2016 Robert L. Schneider ISP Dissertation Award

Contest Winner: Vincent A. Fusaro

The Spirit of ’96: Race and the Implementation of Temporary Assistance for Needy Families

Personal Statement

My dissertation examines the implementation of Temporary Assistance for Needy Families (TANF) in the states, particularly the relationship between race, majority group racial bias, and TANF policy and the drift of TANF away from cash assistance and work supports and toward other priorities. Using advanced quantitative methods, the dissertation proceeds by first clarifying the mechanism underlying the relationship between racial demographics and state welfare policy (i.e., states with larger racial/ethnic minority populations tend to design less generous and more punitive cash assistance programs). Next, it examines whether these patterns are reflected in the use of TANF resources. Structured as a block grant affording states considerable flexibility, over two-thirds of all TANF funds are now directed to initiatives other than basic assistance and work support (e.g., tax credits, two-parent family promotion). Welfare is no longer primarily a “safety net” program and is instead a funding stream for a host of interventions. Are states with larger minority populations more likely to fund particular types of programs, such as family formation initiatives? Finally, I examine whether this drift away from cash assistance and work supports and toward other priorities has affected the material wellbeing of low-income families with children. The dissertation project has a number of implications both for TANF and for redistributive policy more generally, and I envision it informing both federal and state social welfare policy.

As a cross-state study, this investigation is relevant for many U.S. households. Approximately 1.5 million families received basic cash assistance in any given month in 2014 (Administration for Children and Families, 2014). Many more families are in economic need, however—approximately 16 million households with children fell below 150% of poverty during the same period (Census Bureau, 2014). These figures include the 1.5 million households experiencing extreme poverty (less than $2 per person per day in income), a number that has been increasing as the cash safety net has weakened (Shaefer & Edin, 2015). There have also been proposals, most notably by Speaker of the House Paul Ryan, to convert programs such as the Supplemental Nutrition Assistance Program (“Food Stamps”) and Medicaid to block grants modeled on TANF. Findings from the dissertation also, then, have implications for many more of the 47 million people in poverty in the United States (Census Bureau, 2014).

The politics of “welfare” in the U.S. has long been tied to the politics of race, beginning as early as post-Civil War pension programs for widowed mothers (Ward, 2005). Previously excluded from welfare participation through both policy and administrative practice, changes in response to the civil rights and welfare rights movements in the 1960s allowed greater access to cash benefits for minority families. In 1996, welfare reform ended entitlement to cash assistance, imposing work requirements and time limits, and gave states substantial discretion over program design. States with larger populations of color on the welfare caseload, particularly blacks, tended to offer less generous benefits and implement more punitive rules (Soss, Fording, & Schram, 2011). Public policy under TANF therefore magnifies pre-existing social disparities.

My dissertation clarifies the mechanism linking race and state welfare policy—is it the demographics of the caseload, the attitudes of the local white population, or some combination? Identifying the mechanism, in turn, will lead to more effective and precisely targeted policy advocacy. If racial context in the form of caseload demographics is the stronger predictor, it suggests a national re-centralization of policy to alleviate disparities across states is needed. If racial attitudes are the key driver, then anti-racism interventions in the relevant states may be productive.

The dissertation also addresses whether racialized patterns of policy implementation have continued as TANF has drifted away from a focus on cash assistance and work support. Childcare, alternative forms of cash transfer such as tax credits, and programs to prevent pregnancy and promote two-parent family formation are all permissible uses of TANF funds. Do states in which race and racial affect are salient aspects of policymaking direct more resources to corrective programs and fewer to cash transfers and work supports? If so, the study identifies an additional inequity brought about by devolved policymaking—households in some states might receive material supports during times of hardship, while households in other states might receive corrective behavioral interventions at the cost of meeting material needs. It is even possible that the racialized pattern influences the degree of drift from basic assistance and work support, with states in which race is salient moving away from these components more quickly than others.

