The Consequences of Decentralization: Inequality in Safety Net Provision in the Post–Welfare Reform Era
Decentralized safety net programs provide much of the social provision in the US, yet the consequences for social provision have received remarkably limited attention. In this article, we examine cross-state inequality in social safety net provision from 1994 to 2014. We ask whether programs that are more decentralized in terms of policy design are more variable across states in terms of the generosity of benefits and inclusiveness of receipt and whether there has been convergence or divergence in programs affected by the Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) as well as in those that were not. We find substantial cross-state inequality in provision, with greater cross-state inequality in programs with more state discretion. In examining change over time, we find remarkable consistency in the levels of cross-state inequality; however, we also find that the devolution of authority under PRWORA increased cross-state inequality in programs affected by this legislation.
In recent years, inequality has received increasing attention in political, policy, and academic circles. In the United States, the conversation has been overwhelmingly national in scope with a focus on inequalities of income and wealth and differentials in access to remunerative employment, quality schooling, safe neighborhoods, and affordable health insurance. This national focus misses another enormously consequential axis of American inequality, one that has received inadequate attention in contemporary academic and policy circles—that is, how decentralized provision of social and health assistance has shaped geographic inequalities across the 50 US states. Policy scholars have studied cross-state policy variation, often leveraging this variation as a methodological tool for policy evaluation. Yet state-to-state policy variation in social and health assistance is rarely conceptualized as a form of inequality per se. In our view, that is precisely how it should be viewed. As Wildavsky (1985) famously observed 30 years ago, federalism means inequality.
In this article, we shine a spotlight on how federalism produces inequality through the decentralized policy designs of safety net programs. To do so, we consider two research questions. First, is the magnitude of cross-state variation in provision associated with the degree of state discretion in administration, financing, or rule making? We examine 10 federal-state programs that make up much of the safety net for low-income or unemployed adults and their families: cash assistance (Aid to Families with Dependent Children [AFDC]/Temporary Assistance to Needy Families [TANF]), food assistance (Food Stamps/Supplemental Nutrition Assistance Program [SNAP]), child health insurance (Medicaid and Child Health Insurance Program [CHIP]), child support enforcement, child-care subsidies (Child Care Block Grant [CCBG]/Child Care Development Fund [CCDF] and TANF), early childhood education (Head Start and state pre-K [prekindergarten] programs), Unemployment Insurance [UI], targeted work assistance through AFDC/TANF, child disability assistance (Supplemental Security Income [SSI]), and state income taxes for families at the poverty line. Second, has there been convergence or divergence in safety net programs from 1994 to 2014? In answering this second question, we pay particular attention to the case of the Personal Responsibility and Work Opportunity Reconciliation Act of 1996 (PRWORA), which modified state discretion over some but not all safety net programs. In examining change over time, we find remarkable consistency in the levels of cross-state inequality; however, we also find that the devolution of authority and responsibility to states under PRWORA increased cross-state inequality in programs affected by this legislation.
The Decentralized US Safety Net
Social provision in the US is unequal by design, providing tiered and categorically based assistance that varies—across jurisdictions and citizens—in both quantity and quality. Programs in the top tier are standardized or uniform in terms of their benefits and broad in terms of their coverage; programs in the bottom tier—the focus of our analysis—are narrowly targeted, means tested, and more variable in terms of the benefits they provide and what potentially eligible populations receive their benefits. While programs in the top tier are financed and administered at the federal level, the majority of the programs in the bottom tier have some degree of devolved authority or discretion to lower levels of government.
The primary programs that comprise the contemporary safety net for working-age individuals began as local charity efforts and were partially, but not completely, federalized during the New Deal Era of the 1930s and the War on Poverty and Great Society of the 1960s. Each reflects a particular negotiation of power between local and federal policy makers to determine the type and extent of state discretion, authority, and responsibility. Individual programs have evolved over time as a function of their original policy design, the negotiated settlements of federalism, and state-specific factors. This evolution has not altered the most fundamental structural feature of the safety net, however, which is decentralization of authority to state and local governments.
Assistance through what is commonly termed the “safety net”—the only assistance for most economically needy, nondisabled, working-age adults and their dependents—is a patchwork of income transfers, in-kind assistance, and services, all of which are funded with a combination of federal, state, and local tax revenues and managed either jointly by federal and state governments or wholly at the state or local levels. The primary national program that supports the economic security of these individuals is the Earned Income Tax Credit (EITC), which, while immensely important to those who receive it (Halpern-Meekin et al. 2015), is restricted to tax-filing individuals who have employment income in the previous year.
Cross-State Inequality in Safety Net Provision
This decentralized structure has produced substantial inequalities in provisions across states and across populations within states (Meyers, Gornick, and Peck 2001; Allard 2008; Lobao and Kraybill 2009; Soss, Fording, and Schram 2011). In the late 2000s, the concern about unequal responses to people’s needs was further magnified by the Great Recession. As Jason DeParle observed in reporting on the state of safety net assistance during the Great Recession, the most vulnerable victims of the Recession confronted “a jumble of disconnected programs that reach some and reject others, often for reasons of geography or chance rather than difference in need” (2009).
