Table of Contents | Previous | Next |
IV. Recommendations for Future Research
In this synthesis, we have reviewed data and research on the changing status of current and former welfare recipients during the first decade of TANF. Welfare as we knew it in the 1970s and 1980s under the AFDC program can no longer be considered a relevant analog to the TANF program and its myriad incarnations at the state and local levels. Consequently, future research on TANF need not focus on how the transition from AFDC to TANF affected caseloads and the status of current and former welfare recipients. Rather, in moving forward, research should examine how TANF is evolving and how well it can meet the needs of low-income families today and in the future. We see five broad areas for future research on TANF: (1) data needs and capacity building; (2) understanding changes in welfare participation; (3) tracking current and former welfare recipients to identify persistent and emergent problems and needs; (4) understanding how specific features of states’ TANF and public assistance programs influence the well-being of current and former recipients; and (5) expanding beyond the TANF program to learn how other public assistance programs are interacting with and serving the needs of low-income families. We discuss each category in turn below.
Data Needs and Capacity Building
Before outlining suggestions for future research, we address some broad data needs that are critical for research on TANF-related topics and make recommendations for how the federal government can help to address these needs. Our recommendations here concern both administrative and survey data collection. Although there are pros and cons to each type of data (discussed in chapter 1), they both have the potential to support future research.
Survey Data Improvements. High quality survey data provide a promising source of information on the status of current and former TANF recipients at the national level. These data can identify characteristics and outcomes for welfare populations that are not well captured in administrative records. They can also detect employment not captured in state UI systems, allow analysts to compute wage rates, assess such job-related features as benefits and on-the-job training, and obtain data about difficulties on the job and getting to and from work.77 Survey data can also be used to track the well-being of children across multiple dimensions (parent-child interactions, behavioral problems, school performance, health status, etc.).
Changes in the welfare population and the policy and programmatic responses to its needs necessitate refinements to the information collected through survey sources. A growing share of welfare cases have no adult/parent in the assistance unit—they are child-only cases. Given the growth in these cases, it is important for research to determine how well survey data do in identifying them. Survey respondents may not consistently and accurately be able to identify the family members that welfare income is meant to benefit from the standpoint of program rules. In addition, many TANF agencies are providing noncash services to families, such as child care, transportation assistance, and post-employment supports. It will be useful to improve our survey measures of TANF services, possibly by asking directly about receipt of these specific noncash services.
For both monitoring and analytic purposes, it is important that general surveys have a sufficiently large sample sizes or sampling strategies to ensure they capture enough welfare recipients for meaningful study. With the decline in welfare caseloads, even a survey that oversamples low-income households may capture too few welfare-reliant households to sustain meaningful analyses. For example, in the 1997 NSAF there are over 1,000 families identified as TANF leavers; in the 2002 round, the number falls to 537.
In addition, the 2004 panel marks the end of the SIPP, which is being replaced by the Dynamics of Economic Well-Being System (DEWS) in 2008/2009. For this and any other new surveys to be useful to TANF-related research, they must identify and collect information on a sufficiently large number of TANF families to provide reliable estimates of the characteristics, needs, and outcomes of the welfare population.
Another important concern about major existing national survey efforts, such as SIPP and CPS, is how well they capture TANF participation. (See our discussion earlier in the report.) The extent of potential underreporting of TANF, particularly increases in underreporting over time, needs to be understood. There also needs to be additional study of whether certain characteristics are systematically associated with underreporting and whether these relationships have changed over time.78 This type of information could potentially lead to the development of adjustments researchers could use or at least an understanding of whether underreporting can be treated as essentially random. Studies could be done that compare matched administrative records with survey data to analyze differences in reporting.