The vast majority of TANF policymaking scholarship focuses on the rules governing cash assistance. As cash assistance matters less within the broader program structure, it is unclear whether and how this research can assist in advocacy efforts. The dissertation will draw attention to disparities produced by alternative uses of TANF resources, particularly if different uses of funds are related to the material well-being of low-income households. In turn, advocates and policy practitioners at both the state and federal levels can work toward policy changes to eliminate such inequalities (e.g., at the state level, advocate for particular uses of resources; at the federal level, advocate for structural change in TANF policy design).

Finally, the dissertation offers more general lessons for other social welfare programs. TANF is often referred to as an element of the “social safety net” for the most disadvantaged American families, providing time-limited resources to maintain wellbeing during periods of hardship. States, though, have a number of incentives both formal (e.g., credits for caseload reduction) and informal (e.g., an imbalance between revenue and outlays in the overall state budget) to use TANF resources for other activities. It is not surprising, then, that during the “Great Recession” cash assistance rolls were only marginally responsive to the increase in economic need—states had no requirement or incentive to boost spending on basic assistance (Center on Budget and Policy Priorities, 2011).

Converting programs such as SNAP to a similar structure could further erode the public safety net, yet such proposals have gained traction in recent years. The dissertation offers lessons to policymakers and advocates on the incentives of block grant programs and the risks inherent in those structures, particularly with respect to socially disadvantaged groups. In turn, it will aid in either challenging such developments or, if they proceed, in designing policies that preserve responsiveness to economic conditions.

Overall, the dissertation addresses the challenges posed by the devolution of policy from the federal to the state level. Existing research suggests that, at least for “welfare” as cash assistance, devolution introduces racially patterned disparities in policy design. Better understanding the processes underlying this connection, whether and how they apply to other aspects of TANF, and the consequences for household-level wellbeing updates and refines this work. The dissertation identifies areas not only for improvement in TANF, but also provides guidance for the future of other components of the social safety net.

Executive Summary

The 1996 Personal Responsibility and Work Opportunity Reconciliation Act (“welfare reform”) is most well-known for ending cash benefits for low-income families as an entitlement in the United States. It also changed welfare’s policy structure. Under Aid to Families with Dependent Children (AFDC), the previous cash assistance program for families in poverty, the federal government matched state expenditures. Temporary Assistance for Needy Families, the successor to AFDC, provides states with a fixed block grant. States partially match federal contributions through maintenance-of-effort activities, programming consistent with the goals of welfare reform (ending dependence on public support, caring for children in their own homes, promoting two-parent families, and reducing out-of-wedlock births). This structure gives states flexibility over program design. It also affords them broad leeway over the use of funds. States vary in the uses to which funds are put, ranging from alternative forms of cash transfer (e.g., tax credits) to pregnancy prevention and two-parent family formation initiatives. The bulk of TANF resources are now devoted to priorities other than basic assistance, but most scholarship on the state politics of welfare examines determinants of the rules governing traditional cash benefits (Fellowes & Rowe, 2004; Soss, Fording, & Schram, 2011; Soss, Schram, Vartanian, & O’Brien, 2001). What state-level factors account for variation in the use of TANF resources? In particular, do TANF alternatives follow the racialized pattern of implementation observed elsewhere in state welfare policy? Has the transition from cash transfer to other priorities negatively affected low-income families?

The dissertation quantitatively models various aspects of state welfare policy using a sample of all fifty states. It proceeds in three parts. The first component develops an alternative explanation for racialized patterns of state policy implementation and creates a new measure to test this theory. Studies of welfare benefits and rules have long found a relationship between racial context and policy design; states in which people of color, particularly blacks, make up a larger portion of the population or the welfare caseload tend to implement less generous and more punitive programs (Fellowes & Rowe, 2004; Orr, 1976; Soss et al., 2011, 2001; Tropman & Gordon, 1978). A recent theoretical advance in this line of inquiry is Soss et al.’s (2011) Racial Classification Model (RCM), which holds that, when race is salient in a policy area, decision makers will use racial/ethnic stereotypes as a decision-making shortcut. It is possible, however, that the relationship between population demographics and policy is spurious. Individual-level studies of political behavior indicate whites often assume welfare primarily aids blacks (Gilens, 1999; Federico, 2004). Prevalence of negative racial affect among whites, not racial context, might then drive policy.