The extent and implications of the decentralized structure are among the most underappreciated features of the US welfare state (Pierson 1995; Howard 1999). Decentralization of control over social and health assistance reflects a trade-off between uniformity through federal provision that is reflective of equality in social rights and variability through state or local provision that is reflective of inequality in social rights (Obinger, Castles, and Leibfried 2005). Horizontal equity and economic rights are two principles used to critique this form of inequality (Marshall 1949; Finegold 2005). “Equity arguments suggest that all citizens should have access to equivalent public assistance when in need, while economic rights arguments claim that access to basic economic resources should have the same standing as citizenship rights” (Blank 1997, 193).
We leverage the decentralization of US safety net provision to assess the degree of cross-state inequality in provision. We begin by examining the magnitude of cross-state variation in the generosity and inclusiveness of assistance provided through safety net programs that differ in the extent of state discretion for financing, rule making, or administration. To examine this, we identify 10 primary federal-state safety net programs and categorize the extent of state discretion created through policy design in three domains: (1) financial, joint federal-state funding arrangements, partial state funding for programs (state supplements, etc.), or state discretion in spending federal funds; (2) rule-making authority, authority to determine rules regarding eligibility, benefits, and other aspects of the program (conditions of receipt, etc.); and (3) administration, flexibility and discretion in the implementation, management, and frontline delivery of assistance.
We draw on these categorizations of levels of state discretion to formulate two expectations. First, we expect less inequality in the generosity of benefits in programs that are primarily federally funded and correspondingly greater inequality in those programs in which states have more responsibility for financing and exercise more discretion in setting benefit levels. This expectation is drawn from the cross-national literature on fiscal federalism, which posits a general trade-off between subnational financial responsibility, inequities in the redistributive capacities of subnational governments, and uniform social welfare benefits (Broadway and Shah 2011). Second, we expect that state inequality in inclusiveness will be highest in programs for which states claim high levels of both rule-making authority and administrative flexibility. Researchers document substantial geographic variation in programs over which states have authority to set eligibility rules, benefits, and conditions of receipt, which are the primary mechanisms used to control program access (e.g., Fender, McKernan, and Bernstein 2002; Soss et al. 2011; Lobao et al. 2012; Schott, Pavetti, and Floyd 2015). Additional variation in access is introduced by localized administration and management; states vary, for example, in the vigor of their outreach efforts, ease of application procedures, and stringency in enforcing conditions of eligibility, often strategically managing claims for assistance in order to minimize costs to the state and maximize federal dollars received by state residents (Beamer 1999; Keiser 2001; Wolfe and Scrivner 2005; Soss and Keiser 2006; Nicholson-Crotty 2007; Ratcliffe, McKernan, and Finegold 2008; Miller and Keiser 2013).
Policy Convergence in the Post–Welfare Reform Era
The concern about unequal responses to citizen needs was heightened by the significant changes made to safety net policies in the welfare reform era of the 1990s. The most substantial change to the safety net available to low-income families with children in this time period was the passage of a historic welfare reform bill (PRWORA), which President Bill Clinton famously declared “ended welfare as we know it.” This welfare reform legislation eliminated a federal entitlement (AFDC) and created a conditional cash assistance program (TANF), in which state governments had discretion in financing, administration, and rule-making. Touted by some as the devolution revolution, in which increased authority of state and local officials would allow for the flourishing of what Justice Brandeis called laboratories of democracy (Pierson 1995; Volden 2006), others described the changes as a more complicated form of “load shifting,” through which federal authorities increased the responsibility of state and local governments while retaining authority to determine the metrics against which policy outcomes are measured (Peck 2002; Holzinger and Knill 2005; Obinger et al. 2005; Terman 2015). PRWORA in this regard was a mix—it increased state discretion in some programs (e.g., cash assistance, child care, targeted work assistance) but also specified additional federal regulations, guidelines, and incentives for states that reduced state discretion (e.g., child support enforcement, aid to immigrants and the able-bodied without children).
Twenty years after the historic welfare reform of the mid-1990s, safety net provision remains structured by a variety of negotiated settlements between local, state, and federal governments as to levels of rule-making authority, administration, and financial responsibility across programs. Whether these shifts in federal-state relations led to increasing variation in social provision across the states, a convergence as states learned and responded to similar economic conditions, or the persistence of initial state differences over time is an outstanding empirical matter. Therefore, we examine whether states have pulled closer together or drifted further apart in the inclusiveness and generosity of their administration of 10 programs in the post–welfare reform era.
Within a set of related literatures on policy adoption and diffusion (Berry and Berry 1990; Dobbin, Simmons, and Garrett 2007), policy convergence (Heichel, Pape, and Sommerer 2005; Knill 2005; Starke, Obinger, and Castles 2008), and policy retrenchment (Pierson 1994; Béland and Vergniolle de Chantal 2004; Obinger et al. 2005), there are several competing expectations regarding the trends in the degree of cross-state variation in social provision. Drawing on what is known about the factors and conditions under which states adopt similar or divergent policies, we might expect convergence over time as states adopt more similar policies as a result of learning about policy successes (Volden 2006; Volden, Ting, and Carpenter 2008) or respond to similar economic cycles and federal policy changes (Shipan and Volden 2012). We might also expect convergence as states compete in a “race to the bottom” in the generosity of benefits that are seen to attract less productive, more dependent individuals to their state (Peterson and Rom 1990; Schram and Soss 1998; Volden 2002; Bailey and Rom 2004).
