be a straightforward matter to determine who in the population would benefit from student loan forgiveness. While all the individual datasets currently exist, they exist in isolation, they are not easily linked, and their use is restricted to a narrowly defined set of purposes.
In this section we discuss available data, including federal censuses and surveys and tax and other administrative records, and the opportunities and challenges for linkages, using as our prime example the domain of economic
___________________
3 National Student Loan Data System-Catalog. https://catalog.data.gov/dataset/national-student-loan-data-system
resources (income, assets, debts, employment). There are important data sources and linkage possibilities for other domains as well, including health, education, and criminal justice, which we summarize in Box 11-3. We also list potential enhancements to existing panel surveys for intergenerational poverty analysis in Box 11-4.
Data Sources and Linkage Possibilities for Economic Resources
Federal Censuses and Surveys
Since 1940, the U.S. Census Bureau has collected data on household and family income and employment in censuses and surveys, including what is now known as the Current Population Survey Annual Social and Economic Supplement (CPS ASEC), the American Community Survey (ACS), and the Survey of Income and Program Participation (SIPP). Because they gather a wide range of contextual variables, including race/ethnicity, educational attainment, and occupation, which allows for the examination of disparities among population groups, these surveys facilitate the assessment of economic well-being both at a given point in time and across time. However, on their own they do not provide an adequate platform for analysis of intergenerational poverty or mobility. Income reporting in surveys has become substantially less complete over the decades, and data on wealth are not sufficiently detailed. Furthermore, while these surveys support repeated cross-sectional (time-series) analyses, they do not provide linked longitudinal data on individuals and families, with the exception of short periods in the CPS ASEC and SIPP.
Tax Data
Painting an accurate picture of economic resources for U.S. households over time requires IRS tax data, linked for families across years, as well as relevant survey and other administrative data. IRS data are highly sensitive, and access to them, including for program administration and other legal purposes, is tightly constrained by Title 26, Section 6103, of the U.S. Code, Confidentiality and disclosure of [tax] returns and return information. Section 6103(a) states that, generally, “return information shall be confidential.” Nowhere is there a provision for research or policy analysis uses of tax data, except for specific purposes by specific agencies.
Several departments and agencies have limited access to IRS data for statistical purposes under Section 6103(j), including the Department of Commerce, for which the rule reads as follows:
BOX 11-3
Data on Health, Education, and Criminal Justice
Health Data
Birth and death records going back to 1938 contain relevant information, such as birthweight and cause of death, for analyzing the effects of health on economic wellbeing. These records are owned by vital registration areas (the 50 states, the District of Columbia, New York City, and the territories), and each area determines who can use them, for what purposes, and at what cost. They are available going back to 1968 for linkage projects at the National Center for Health Statistics (NCHS) but only after a lag, at a cost, and within the NCHS secure data center.a
Medicare and Medicaid claims records, available to the Census Bureau and researchers (in the case of Medicaid with significant lags), are another potentially useful source of information on health conditions and treatments among poor populations. The Health and Retirement Study (HRS) regularly links Medicare and Medicaid claims records and Social Security and Veterans Administration records with responses from survey participants who have consented to the linkage.b While the HRS is not designed for analysis of intergenerational poverty, as it follows only about 20,000 couples and people living alone aged 50 and older, the HRS illustrates what can be done with linkages to medical claims and beneficiary records accessed in a secure online environment (see also Box 11-4).c
Local data on children’s health can contribute to experiments and projects that scale up promising interventions at specific sites. Public health databases may include information on child blood levels, and hospital inpatient and emergency department records may provide useful data. Because children are relatively healthy (as compared with older adults, for example), access to administrative records is vital for meaningful analyses of the determinants and consequences of child health.
Education Data
Many states have built detailed Statewide Longitudinal Data Systems (SLDS) that follow school children at least through high school and in some instances through college and into the labor market.d These data could be invaluable for intergenerational poverty research. However, each state has its own arrangements for use, and inter-
pretations of the Family Educational Rights and Privacy Act (FERPA) at present make it difficult for researchers to link the SLDS to other data except for state-commissioned projects.e
Criminal Justice Data
The involvement, through incarceration or parole, of people in poverty with the criminal justice system severely constrains their families’ current and future economic well-being. The Bureau of Justice Statistics, beginning in 2008, developed a system to collect records from all state and federal criminal justice agencies to track recidivism, which linked with other data could potentially provide useful information for intergenerational poverty research.f The Criminal Justice Administrative Records System (CJARS), begun in 2016, is another potentially useful data system for such research. At the University of Michigan, CJARS collects longitudinal electronic records from criminal justice agencies and harmonizes these records to track a criminal episode across all stages of the system. At the Census Bureau, harmonized criminal justice records can be linked anonymously at the person-level with extensive social, demographic, and economic information from national survey and administrative records. At present, CJARS has records from 35 states, with longitudinal records covering 68% of U.S. population.g
__________________
SOURCE: Committee generated.