Linking Administrative Data. At the state and local level, program administration data can be used to measure how long families have stayed on TANF, the extent to which former recipients return to the program, and the state’s view of TANF-related services the family has received. Data systems across states will vary in the amount of detail they contain regarding families’ demographic characteristics and barriers to work. It would be very useful as a matter of course to link TANF program data to UI earnings records along with program data from other forms of public assistance, such as food stamps, Medicaid, WIC, housing assistance, and child care assistance. Much of this information is already collected, and linking these administrative data sources would help states track the changing profile of their welfare recipients and assess how former recipients fare in terms of employment and use of other forms of public assistance. While some efforts to link administrative data for specific states are ongoing (such as Chapin Hall’s efforts for Illinois administrative data), it is possible that the federal government could play a role in supporting other states in carrying out more linkages. This role could include technical assistance and capacity-building grants as well as potentially provide a clearinghouse or some way of streamlining research access to these data. The federal government could also conduct a review documenting the status of existing efforts to link data on a state-by-state basis.
Understanding Changes in Participation
Understanding Changes in Welfare Take-Up Rates. A central concern for TANF program administrators and policy makers is whether we have entered a new era where TANF caseloads will hover around 2 million families in the average month, or whether caseloads will revert back to their levels of decades past. This essentially is a question about welfare entry, welfare exit, and the reasons why families are increasingly likely to eschew cash assistance. Over the years since welfare reform, there has been a decline in the “take-up” rate of TANF, that is, the number of families receiving TANF as a percentage of eligible families. Additional research focused on understanding this decline in participation rates is important to understanding whether TANF is meeting its goals. This type of research requires analysis of the broader low-income population and the ability to accurately estimate potential eligibility.
A potential research project to investigate changes in welfare participation must begin with nationally representative data on low-income families from various points in time, and these data must contain enough information on living arrangements and family income so that researchers can compute program eligibility. The most obvious choice would be the Current Population Survey, which not only provides annual data from the past, but also is ongoing. In addition to the data, researchers must gather accurate information on state-specific welfare program rules and apply them to the data to determine which low-income families are in fact eligible for benefits. Finally, the data must contain accurate information on program participation or, at the very least, information on participation must be reliably imputed. These data demands are quite high; however, the TRIM3 micro-simulation model, funded by HHS, routinely takes CPS data, adjusts for under-reporting of welfare receipt to match administrative totals, determines eligibility, and computes take-up rates. Thus, researchers could use TRIM3-enhanced CPS data to assess why take up rates have changed.
Specifically, one could use regression-based decompositions to assess the factors behind declining welfare take-up rates. One could take a base year, say 1996, and a run a regression to predict welfare participation controlling for the demographic characteristics of the eligible population as well as state-level economic and policy conditions. Using the coefficients from the model, one could then take the characteristics and state conditions from a future year, say 2005, and generate predicted take-up rates. To the extent that take-up rates from 1996 differ from the 2005 predicted rates, differences in take-up rates can be attributed to differences in the eligible populations, specific state policies, and economic conditions. Differences between the predicted and actual 2005 rates are due to unobserved factors, such as changes in the practices of welfare offices that are not captured in specific policies and changes in the underlying attitudes toward welfare.
Alternatively, to better understand changes in welfare take-up rates, one could study applicants who do not end up receiving TANF benefits as well as families that are formally diverted from the program. Such a research project would require access to state administrative data to identify rejected applicants and diverted families, and then contacting and surveying them. HHS has already funded a series of similar studies on applicants from the late 1990s79 and it would be useful to replicate this work as TANF programs have become more established. Further, learning how eligible nonparticipants get by and whether they would be better off in the short and long runs had they entered TANF are key questions for policymakers.
The Role of Nonmarital Births. In addition, changes in long-term societal trends may play a part in lower caseloads. Historically, the most common reason for welfare entry was a nonmarital birth. Teenage women who entered welfare with nonmarital births were particularly prone to long periods of receipt. Teen birth rates fell during the 1990s and into the 2000s, but it is not clear whether this was in response to changes in welfare policy. Further, we do not know whether teens who do have children outside of marriage are much less likely to enter welfare today than they were in the past. If teen birth rates continue to decline (or at least remain steady) and if unwed teen mothers are now less likely to enter welfare, then we probably have entered a new historical period in which caseloads hover around 2 million with some variation around economic cycles.