A mixed expectation of both divergence and convergence can be derived from theories of federalism that point to the ongoing, strategic competition for policy control within multilevel governance systems (Mashaw and Calsyn 1996; Obinger et al. 2005). As Béland and Vergniolle de Chantal (2004) argue, in the intrastate structure of US federalism, states attempt to influence policy both locally, through state legislative and administrative action, and nationally, as institutionalized interest groups seeking to advance their regional and ideological interests. The PRWORA reflects a particularly interesting settlement of political demands for increasing the so-called personal responsibility of individuals while reducing the power of the federal government to specify the terms for assistance. The legislation increased state control over policy in some programs, most significantly by replacing individual entitlements with state-controlled block grants, while also imposing more centralized control in other areas, such as mandatory state enforcement of private child care obligations for recipients of cash assistance. Over time, it is possible that the welfare reforms sparked both convergence and divergence in state provisions, depending on the specific rebalancing of centralized mandates and decentralized discretion within each affected program.
Finally, institutional theory suggests an expectation of consistency over time in the degree of cross-state inequality, assuming path dependence and feed-forward effects through the influence of current policy on the organization of bureaucratic capacity and political interests (Pierson 1994, 2000). Although US states share a common institutional structure for safety net programs, Schneider (2012, 195) observes that policy design and action at the state level does create distinctive “public policy cultures that cohere around certain themes and endure for many generations” and influence subsequent policy decisions. Absent major policy shocks that change the balance of federal and state control over financing, rule-making, or administration, we expect to observe consistency in state responses to new economic and policy conditions. Although the overall levels of assistance may rise or fall over time, unless state-level policy discretion is substantially increased or curtailed, the degree of variation across the states is unlikely to change.
Data and Measures
We use the State Safety Net Policy (SSNP) data set, a unique data set that the authors have assembled from publicly accessible state and federal administrative records, secondary sources of these records, and original population estimates calculated using the Annual Social and Economic Supplement of the Current Population Survey. These data include 10 federal-state programs for low-income or unemployed working-age adults and their families: cash assistance (AFDC/TANF), food assistance (Food Stamps/SNAP), child health insurance (Medicaid and CHIP), child support enforcement, child care subsidies (CCBG/CCDF and TANF), early childhood education (Head Start and state pre-K programs), Unemployment Insurance (UI), targeted work assistance through AFDC/TANF, child disability assistance (SSI), and state income taxes for families at the poverty line. We use two criteria in selecting programs for inclusion in the SSNP data set. First, we focus on programs in which the state has discretion in financing, administration, or rule-making. Second, we select programs that influence the economic resources directly (by providing cash) or indirectly (by providing other goods or services) to low-income or unemployed working-age adults and their dependents.
For each type of assistance, generosity is calculated by dividing total benefit spending (federal, state, or both, as appropriate) by a state’s caseload or number of recipients.1 The generosity measures are adjusted to constant (2012) dollars using the Bureau of Labor Statistics Consumer Price Index Research Series (CPI-U-RS). Inclusion is calculated by dividing the number of actual program recipients in a state by the number of potentially needy individuals or families in the state. For means-tested programs, the estimate of the potentially needy is the number of individuals or families who (a) fall into categorically eligible groups and (b) have market (or pretransfer and tax) incomes below the federal poverty threshold or below some percentage of the threshold depending on the income eligibility criteria of the program (estimated using 3-year moving averages from the Annual Social and Economic Supplement of the Current Population Survey).2 Table 1 provides a description of the construction of each policy indicator, including specific data sources for each policy indicator. Generosity of benefits and inclusiveness of receipt are calculated for each type of assistance for all 50 states for 1994–2014.3 The generosity and inclusion policy indicators are smoothed using 3-year moving averages to reduce the year-to-year fluctuations and top and bottom coded at 2 standard deviations above and below the 50-state mean.
|Generosity||From 1994 to 1996, average yearly cash benefit in AFDC. From 1997 to 2014, calculated as state and federal dollars spent on cash benefits in TANF programa divided by the monthly average number of recipient families.b|
|Inclusion||From 1994 to 1996, numerator is monthly average number of families receiving AFDC.c From 1997 to 2014, numerator is monthly average number of families receiving TANF.b Denominator is number of pretax and transfer poor families with children (at 100% FPL).|
|Generosity||Child support distributions per child support case in which a collection was made on an obligation.d|
|Inclusion||Number of child support cases for which a collection was made on an obligationd divided by the number of single-parent families with children.|
|Generosity||Expenditures on benefits divided by the number of participating households.e|
|Inclusion||Number of households with children participatingf divided by the number of pretax and transfer poor families with children (130% FPL).|
|Generosity||Average weekly benefit received multiplied by the average number of weeks of receipt.g|
|Inclusion||Number of recipients in all programs divided by the total number of unemployed.g|
|Supplemental Security Income:|
|Generosity||Average yearly child disability benefit received (includes federally administered state supplementation payments).h|
|Inclusion||Number of children <18 receiving SSIh divided by the number of pretax and transfer poor children <18 (200% FPL).|
|State income tax:|
|Generosity||State income tax that a single-parent family of three pays when their income is at the poverty line.i|
|Inclusion||Proportion of poor single-parent families of three (100% FPL) under state income tax threshold for single-parent family of three.i|
|Preschool and early education:|
|Generosity||Federal and state expenditures on Head Start and state pre-K divided by the number of children enrolled in Head Start and state pre-K.j|
|Inclusion||Children enrolled in state pre-K and Head Start divided by the number of children 3–4 years old.j|
|Targeted work assistance:|
|Generosity||Federal and state expenditures on work-related activities including transportation divided by the number of participating families.k|