NOTES: aSee National Vital Statistics System - Health, United States (https://www.cdc.gov/nchs/hus/sources-definitions/nvss.htm) and Research Data Center Homepage (https://www.cdc.gov/rdc/). bAvailable Restricted Data Products, Health and Retirement Study (umich.edu). It also has links for some respondents to 1940 census data—see Box 11.4 (https://hrs.isr.umich.edu/dataproducts/restricted-data/available-products). cSee, e.g., Davis et al. (2022), Time to Dementia Diagnosis by Race: A Retrospective Cohort Study, which identifies lags in diagnosis for Black Americans. dStatewide Longitudinal Data Systems Grant Program – “About the SLDS Grant Program” (ed.gov). https://nces.ed.gov/programs/slds/about_SLDS.asp. eFamily Educational Rights and Privacy Act (FERPA) (https://www2.ed.gov/policy/gen/guid/fpco/ferpa/index.html); see also National Research Council (2009) (https://nap.nationalacademies.org/catalog/12514/protecting-student-records-and-facilitating-education-research-a-workshop-summary) fNational Recidivism and Reentry Data Program, Bureau of Justice Statistics (ojp.gov). https://bjs.ojp.gov/recidivism-program gCriminal Justice Administrative Records System (cjars.org).
BOX 11-4
Enhancing Panel Surveys for Intergenerational Poverty Research
Several long-running panel surveys have contributed to understanding long-term poverty, including the Panel Study of Income Dynamics (PSID), the National Longitudinal Surveys of Youth, and the National Longitudinal Study of Adolescent to Adult Health. For cost and feasibility reasons, panel surveys generally have relatively small overall sample sizes. Yet there are opportunities, outlined below, to enhance their ability to identify areas for targeted research and follow-up of policy and program interventions that could ameliorate intergenerational poverty.
Increasing Samples for High-Risk Groups
To respond to the Equitable Data Working Group’s report, issued in April 2022, which calls (p. 6) for more “disaggregated statistical estimates . . . for understanding potential disparities in life experiences and outcomes across demographic groups,” panel surveys would need to find cost-effective ways to increase sample size for groups most at risk of intergenerational poverty. Panel surveys could gather some of their information from administrative data records or possibly cut back on sample subgroups at the higher end of the income and wealth continuum. Funders could also support small additional samples of high-risk groups in several surveys, with results pooled for greater statistical reliability.
Adding Questions on Respondents’ Place of Birth and Childhood
Adding questions to panel surveys (and repeated cross-section surveys named in the text) on where the respondent was born and grew up would increase their value
-
Statistical use
-
Department of Commerce
Upon request in writing by the Secretary of Commerce, the Secretary [of Treasury] shall furnish—- such returns, or return information reflected thereon, to officers and employees of the Bureau of the Census, and
- such return information reflected on returns of corporations to officers and employees of the Bureau of Economic Analysis, as the Secretary may prescribe by regulation for the purpose of, but only to the extent necessary in, the structuring of censuses and national economic accounts and conducting related statistical activities authorized by law (Confidentiality and Disclosure of Returns and Return Information, 1979).
-
Department of Commerce
for intergenerational poverty research. Indeed, the PSID and several other panel surveys, including the Health and Retirement Survey, not only have survey information about respondents’ childhoods, but also undertook a project to link public 1940 census records to respondents who were alive at the time. The 1940 census provides a wealth of detail for linked respondents’ families and neighborhoods.a
Obtaining Additional Detail on Race and Ethnicity
It would be useful for panel surveys to gather additional details regarding identity beyond the standard race and ethnicity (Hispanic or Latino) categories—for example by asking about membership in specific Asian groups or American Indian tribes or about White or Black origins. The U.S. Office of Management and Budget issued a Federal Register Notice on January 27, 2023, which proposes as the new federal standard a detailed combined question in which Hispanic or Latino would be a race/ethnicity category and a new category would be added for Middle Eastern and North African.b
__________________
SOURCE: Committee generated.