Thus, it would be valuable to undertake a research project assessing whether the relationship between nonmarital fertility and welfare entry has changed. Existing data tracking teenagers such as those available in the NLSY 97 cohort (12–16-year-olds in 1997) can be used to address this question. Specifically, one can use data from the NLSY 97 and compare them with data from the NLSY 1979 cohort containing information on youth between the ages of 14 and 21 in 1979. Using these data, one can use regression-based models to assess changes in the probability that unwed teens have children and to determine whether those that do have nonmarital births are less likely to enroll in welfare or wait longer before signing up for benefits. Further, if it can be established that young parents and their children are less likely to enter welfare today than in the past, it is important to then assess how these young families are faring.
Understanding Child-Only Cases. As overall caseloads dropped, the share of TANF cases composed only of children (i.e., child-only cases in which the TANF grant is meant to support only children and not the adults in the family) grew from 23 percent of the total caseload in 1997 to 35 percent in 2001 and to 46 percent in 2005. It is important to note that the actual number of child-only cases has not risen; in fact, there are fewer child-only welfare cases in 2005 (870,000) than in 1997 (919,000). Rather, the number of adult-headed TANF cases declined by a far larger amount than child-only cases.80 There is very little research examining the characteristics and well-being of these families and their rates of entry into and exit from TANF. It would be useful to better understand the consequences of subjecting adults in these families to work requirements and other TANF policies. We would also like to know how these consequences vary across different types of child-only cases (e.g., foster families, undocumented immigrant parents of citizen children, and children in the care of their grandparents or other relatives). This kind of research requires data that identify child-only cases. Additional research needs to be done to assess how accurately current national data sets identify these cases. In addition, one could survey families receiving a child-only TANF grant drawing samples from state administrative data.
Tracking Recipients to Identify Needs
Ongoing information on the status of current and former welfare recipients is vital for program administrators and policymakers. Linked administrative data, if available, can measure how long families stay on TANF and how this changes over time, how quickly families return to TANF and the relation of returns to TANF and employment, and the extent to which TANF families connect to other programs, such as Medicaid, Food Stamps, or child welfare while on TANF and after leaving. Routine, continuous updating of linked administrative systems allows monitoring of these and other questions over time. This will be particularly important if the United States experiences a sharp or prolonged downturn in the economy during which welfare caseloads swell and families find it increasingly hard to find and keep jobs.
In addition, it is important to assess the longer-term outcomes of welfare leavers. Questions to be addressed include, what are the post-exit employment and well-being trends for former recipients? Is there a steady progression toward higher wages? Are there key points in time that are associated with quantum improvements in jobs and well-being (e.g., mothers may eschew promotions at work until their children reach middle school)? This will require analysis of longitudinal data from both administrative and survey sources. Ultimately, it will be useful to track welfare leavers for five or more years after exit to assess the extent to which these families are still struggling, what types of longer-term supports they require after leaving TANF, and whether certain types of services provided while on TANF contribute greater success in the years after exit.
Aside from gathering new data, which could be challenging and expensive, the NLSY 97 may provide the best source of data for tracking a large sample of welfare leavers for multiple years into the future. The youth in the sample are between the ages of 12 and 16 in 1997. As such, they are in their early to mid-20s today, a period of time in which low-income unwed mothers are likely to be cycling on and off welfare. Data sets such as the SIPP may not track families for a sufficiently long time to ascertain long-term outcomes. Linked administrative data can provide longer-term data on employment and program participation but little else.
Ultimately, research efforts that link survey and administrative data such as the Longitudinal Employer-Household Dynamics (LEHD) project might provide very useful longer-term information if they could be linked to administrative data on current or former welfare recipients. The LEHD is an innovative program within the U.S. Census Bureau. It uses modern statistical and computing techniques to combine federal and state administrative data on employers and employees with core Census Bureau censuses and surveys while protecting the confidentiality of people and firms that provide the data.