|Inclusion||From 1994 to 1996, number of JOBS participants divided by average number of families receiving AFDC. From 1997 to 2013, number of families meeting work requirements divided by average number of families receiving TANF.l|
|Child health insurance:|
|Generosity||Federal and state expenditures on Medicaid child eligibles (1994–98), beneficiaries (1999–2012), and SCHIP enrollees divided by the number of Medicaid child eligibles (1994–98), beneficiaries (1999–2012), and SCHIP-enrolled children.m|
|Inclusion||Medicaid eligibles (1994–98), beneficiaries (1999–2012), and SCHIP-enrolled childrenn divided by the under-18 pretax and transfer poor population (300% FPL).|
|Generosity||Total spending (CCDF and TANF) on child care per child served by TANF and CCDF.o|
|Inclusion||Number of children served by TANF and CCDFn divided by the number of pretax and transfer poor children under 13 (100% FPL).|
These data have several unique strengths that allow us to build on previous scholarship on geographic inequalities in social provision. First, although several scholars have provided detailed accounts of individual program variation (Campbell 2014; Schott et al. 2015; Hahn et al. 2017) or different aspects of social provision (Allard 2008; Newman and O’Brien 2011), we examine the decentralized safety net more comprehensively. These data provide the only comparable measures of safety net provisions across multiple programs of the decentralized safety net. Second, these data measure policy as a net output of federal and state government actions, including policy choices, funding levels, and administrative and management practices. They also measure this output as it is delivered or available to individuals and families within each state. Unadjusted measures of state expenditures and program caseloads, used in many policy studies, may be reasonably comparable. But without adjustments for the number of people served or the level of underlying need in the state, such measures do not capture either the level of state safety net provision, relative to need in that state, or how the safety net is experienced on average by individuals and households in need (i.e., their chances of getting assistance and the dollar value of that assistance). Third, these data provide comparable yearly estimates of program generosity and inclusion from 1994 to 2014, allowing us to map policy trajectories across periods of dramatic policy and economic changes. This allows us to conduct one of the only longitudinal studies of multiple programs that make up the safety net in each state.
While these data have several strengths, they also have limitations. The 10 programs included in the SSNP data set are the primary sources of support for most low-income or unemployed working-age adults and their dependents but are not exhaustive of all programs. For example, because of data limitations, the SSNP data set does not include a handful of other programs that are available to this population, including the Low Income Energy Assistance Program (LIHEAP), the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC), or subsidized school meal programs. Second, as measures of the output of government policies and programs, the indicators capture both factors that state policy makers can control through explicit policy choices and administrative decisions as well as factors outside their control such as market prices for services and individual behavioral responses to program rules. Third, during the time period under investigation, various safety net programs were revised, eliminated, or replaced, and data availability and reporting formats changed. This makes rendering exact over-time comparisons difficult; however, we have attempted to create measures that are as comparable as possible given program and data changes (see table 1 for descriptions of each program indicator measure, including changes in programs and data reporting or sources).
To answer our first research question concerning the extent of variation in social safety net provision across the US states, we estimate several measures of variation and dispersion. To describe the magnitude of difference across states in a readily interpretable metric, we provide the absolute values observed at different points in the distribution of states (90th and 10th percentiles). To estimate the level of cross-state variation or inequality, we estimate the range (using the 90/10 ratio), the variance (using the standard deviation), and the coefficient of variation (COV), which is calculated as the mean divided by the standard deviation. We rely most heavily on the COV because it is a standardized measure of dispersion or variance, and it is the most widely used measure in the study of policy convergence (Heichel et al. 2005; Knill 2005). Although the COV is the most commonly used measure, it does have an important limitation. Values of the COV are affected by the underlying metric of the data. Because the COV is calculated as deviations from the mean, the size or magnitude of the COV is affected by the size or magnitude of the mean. Therefore, one can compare COVs that are based on the same underlying metric (in our case spending per recipient or proportion of potentially eligible receiving benefits) but not across metrics.
In order to assess the correspondence between the extent of state discretion and the magnitude of cross-state variation, we categorize each program as providing high, medium, or low levels of control (see table 2). Although states exercise discretion in each of the programs, we are interested in differences in the extent of discretion associated with the particular balance of federal and state authority embedded in the institutional program design.
|State income tax||High||High||High|
|Targeted work assistance||High||High||High|
|Child health insurance||Medium||Medium||Medium|
|Supplemental Security Income||Low||Low||Low|
For benefit financing, discretion is coded as low if federal funds or services are provided as categorical entitlements to individuals on the basis of federal eligibility and benefit rules (with opportunities in some programs for states to supplement benefits).4 Discretion is considered medium if federal funds are provided through formula-based matching funds or block grants that impose specific rules and limitations on the use of funds or if funds are obtained through federally mandated collection of payments from individuals or employers based on state specific formulas. And discretion is coded as high when benefits are fully funded at the state level or are funded through federal block grants that require matching funds and give states broad authority over the use of those funds, including transferring funds to other programs or purposes.
In program rule-making, discretion is coded as low in programs with standardized federal rules for coverage, eligibility, service type, and other program elements (with opportunities in some programs to apply for waivers of federal rules). Discretion is coded as medium if the program is governed by federal mandates, performance standards, or broad policies but states determine specific policies for eligibility thresholds, benefit levels and duration, conditions of receipt and termination, and other policies and procedures for the treatment of individuals claims. And discretion is considered high if states set all or most specific policies governing eligibility, benefit receipt, and termination.