NOTES: aCensuses are opened to the public 72 years after completion—the 1950 census forms were released in 2022 and are now being digitized—and it will be valuable to be able to link future surveys back to them. Surveys with adequate locating information can be linked to more recent censuses, not yet public, inside secure Census Bureau computing facilities. bFederal Register: Initial Proposals for Updating U.S. Office of Management and Budget’s Race and Ethnicity Statistical Standards. https://www.federalregister.gov/documents/2023/01/27/2023-01635/initial-proposalsfor-updating-ombs-race-and-ethnicity-statistical-standards?et_rid=35386254&et_cid=4581739
Research and policy analysis access to the IRS information available to the Census Bureau (comprising many but far from all items, as spelled out in regulation 6103(j)(1)-1)4 is provided through the network of Federal Statistical Research Data Centers (FSRDCs) that the Census Bureau operates for the federal statistical system. Research proposals must undergo a lengthy review process by both the Census Bureau and the IRS and be justified in terms of utility to both agencies, and outputs must clear high hurdles for confidentiality protection. In addition, many of the institutions that house research data centers charge researchers for a “seat.”5 The Statistics of Income program of the IRS has its own small program of research access to tax data (which must be carried out at IRS facilities when individual
___________________
4 Appendix C: Chapter 11, Table 11-1 documents types of IRS data available and not available to the Census Bureau under regulation 6103(j)(1)-1.
5 Federal Statistical Research Data Centers (census.gov). https://www.census.gov/about/adrm/fsrdc.html
records are analyzed), and approved research must serve the purposes of tax administration.6
Administrative Data on Other Sources of Economic Support
Gaining a comprehensive picture of available economic resources requires access to administrative records from programs that provide nontaxable benefits to families. These include such income support programs as the Supplemental Nutrition Assistance Program (SNAP), the National School Lunch Program, housing subsidies, the Low Income Home Energy Assistance Program, the Supplemental Security Income program, and others. Many of these records are held by the states, only some of which provide access to the Census Bureau or other agencies for statistical purposes (see Appendix C: Chapter 11, Table 11-2).
A complete picture also requires records on major assets and debts, such as financial assets (savings, checking, securities), home equity, retirement equity, education tax-deferred accounts, and health insurance coverage on the asset side, along with debts for mortgages, credit cards, student loans, and medical care on the debit side. Tax returns provide or can be used to infer some of these items, and surveys ask about some of them, but additional sources are needed to obtain a complete picture—for example, records from the National Student Loan Data System.7
Promising Developments
Data Linkage Projects
Several statistical agencies are engaged in linkage efforts to help understand economic mobility and child outcomes that can foster mobility. For example, the National Center for Health Statistics has linked 20 years of data (1999–2018) from the National Health Interview Survey and the National Health and Nutrition Examination Survey with HUD records on major housing assistance programs.8 Over a dozen research reports have already benefited from the linked data9—see, for example, Fenelon et al.
___________________
6 SOI Tax Stats - Joint Statistical Research Program | Internal Revenue Service (irs.gov). https://www.irs.gov/statistics/soi-tax-stats-joint-statistical-research-program
7 The Federal Reserve Board’s Survey of Consumer Finances provides comprehensive estimates of income and wealth but is small in sample size, is conducted only every 3 years, and cannot be linked with other datasets because it includes a sample of high-income households from tax returns, which cannot be used for any other purpose.
8 NCHS Data Linkage - HUD Administrative Data (cdc.gov); see also NCHS Linked Data Table (cdc.gov).
(2018), who found that children living in public housing have better mental health outcomes than a control group (no effect appears for children whose families use housing vouchers). Three linkage projects at the Census Bureau are particularly promising for retrospective and prospective research on intergenerational poverty owing to their content and time span: the American Opportunity Study (AOS), which is supported by the Decennial Census Digitization and Linkage (DCDL) project; the Comprehensive Income Dataset (CID); and the Opportunity Atlas (see Box 11-5). There are also important efforts that take advantage of data that the states maintain
BOX 11-5
Data Linkage Projects at the U.S. Census Bureau
The AOS is an ongoing effort at the Census Bureau in collaboration with researchers at Opportunity Insights, the University of Michigan, and Stanford University to link the 1960–2020 censuses and ACS, thereby transforming cross-sectional data into longitudinal data representing the full U.S. population over the last 70 years. This panel can be continuously refreshed as additional census and ACS data become available and serve as a spine on which to append tax records, earnings reports, and program records. The Census Bureau, the National Academies, Opportunity Insights, and several foundations funded the infrastructure for the AOS. The DCDL project is linking the 1960–1990 censuses, which is the last component to be completed for the AOS (see https://www.census.gov/programssurveys/dcdl.html).