Testing these two explanations against one another quantitatively requires a measure of aggregate white racial affect at the state level. In the first portion of the dissertation, I construct such a variable using multi-level regression with post-stratification (MRP), a technique that allows for estimation of subnational aggregate opinion from national survey data, on the 2008 National Annenberg Election Survey’s racial attitudes battery (Lax & Phillips, 2009). In contrast to surveys such as the General Social Survey, which have relatively small samples and complex sampling schemes, the NAES survey has a very large sample—approximately 20,000 respondents to the relevant questions—and is therefore more appropriate for producing state estimates. The MRP procedure begins by estimating a multi-level model of an individual-level survey response (here, whether the respondent scores in the top quartile of the racial affect scale) as a function of individual demographic (e.g., age, gender) and geographic (i.e., state) characteristics. These models are then used to predict the probability of holding that opinion for every possible “type” produced from permutations of the predictors (e.g., college educated males ages 25-36 living in Arkansas). The predictions are weighted using counts of each “type” drawn from Census data. Summing these values within each state, then dividing by the population produces the estimated proportion of the state populace, restricted here to whites, holding the given opinion.

After constructing the aggregate racial attitudes measure, I test it as a predictor of cash assistance rules (e.g., time limits, sanctions) based on models used in the canonical state welfare reform studies. Here, the states are analyzed cross-sectionally within a single year at a time. For continuous variables such as benefit levels, linear regression models are estimated. Categorical variables, such as whether the state has a time limit shorter than the federal standard, are examined with logit models. The procedure is implemented for 1997 and 2013 to provide a comparison between early and contemporary TANF policy. In addition to the new racial attitudes variable, I include caseload demographics to represent the existing racial context-based theories and control for other relevant state characteristics (e.g., government ideology, economic conditions). This component validates the new variable and connects the dissertation to the larger state welfare policy literature.

The second component, the core of the dissertation, examines state TANF implementation beyond cash assistance. Applying theories of race and welfare policy, it might be expected that states with larger populations of color and/or more negative white racial attitudes will devote fewer resources to cash transfers and more to corrective programs, such as two-parent family formation.

Here, I model categorical TANF expenditures (e.g., basic assistance, tax credits, family formation and pregnancy prevention) as reported to the Administration for Children and Families using the racial demographics and racial attitudes variables as key predictors. The models are cross-classified multi-level linear models, with state-years from 1997 to 2013 nested within states and years, an approach that allows me to use multiple years of data while examining characteristics, such as racial demographics, that change only slowly. I also model change over time to identify which, if any, of the predictor variables are associated with movement away from cash assistance over the entire history of TANF. Expenditure dollar values are adjusted to constant dollars and indexed to the number of families in poverty within the state. This portion of the dissertation not only provides an additional test of theories of race and welfare, but updates the larger literature to better reflect the contemporary nature of TANF as more than basic assistance and work supports.

A key question the policymaking studies are unable to address is whether state differences in TANF priorities are related household-level outcomes. Has the drift away from cash assistance increased the risk of material hardship in low-income households with children? Are different state uses of TANF resources related to household hardship? The final portion of the study uses the scaled TANF expenditure variables as predictors in models of food insecurity among low-income (less than 185% of poverty) households with children from 1997 to 2013. Data are drawn from the Current Population Survey, a large, continually-administered social and economic survey jointly fielded by the Census Bureau and the Bureau of Labor Statistics. I construct the household material hardship measure using the food insecurity scales in the CPS’s Food Security Supplement. The models control for both household (e.g., employment, education, age and number of children) and state (e.g., unemployment rate) factors that influence food insecurity. State and year indicator variables account for other unobserved group-level characteristics. This final component connects scholarship on welfare policymaking to research on policy outcomes, two literatures that have otherwise been distinct in the investigation of welfare implementation.