In program administration, discretion is coded as low if programs are administered through federal agencies at the subnational level or if state-designed systems are governed by specific federal rules and subject to regular oversight and monitoring of operations. Discretion is considered medium if states develop their own administrative systems for providing or contracting for assistance, subject to federal performance standards and monitoring of compliance. And discretion is coded as high if states design and manage their own systems within very broad federal guidelines with minimal review or direct monitoring of operations at the local levels.
To answer our second research question, we use 20 years of data to compare the trajectories of change in cross-state variation across programs. The analysis of change over time examines two aspects of convergence: the degree or magnitude of change, observed as change in variation, and the location of change, observed by examining change at different points in the distribution (Heichel et al. 2005; Holzinger and Knill 2005). The degree of convergence is assessed by comparing changes in the COV from 1994 to 2014. To determine when we should interpret the changes in the degree of variation as substantively meaningful, we statistically test the difference in COVs. Although there is no standard statistical test for differences in measures of variation such as the COV, we use a bootstrap method that generates a sample of COVs for each yearly comparison and estimate the probability that the observed difference is random.5 To further assist in the interpretation of the changes in the COV, we also estimate significant changes over time in the two components of the COV (the mean and variance/standard deviation). To test for significant changes in the mean, we use t-tests, and to test for significant changes in the variance we use the Levene test. Examining all three together (COV, variance, and mean) provides insight into why the COV is increasing, decreasing, or remaining stable over time (e.g., the mean and variance may both increase, which would not lead to an increase in the COV). The location of convergence is assessed by comparing the values at the 10th and 90th percentiles, which allows us to identify whether there is evidence of states at low levels of provision catching up with others or if states at high levels of provision are reducing their levels of provision more than other states.
Extent of Cross-State Inequality in Safety Net Provision
A comparison of cross-state variation in safety net provision across the 10 programs reveals that programs differ in the extent of variation and whether states vary in generosity, inclusion, or both (see table 3). As we expected, in 2014 we observe greater variation in the generosity of benefits in those programs over which states have greater financing responsibility and control (see fig. 1). The COV measures are largest—ranging between 0.40 and 0.87—in the three programs that have high levels of state control over funding (cash assistance, state income taxes, and targeted work assistance). For example, state income taxes (COV = 0.71) are financed entirely at the state level and reflect state policy choices regarding refundable tax credits for low-income families (e.g., state EITC) and minimum thresholds for tax liability.6 Cash assistance (COV = 0.40) and targeted work assistance (COV = 0.87) are financed through the federal TANF block grant that gives states large discretion over how much to spend overall and the share of funding that is dedicated to recipient benefits or services. The COV ranges between 0.18 and 0.26 in five programs that have medium levels of discretion and responsibility for financing. Employment-contingent subsidized child care (COV = 0.27) is funded with a mix of block grant funds, state matching funds, and supplements that allows states considerable latitude in both total and per-recipient expenditures. Variation is also in the middle range for UI (COV = 0.24) and child support benefits (COV = 0.18), in which states are mandated to collect and distribute assistance within broad federal rules, and in child health programs through Medicaid and CHIP (COV = 0.24) that are financed with a mix of state and federal matching funds. In contrast, the two programs with the least cross-state variation are largely or entirely federally funded, leaving states with limited discretion for determining total spending or individual benefit levels: food assistance (COV = 0.10) and SSI (COV = 0.03).7
|Mean||SD||COV||Range (90/10)||10th Percentile||90th Percentile|
|Percentage Δ 1994–2014||−36||−30||8||<−1||−37||−38|
|Targeted work assistance:a|
|Percentage Δ 1994–2014||444||926||89||391||29||720|
|Percentage Δ 1994–2014||1||2||0||<−1||<1||<1|
|Percentage Δ 1994–2014||27||<−1||−20||−14||30||11|
|Supplemental Security Income:|
|Percentage Δ 1994–2014||−3||−35||−40||4||−3||−8|
|State income taxes:|
|Percentage Δ 1994–2014||…b||97||34||30||68||7,738|
|Percentage Δ 1994–2014||46||54||4||7||35||45|
|Percentage Δ 1994–2014||63||40||−16||−5||66||54|
|Percentage Δ 1994–2014||−32||−42||−14||−11||−23||−31|
|Child health insurance:a|
|Percentage Δ 1994–2014||84||68||−11||−6||93||78|
|Percentage Δ 1994–2014||−67||−25||125||210||−84||−53|
|Targeted work assistance:a|
|Percentage Δ 1994–2014||13||0||−21||−19||13||7|
|Percentage Δ 1994–2014||48||56||0||−3||48||46|
|Percentage Δ 1994–2014||3||0||−3||7||5||−4|
|Supplemental Security Income:|
|Percentage Δ 1994–2014||33||0||−6||3||0||25|
|State income taxes:|
|Percentage Δ 1994–2014||21||−8||−29||−24||26||2|
|Percentage Δ 1994–2014||53||117||53||130||−13||87|
|Percentage Δ 1994–2014||−15||14||36||19||−20||−3|
|Percentage Δ 1994–2014||85||39||−26||−29||138||68|
|Child health insurance:a|
|Percentage Δ 1994–2014||85||30||−29||−13||100||72|
The extent of cross-state variation by program also conforms to our second expectation, that variation in inclusiveness would be greatest in programs over which states exercise greater discretion in rule-making and administration (see fig. 2). Five of the 10 programs are characterized by high levels of both rule-making authority and administrative flexibility; these programs are also among the most variable across states: cash assistance (COV = 0.63), preschool/early education (COV = 0.58), child care (COV = 0.49), and targeted work assistance (COV = 0.37). High levels of state variation in the TANF-related programs is not surprising given the explicit devolution of authority to set eligibility criteria and rules in the TANF block grant (Schott et al. 2015). The variation in inclusiveness of preschool/early education programs reflects the combination of Head Start, a federally administered program, and state initiated and managed pre-K programs, which vary dramatically across states (Barnett et al. 2015). In contrast, the programs with the least variation in the inclusiveness of receipt—food assistance (COV = 0.15) and health insurance (COV = 0.15)—are both subject to standard federal eligibility criteria and require states to seek waivers for significant deviations from these criteria, and they are also subject to direct federal oversight and monitoring.8 Variation in the children’s SSI program is also substantial (COV = 0.34) despite the direct administration of the program by federal authorities. One possible explanation for this is variation in state effort to increase participation in the federally funded program, which has been associated with the aggressiveness of state TANF reforms (Schmidt and Sevak 2004), state revenue and expenditure changes (Kubik 2003), and geographic region (ASPE 2015).