Linked survey and administrative data could also be used to study the employment, TANF duration, and non-TANF benefit use of recipients with barriers. Information from past surveys of TANF recipients funded by HHS, which included measures of a range of barriers, could be linked to administrative data to ascertain differences in outcomes across those with and without multiple barriers or specific types of barriers.
A key feature of the TANF program is the lifetime limit on the receipt of federal cash aid. Whether in response to labor market opportunities, welfare policies, or a combination of both, welfare caseloads have fallen, and even after 10 years of the TANF program, few families have exhausted their lifetime benefits. It is important to see if greater proportions of TANF families exhaust their lifetime benefits in the coming years and to examine how families that have reached time limits are faring. Survey or qualitative research to determine the reasons why more families are not exhausting benefits is also important, including studies of whether states have intervention strategies for those about to reach limits that are helping families find work before hitting the limit.
To this end, a series of research projects modeled after HHS’s welfare leaver and stayer studies would be useful. A selection of states could be funded to identify TANF cases that closed due time limits over a specified time, say a three-month window. Families that have recently exhausted their benefits could be contacted and surveyed, perhaps multiple times in the months and years following the termination of benefits to assess their well-being and how they are getting by. The process could be repeated using subsequent cohorts of families that have reached their time limits to see if the number of cases exhausting benefits is growing and to see if the well-being and survival strategies of these families are changing. The study may also want to include cases that closed for other reasons but were nearing their lifetime limit as well because these families may have left the rolls in anticipation of the looming time limit.
Understanding How State Policy Choices Influence Recipients
The well being of current and former TANF recipients may vary substantially from state to state and may be influenced by state policy choices. States may choose policies that have offsetting effects on various outcomes for welfare populations. For example, generous earnings disregards may reduce TANF exit rates while strict time limits may increase them. As states vary their TANF policy packages in the future, perhaps in response to the changing rules and regulations pursuant to TANF reauthorization in 2006, it will be useful to continue efforts to assess how states choose policies and how policy choices influence the employment, earnings, incomes, and well-being of current and former TANF recipients. In particular, it will be important to assess the effectiveness of the differing approaches states take toward meeting the needs of their “hard-to-serve” clients.
Several efforts are underway, funded by government and nongovernment sources, to identify the variety of approaches states are using to address the needs of clients with multiple barriers to work and how these are changing in reaction to reauthorization. A next step is research that could link the outcomes of families to the services they received. To be able to connect individual outcomes with services, information on service receipt would need to be available in administrative data or participants would need to be surveyed. Initial work could be funded on better understanding the extent to which the services families receive can be identified using state administrative data, including the quality of that data and which states’ data are most promising. In addition, a sample of families receiving “hard-to-serve” services could be contacted for more in-depth interviews. Such research would be far more effective if data from multiple states could be brought together.
Demonstration Projects. Even with a detailed set of observational data, it may still be challenging to tease out program effects because of hard-to-observe differences between states and each state’s hard-to-serve population, and because the decisions about who receives what services may reflect varying state selection criteria and client decisions about the services they wish to receive and whether they continue in the TANF program.
Ultimately, it may be useful to establish multiple hard-to-serve demonstration projects in which states (and their research partners) are asked to develop a set of services for this population and selection criteria or where the impact of specific models in place could be evaluated. This could follow the model of the Enhanced Services for the Hard-to-Employ Demonstration and Evaluation (HTE), currently funded by ACF, but with greater focus on TANF recipients (only one of the four sites in the HTE evaluation concerns TANF recipients exclusively, although others include TANF recipients). The state would randomly assign their hard-to-serve populations to different services and the effectiveness of the services could be assessed using standard experimental evaluation techniques. Multiple experimental treatments could be conducted within a single state, and different states could use an array of different approaches. Although this experimental demonstration approach is expensive and it would be years until it produced results, it could be very useful for identifying the best ways to address the diverse needs of hard-to-serve clients. This approach could enable us to answer questions such as the impact of intensive case management or the relative benefits of combining services that address barriers while requiring work versus a sequential approach.