The CID is a foundation-funded project of Bruce Meyer et al. at the University of Chicago working with the Census Bureau to combine surveys, tax records, and federal and state benefit program records. The goal is to overcome inaccuracies in basic understanding of economic well-being. To date, the CID project has linked tax records and 12 sources of federal and state administrative program data with the CPS ASEC, the ACS, and the SIPP, and has produced new research on extreme poverty and homelessness. The intention is to extend the dataset back in time for two decades and to update it continuously going forward (Corinth & Meyer, 2021; see also the CID website at https://cid.harris.uchicago.edu).
The Opportunity Atlas is a collaborative, foundation-funded effort by Opportunity Insights in cooperation with the Census Bureau to construct a small-area (census tract level) linked dataset for analyzing the economic mobility of children born between 1978 and 1983. The atlas, which is built from the 2000 and 2010 decennial censuses linked to tax returns and the 2005–2015 ACS, contains estimates of children’s outcomes in adulthood, including earnings and incarceration (see The Opportunity Atlas website at https://www.opportunityatlas.org/).
SOURCE: Committee generated.
that are not available at the federal level—for example, the Texas Education Research Consortium has linked K-12 and higher education records to children’s later labor market outcomes, and the California Policy Lab has linked a range of state-level data covering education, safety net participation, and criminal justice involvement.
Legislation, Regulations, and Other Support
There is growing recognition that research and policy analysis using linked survey and administrative data is key to advancing knowledge in many areas, including intergenerational mobility and poverty. New legislation—the Foundations for Evidence-Based Policymaking Act of 2018 (Foundations Act)—and reports of commissions that led to and were established by the Foundations Act represent important steps forward (see Box 11-6). Reports of the National Academies’ Committee on National Statistics (National Academies, 2017a,c, 2022a) stress the need for “blended
BOX 11-6
Relevant Legislation and Statements of Support for Linked Data for Evidence
The U.S. Commission on Evidence-Based Policymaking, chartered in March 2016 by the Evidence-Based Policymaking Commission Act (P.L. 114-140), in its final report envisioned “a future in which rigorous evidence is created efficiently, as a routine part of government operations, and used to construct effective public policy” (Commission on Evidence-Based Policymaking, 2017, p. 1). The report’s 22 recommendations address four areas: modifying federal laws to facilitate data use; establishing a National Secure Data Service (NSDS, which would not store data, but instead link and return privacy-protected data to the requesting users); instituting processes to improve data access and transparency; and designating leadership positions to support evidence generation and use in government.
The Foundations for Evidence-based Policymaking Act of 2018 (Foundations Act) enacted almost half of the commission’s recommendations. Title III contains important provisions for linking data to analyze economic mobility over time: It incorporated the Confidential Information Protection and Statistical Efficiency Act (CIPSEA), originally enacted in 2002 (CIPSEA enables researchers to become sworn agents of a statistical agency to gain access to confidential data in an FSRDC); codified Statistical Policy Directive No. 1 on the responsibilities of statistical agencies to produce relevant, objective data; added the presumption that statistical agencies may, on request, obtain federal data for evidence-building; expanded secure access to CIPSEA data assets; required a standard data application process for researchers to use confidential data; and charged the U.S. Office of Management and Budget (OMB) to coordinate statistical
data” to improve the timeliness, relevance, quality, and granularity of federal statistics and data for evidence-building. The value of research access to confidential datasets through the FSRDC network was shown in a recent study that took advantage of the natural experiment afforded by the roll-out of FSRDC sites throughout the country (Nagaraj & Tranchero, 2023).
Remaining Challenges for Economic Opportunity Research
Significant technical, privacy, feasibility, and legal challenges must be overcome to achieve the vision outlined above. These are outlined next.
Technical Challenges
The Census Bureau has developed robust linking software, which performs very well for people with Social Security Numbers or Taxpayer Identification Numbers and reasonably well using date of birth and name
agencies’ confidentiality and disclosure policies. The chief statistician’s office in OMB is to issue regulations for these provisions, which will be key for researcher access to linked data.