Taken together, these findings reveal substantial cross-state variation in safety net provision, resulting in highly unequal access and benefits provided through the same programs in different states. Direct federal funding and nationally uniform eligibility criteria appear to reduce geographic inequality in state provision. Even in programs with consistent federal rules, however, state administrative actions appear to introduce variation in treatment, particularly in access to benefits. The weaker the federal role, the further apart are the states with respect to both the share of the needy they help and the level of assistance they provide.
Magnitude of Cross-State Inequality of Safety Net Provision
To give substantive meaning to the extent of variation measured by the COV, we also examine the absolute differences in the value of benefits and share of the potentially needy served in higher- and lower-provision states (see table 3). We find meaningful levels of geographic inequality in the generosity of benefits. For example, a poor family receiving cash assistance in 2014 in a state near the 10th percentile receives an average benefit of $1,957 (in 2012 dollars); a similarly poor family in a state near the 90th percentile receives an average benefit of $5,811—a $3,854 or 66 percent difference. In states with an income tax, to take another example, a one-parent family of three with poverty-level income would receive a $1,019 tax refund in the state around the 90th percentile, due to a progressive tax schedule and targeted benefits; a similar taxpayer would face a $131 liability in the state at the 10th percentile.9 In fact, in eight of the 10 programs the difference in average benefits between low- and high-provision states is more than $1,000, which is nearly 10 percent of the federal poverty threshold for a single-person household.
Inequalities between states are even more pronounced in the inclusiveness of social safety net programs. The inclusion measures control for level of need within each state by calculating recipients as a share of the relevant poor (or unemployed) population. Although targeted on the neediest, most programs serve only a fraction of those at risk. In seven of the 10 programs, the average rate of inclusion is less than half in 2014, and even states at the 90th percentile of inclusiveness served fewer than two-thirds of those in need. Only two programs—food assistance and children’s health insurance—effectively reached not only those in poverty but a share of those over the FPL.10 With the exception of these two relatively expansive programs, levels of inclusion are generally low and vary by 50 percent or more between the more and less inclusive states. In cash assistance, for example, the average inclusion is just under 20 percent—or two out of 10 poor families with children—across all states. But states near the 90th percentile reach about one in three such families (respectively), whereas those near the 10th percentile reach fewer than one in 10. The primary alternative form of cash assistance, UI, reaches only about one out of every three unemployed adults nationwide, because of restrictive coverage and eligibility rules. In states near the 90th percentile, however, the rate is as high as one out of two, and in states near the 10th percentile it is as low as one out of four. These differences create geographic inequalities in the treatment of similar claimants and, by allowing some states to provide very low benefits to a very small fraction of the needy, exacerbate the weakness of the safety net as a whole.
Convergence, Divergence, or Stasis in State Provisions
Turning to our second research question, we examine changes in the extent of cross-state inequality since the mid-1990s using the COV. Overall, we find little evidence to support a “race to the bottom” in state safety net provisions. Instead, the majority of programs and measures conform to the predictions of institutional stability over time in state approaches. Levels of generosity and inclusion changed significantly in nearly all of the programs. When the change in dispersion (standard deviation) is standardized for change in levels (mean) using the COV, however, we see considerable consistency in the extent of state-to-state variation (see table 3). Significant changes in the COV are observed in the generosity of benefits for only two programs and for inclusiveness in a somewhat larger group of five programs.
At both the beginning and the end of the period, states were relatively tightly clustered in measures of generosity with COV values in the 0.12 to 0.35 range for most programs (see fig. 3). The COV for generosity changed significantly in only two programs over the total period: states pulled much further apart in spending per participant in targeted work assistance for TANF recipients, and states pulled somewhat closer together in the average benefit received by disabled children in the SSI program.
Inequality was substantially higher in the inclusiveness of safety net programs than in the generosity of benefits at both points in time, and there were more marked changes in variation in the inclusiveness of provisions over time (see fig. 4). States diverged to a significant degree in the inclusiveness of three programs: cash assistance, preschool/early education, and child care. They pulled closer together in two: child support collections and child health insurance.