Services from a Client Perspective. It is also important to discern how clients view the services they are receiving. A program as described may vary considerably from the way a program is experienced. Given the diversity of state and local approaches, it is probably not feasible nor would it be meaningful to survey current and former recipients about the services they received using a national sample. However, periodic local survey efforts would reveal how reliable programs “on the books” translate to programs “on the ground” and whether the needs of TANF clients are truly being identified and met. Such efforts could also help refine analyses on how state choices of programs and policies under TANF influence the outcomes of current and former TANF recipients. Such surveys could also identify whether recipients understand program rules, such as the reason for and level of sanctions or the length of time they can receive benefits. This is critical for understanding if these policies impact behavior.
Research beyond TANF
In 2005, there were 7.6 million families living below the poverty line and millions more living just above it. The average number of families receiving TANF in any given month, however, is about 2 million. Clearly, many low-income families and their children are not turning to TANF for support. As such, it is important to focus on a broader population than just those families that come in direct contact with the TANF program and to consider how other sources of support interact with and substitute for TANF.
Sources of Support. Of particular interest is identifying the sources of income (income packaging) for low-income, non-TANF families. To what extent are they receiving cash and in-kind benefits from public and private sources? Do different types of families (single mothers, cohabiting couples, etc.) use different packages of income? How much of a low-income, non-TANF family’s income comes from earnings? Are child outcomes influenced by the composition of family income? Research in this area could make use of existing secondary data sources, such as the SIPP (and its successor, the DEWS), that provide detailed information on income composition and also include information on related outcomes, such as child well-being.
In addition, linked administrative data can be used to identify families that are participating in noncash assistance programs, such as child care, Medicaid, or food stamps, who have either never or not recently participated in TANF. Researchers could compare available information about these families (including earnings) with data on families entering TANF to better understand differences in who uses what parts of the safety net. In addition, a subset of the families captured in these administrative data could be surveyed to gain a better understanding of their broader circumstances.
Disconnected Families. A substantial share of families has become disconnected from both TANF and from work. Some of the evidence we reviewed showed these families to be disadvantaged but also in transition, with some returning to welfare and some moving on to jobs. Understanding these families’ circumstances over time is critical, particularly as time limits reduce the ability of families to return to TANF. In addition, a substantial number of families in this circumstance have never received TANF benefits.
The biggest challenge in obtaining better information on disconnected families is identifying them. Administrative data can provide list samples of families that stopped receiving welfare and never generated UI wage records, but such data cannot identify families that never came into contact with the welfare system. Ongoing national surveys such as the CPS can identify disconnected families but provide little information on their well-being and survival strategies. It may be interesting to use the March CPS files to identify disconnected families and then create a special April supplement to gather more in depth information about them—about three-quarters of the families identified as disconnected in March could be surveyed in this special April supplement.
Finally, much of the research on former welfare recipients highlights that individuals were able to find jobs but that wage growth was low and job benefits minimal—in short, that these women had joined the ranks of the working poor. This research provides a natural next step to understanding the broader low-wage labor market and how workers can retain and advance in jobs. A focus on low-income families with children, in particular working single mothers, and examining what policies and programs might support their income growth, is an important part of this understanding. Some of this research is now underway. For example, HHS is funding a survey of employers in the low-wage and TANF labor market as well as evaluating retention and advancement strategies for low-wage workers.
Ten years after federal welfare reform, the nature of public assistance to low-income families has changed both in size and scope. Fewer families use cash assistance today than a decade ago. Future research needs to determine if this decline is permanent and if so, why. In addition, there is a continuing need to monitor the status of current and former welfare recipients to ensure their ongoing and emerging needs are being met. This may require creative work with administrative and survey data sources. Further, to make sure policies are having their intended effects, policymakers need to gain a better understanding of how clients perceive state policy decisions and how these decisions directly influence the status of current and former recipients. Finally, many families that are in need of assistance never come into contact with the TANF program. As such, it is imperative that policymakers and program administrators consider how the broader low-income population is faring.
| Table of Contents | Previous | Next |