The Advisory Committee on Data for Evidence Building (ACDEB), chartered by Title I of the Foundations Act, in its interim report (ACDEB, 2021) supported the concept of an NSDS and urged OMB to produce guidance and regulations on the presumption of accessibility for statistical agencies, expansion of access to CIPSEA datasets, the responsibilities for statistical agencies, how agencies should implement “open data by default,” and interagency and intergovernmental data-sharing responsibilities. ACDEB’s final report (2022) focused on establishing the NSDS. It recommended that funding in the CHIPS Act of 2022 (Section 10375) for the National Science Foundation to pilot the NSDS be assigned to National Science Foundation’s America’s DataHub Consortium; that the NSDS be a legally recognized entity owned by the federal government and operated by a contractor; that the NSDS facilitate analysis by researchers of data assets hosted by affiliated organizations, including federal, state, territorial, local, and tribal governments, nonprofits, and other organizations; that NSDS core functions be funded by Congress; and that it develop a mixed funding model to include federal, state, private-sector, and user support.
__________________
NOTES: The full text of the Foundations Act is at PUBL435.PS (congress.gov); Implementing the Foundations for Evidence-Based Policymaking Act at the U.S. Department of Health and Human Services, is a good summary. The ACDEB interim and final reports and other information are available at Advisory Committee on Data for Evidence Building, U.S. Bureau of Economic Analysis. See also America’s DataHub Consortium: Seeing—and understanding—the entire elephant, National Science Foundation.
SOURCE: Committee generated.
as the linking variables. It misses some people, however, particularly immigrants and low-income individuals. Moreover, linking errors have not been well studied. Other technical issues include the handling of data gaps and inconsistencies.
Privacy Protection
Threats to privacy and the unauthorized disclosure of information collected under a pledge of confidentiality (e.g., tax and census data) have increased over the decades. New methods are under development to reinforce protection against disclosure, including algorithms that satisfy differential concepts of privacy and facilitate multiparty computing. These methods are in their infancy, however, in terms of feasibility and their ability to preserve accuracy. In fact, there is evidence that the Census Bureau’s implementation of differentially private algorithms and deletion of previously available data in the 2020 census products has adversely affected data users out of proportion to the gains in privacy protection.10 In contrast, the Year 2 Advisory Committee on Data for Evidence Building (2022) report concludes that:
(1) disclosure risk is on a continuum and is not binary, (2) not all data are equally sensitive, (3) there is shared responsibility between the statistical agency and users for protecting and not disclosing or re-identifying data, and (4) there is a need to protect good faith actors (i.e., data providers and users who take all precautions appropriate for known risks; 2022, p. 34).
The report also explicitly recommends (p. 28) a risk-utility framework for balancing data utility with disclosure risks in determining appropriate levels of privacy protection for federal data assets (see also Hotz et al., 2022).
Feasibility
The Census Bureau is currently the locus for relevant linkage projects, such as the DCDL. However, it has no specific appropriations for such work and thus requires significant funding from foundations and other sources. The Foundations Act and the ACDEB reports clearly envision the NSDS as the future locus of linkage work. However, the NSDS is just getting under way through America’s DataHub Consortium at the National Science Foundation, and it has no guarantee of a sustained funding stream nor that the Census Bureau or the IRS will make their confidential data available to it.
___________________
10 See letter from Steering Committee, Federal-State Cooperative for Population Estimates, and signatories (2022, August 1) to Census Bureau director Robert Santos. FSCPE-SC-LetterToDirectorSantos-8-22-22.pdf (cornell.edu).
Whichever agency handles linking, access to linked data for research and policy analysis would presumably continue in the FSRDC network. The current processes for securing approval and operating the centers leave much to be desired; they are neither timely nor efficient, and they require resources if researchers are to get started, let alone complete their analyses.11 It is hard to imagine the centers, as currently operated and resourced, scaling up to handle much additional load, particularly given the need for enhanced privacy protection, which requires significant agency staff time.
Legal Challenges
Despite undoubted advances in normalizing the use of linked datasets for research and evidence building, significant legal gaps remain. Titles 13 and 26 still restrict the use of census and IRS data to projects that will benefit the Census Bureau and tax administration, respectively. Moreover, IRS regulation 6103(j)(1)-1 omits important tax return data from the list of items available to the Census Bureau. Similarly, laws governing other federal agencies, such as the Social Security Administration, and state laws make it difficult to combine data from multiple sources. No law includes an affirmative presumption of research access.
Another challenge is the absence of any legal provision, except for the authorizing legislation for the Institute of Education Sciences, which houses NCES, that places the onus for privacy protection on data users (including researchers) as well as agency staff.12 Imposing substantial penalties on users as well as agency staff for disclosing confidential information, as the Year 2 ACDEB report endorses (see above), should help statistical agencies strike a reasonable balance between disclosure protection and data usability.