Consistent with the expectations suggested by the changes in federal-state responsibilities in the PRWORA legislation, three of the programs in which we observe substantive divergence in the magnitude of cross-state variation—cash assistance and child care (in inclusion) and targeted work assistance (in generosity)—were directly affected by the welfare reforms of the 1990s that granted states greater flexibility. However, the location of change within the total distribution of state efforts varied. While states pulled further apart on the generosity of targeted employment assistance because of very large increases in spending in a few states, the divergence in the inclusiveness of cash assistance and child care was driven largely by especially steep reductions in states that began the period with low levels of provision. The case of cash assistance is particularly dramatic: states at the 10th percentile in 1994 provided assistance to 38 out of 100 poor families with children but only to six out of 100 poor families in 2014. Divergence is also observed in the inclusion of children in preschool/early education, but it resulted from nearly the opposite change with low-provision states contracting slightly while those near the 90th percentile nearly doubled the share of children served.
It is equally notable that one of the two programs for which we observe significant convergence across the states—child support inclusion—was also addressed in the PRWORA legislation. In this case, rather than increasing state flexibility, federal lawmakers increased expectations for state performance, along with administrative funds for meeting new standards. In response, all states increased their inclusiveness in child support collections. This increase is most dramatic in states near the bottom of the distribution, leading to a convergence in the extent of cross-state inequality. A similar pattern is seen in the inclusiveness of child health insurance. The creation of CHIP soon after the passage of the PRWORA, with more generous federal cost shares and higher eligibility thresholds, both mandated and incentivized greater inclusion in state-run programs. This expansion of federal involvement in children’s health care corresponds to the substantial increases in inclusiveness observed across the board, with larger increases for states at the bottom of the distribution and a reduction in cross-state inequality in provision.
Despite significant changes in levels of provision between 1994 and 2014, as measured by the generosity of benefits and inclusion of the needy, the extent of state-to-state variation did not change significantly on most measures. Although the states were doing more, or less, as a whole, at the end of the period they generally increased or decreased provisions at similar enough levels to maintain the extent of cross-state variation observed in 1994. Most cases in which states were observed to pull further apart or closer together were in programs directly affected by federal legislation in the mid- to late 1990s, including a widening gap in the inclusiveness of programs (cash assistance and child care) in which the conversion from individual entitlements to block grants by the PRWORA increased state discretion and a narrowing of interstate variation in programs for which federal actions mandated (child support collections) or incentivized (child health insurance) greater inclusion of the needy.
The decentralized structure of the safety net is one of most crucial yet least carefully studied structural design features of the US welfare state, and it has dramatic consequences in terms of inequalities in social provision across the states. Using state-level measures to examine geographic inequality in safety net programs, we shed new light on the potential consequences of the decentralized structure of assistance for working-age adults and families.
The most striking finding of our analysis is the extent and persistence of geographic inequality. Scholars have long observed that inequality is an inevitable outcome of a federalist system, especially in the absence of fiscal redistribution. But the extent of inequality in the US safety net has rarely been assessed across the weakly coordinated system of numerous separate programs that make up the American welfare state. When we undertake such an assessment using state-level measures of generosity and inclusion, we find that the magnitude of cross-state inequality corresponds closely to the level of state discretion in financing, rule-making, and administration. The magnitude of inequality in provision, using measures that control for underlying need, suggests unequal treatment of individuals and households with similar needs who live in different jurisdictions. The highest levels of inequality are observed in those programs for which states have the highest level of financial responsibility and greater rule-making and administrative autonomy, especially in regard to the inclusiveness of program receipt.
The implication of these findings is that designing policies with state discretion in financing, rule-making, or administration is likely to lead to greater levels of cross-state inequality in provision than a design in which state discretion is limited. Recent work examining state policy choices and social welfare spending also highlights potentially negative consequences of allowing state discretion. For example, recent work on state spending of the TANF block grant finds that 10 states spend less than 10 percent of their TANF block grant on basic assistance (Schott et al. 2015), which likely substantially weakens TANF’s role in the safety net (Floyd, Pavetti, and Schott 2017). Building on a long line of scholarship, Soss and colleagues also demonstrate the significant role that race plays in state social welfare policy choices, with states with a greater representation of African Americans being more likely to adopt paternalistic policy designs that include more punitive sanctions and more restrictive and invasive conditions on the receipt of social welfare benefits (Soss et al. 2011).
The consequences of devolution can also be seen over time. While we find several cases of changes in the extent of cross-state inequality, the vast majority of programs can be characterized as having relatively stable levels of cross-state inequality in provision. This is likely a result of the substantial path dependence or feed-forward effects of the initial policy designs that established particular federal-state arrangements in terms of responsibility for financing, rule making, and administration.
However, the most notable findings regarding change over time in cross-state inequality in provision is that the change in federal-state relations resulting from PRWORA increased cross-state inequality in three of the programs funded in part by this block grant (cash assistance, targeted work assistance, and child care). This dynamic is underscored by a rare exception: in the one program, child support, in which the PRWORA imposed new and more stringent federal requirements, state outcomes converged.