The chief statistician’s office in the OMB is the relevant locus for facilitating the drafting of legislation, regulations, and guidance that would make it easier to link federal and state census, survey, and administrative program records. Such linkage capabilities would allow for monitoring and
___________________
11 The single sign-on system for researcher access to confidential data, mandated in the Foundations Act, is operational through the University of Michigan, but its functionality is limited to assisting researchers to locate data files of interest. The proposal and approval process for gaining access to files for research has not yet been streamlined. See Standard Application Process to access federal confidential data – National Science Foundation invites comments on process, common form—EconSpark (https://www.aeaweb.org/forum/2997/standard-application-process-federal-confidential-comments).
12 The language is in the Education Sciences Reform Act of 2002, Section 183, parts c and d (see National Center for Education Statistics [2019], Restricted-Use Data Procedures Manual, App. D).
valuable research on long-term economic mobility in the United States and in certain population groups and geographic areas.
CONCLUSIONS AND RECOMMENDATIONS ON RESEARCH AND DATA NEEDS
Experiments and Long-Term Follow-Ups
Conclusion 11-1: In many domains, such as education, there is a lack of strong causal evidence about the effects of policies and programs on intergenerational poverty at the needed scale. Sometimes this is because careful research has failed to establish long-term effects. More often, the issue is a lack of data that would support estimates of long-run program impacts.
Conclusion 11-2: For many reasons, it is difficult to conduct research on intergenerational poverty and effective policies and programs to reduce it. Owing to the scale of effort required, it is suggested that funding organizations (public and private) consider joint grantmaking and the adoption of the following funding principles and research best practices to maximize the likelihood of achieving valid results:
A Federal Data Infrastructure for Research Use
Conclusion 11-3: Existing census, survey, and administrative data linked for families over time and across subject domains—income, wealth, demographics, health, education, and others—can facilitate cost-effective research and evidence building on intergenerational poverty and socioeconomic mobility, looking both backward and forward in time. The research and policy analysis community needs timely, cost-efficient access to linked datasets with appropriate confidentiality protection.
Conclusion 11-4: At present, data for studying intergenerational poverty and related topics are controlled by a variety of federal and state agencies and are difficult to link or use for research or policy evaluation. Recent developments designed to ameliorate this situation include the Foundations for Evidence-based Policymaking Act of 2018, which presumes access to federal data by statistical agencies for evidence-building and calls for a streamlined process for researcher access to such data; supportive reports of the Commission on Evidence-based Policymaking and other organizations; and innovative projects at the Census Bureau and other agencies aimed at building linked datasets.
Conclusion 11-5: Significant challenges remain for access to linked datasets for analysis of intergenerational poverty and related topics. They include technical issues related to constructing and evaluating linked datasets; technical and policy issues regarding new methods of privacy protection and their effects on data accuracy; making access feasible in terms of cost, timeliness, and adequate budgets for the agencies linking the data; and legal barriers (e.g., research and evaluation must be justified in terms of agency benefits).
Recommendation 11-1—To facilitate research and evidence building on economic opportunity, intergenerational poverty, and related topics, the Chief Statistician at the Office of Management and Budget (OMB) should:
-
Work within OMB and with relevant agencies and congressional committees to amend the Foundations for Evidence-based Policymaking Act to:
- include a presumption of secure access to confidential data for research and policy evaluation, explicitly superseding provisions in U.S.C. Titles 26 and 13, which require research to benefit the Internal Revenue Service (IRS) and the Census Bureau, respectively;
- provide secure access for statistical use, research, and policy evaluation to records of state benefit programs that receive federal funds (e.g., the Supplemental Nutrition Assistance Program);
- require federal agencies with custody of confidential datasets to use a risk-utility framework for determining appropriate privacy protection methods for their data; and
- impose penalties on researchers and other data users for willful, harmful disclosure of confidential data, similar to the penalties imposed on statistical agency staff;
- Work with the IRS Statistics of Income Division and the Census Bureau to expand the tax items available to the Census Bureau under regulation 6103(j)(1)-1 for research use;
- Work within OMB and with relevant agencies and congressional committees to secure sustained funding for data linkage projects, Federal Statistical Research Data Centers, and technical capacity in the states to share records to support cost-effective research and policy analysis on intergenerational poverty, economic opportunity, and related topics; and
- Work with relevant agencies to establish guidelines for consent and data storage that will facilitate the re-use of survey and intervention data, linked to subsequent administrative records, for long-term follow-up and for studies not yet anticipated at the time of the original study.