These findings of policy stability with select cases of convergence and divergence have several implications for the field of social policy and the well-being of families with children. First, while these patterns of over-time, cross-state inequality are likely a result of a number of economic and political factors, we demonstrate that policy convergence and divergence are also related to how a policy structures the federal-state relationship. In other words, while political ideology or economic conditions may influence the policy diffusion and adoption process, attention must also be paid to how the policy itself is structured. Second, given the magnitude and general stability of between-state inequalities in provision, any change in the policy environment that is intended to reduce such inequality would need to include changing the level of state responsibility for these programs. In fact, even the most optimistic observers of post–civil rights era federalism concede that state discretion—in the absence of high federal standards and serious benchmarking of outcomes—is likely to do more damage than good (Freeman and Rogers 2007). Indeed, even some champions of the 1996 reforms have beaten a retreat—expressing surprise or dismay at the ability and willingness of states (especially in the case of cash assistance) to eviscerate rather than innovate (Haskins 2016). Given the active role taken by many city and local governments to implement dramatically more progressive policies such as the $15 minimum wage, state actions ranging from preemption laws to drastic cuts in all state spending, and federal proposals to block grant Medicaid or other safety net programs, questions of state (and local) discretion and the relations between federal and state governments are centrally implicated in social welfare policy debates and should likewise be examined as centrally important aspects of social welfare policy research.
Appendix. Consequences of Decentralization
Pretax and Transfer or Market Income
Pretax and transfer or market income uses the following income components: wage and salary; self-employment; farm; interest; dividends; rents, royalties, estate, and trust income; alimony; private and occupationally based retirement, survivors’, and disability pensions (not including Social Security, Veteran’s Affairs benefits, or Workers Compensation); financial assistance from friend/family; and income reported in the “other income” category that was one of the previous types of income. Pretax and transfer poverty (market income) is used as the income measure for determining potential need for assistance in order to capture the income resources available before any direct transfers or taxes. This differs from the official poverty measure, which includes cash transfers. It also differs from the poverty or income guidelines used by many government programs to determine eligibility, which are based on the official poverty thresholds but differ in a number of ways including in income level, the way assets are counted, and the time period considered (see
Population Denominator Estimates
The population denominators are not calculated to precisely estimate the population that would qualify for each program in each state. Instead, they provide the most valid basis for comparing the extent to which social supports reach economically needy, program-relevant populations in different states and years. We use measures of the categorically eligible population with income below the poverty threshold as opposed to estimating the more narrow potentially eligible population based on specific program eligibility rules in order to assess how deep the receipt is into the economically needy population. We are not testing whether states are meeting their own eligibility criteria because they could set very low eligibility thresholds or very restrictive eligibility criteria and serve all those families, but that would not be an accurate measure of serving families in need. The Urban Institute TRIM program can be used along with Current Population Survey data to more accurately estimate potentially eligible populations for various programs based on state-specific eligibility rules. We do not use this approach because we are interested in the potentially needy population and what proportion of this group is assisted. This approach allows for better comparability over time within programs such that, even as the program eligibility rules change, the measure of the potentially needy stays the same.
To assess the extent to which the adequacy measures are associated with the cost-of-living differences across states, we use the state-level, all items Regional Price Parity (RPP) index created by the Bureau of Economic Analysis (Aten, Figueroa, and Martin 2012). Specifically, we apply the RPP to the final adequacy measure for 2008–14.
There are not large differences in means or the range of values from the 90th to 10th percentiles for any of the programs. Using adjusted measures actually increases the cross-state variation in three programs: food assistance (COV = 0.15), SSI (COV = 0.08), and targeted work assistance (COV = 0.91). Adjusted measures results in slightly lower levels of cross-state variation in cash assistance (COV = 0.34), UI (COV = 0.20), state income tax (COV = 0.67), preschool and early education (COV = 0.25), child support (COV = 0.15), and housing assistance (COV = 0.20). None of the reductions in the magnitude of variation are 0.10 or greater (the criteria we use to establish a meaningful difference in COVs). In fact there are only two changes that are above 0.05 (cash and housing assistance). There is no change in the magnitude of variation in child health insurance.
1. We use the yearly total number of recipients or caseloads when available. When not available, we use monthly average caseloads.
2. Pretax and transfer or market income uses the following income components: wage and salary; self-employment; farm; interest; dividends; rents, royalties, estate, and trust income; alimony; private and occupationally based retirement, survivors’, and disability pensions (not including Social Security, Veteran’s Affairs benefits, or Workers Compensation); financial assistance from friend/family; and income reported in the “other income” category that was one of the previous types of income. Pretax and transfer poverty (market income) is used as the income measure for determining potential need for assistance, in order to capture the income resources available before any direct transfers or taxes. This differs from the official poverty measure (which includes cash transfers) and the poverty or income guidelines used by many government programs to determine eligibility (which are based on the official poverty thresholds but differ in a number of ways including income level, the way assets are counted, and the time period considered). Counts of the number of families or children falling below the poverty threshold in a given year are weighted to the state level using the person weight assigned to the family or household head. We use measures of the categorically eligible population with income below the poverty threshold as opposed to estimating the more narrow potentially eligible population based on specific program eligibility rules in order to assess how deep the receipt is into the economically needy population. We are not assessing whether states are meeting their own eligibility criteria. States could set very high eligibility thresholds or have other restrictive eligibility criteria and serve 100 percent of those families; however, that would not be an accurate measure of the reach the program has into the population of families or children in need. The Urban Institute Transfer Income Model (TRIM) program can be used along with Current Population Survey data to more accurately estimate potentially eligible populations for various programs on the basis of state-specific eligibility rules. We do not use this approach because we are interested in the potentially needy population and what proportion of this group is assisted. This approach allows for better comparability over time within programs such that, even as the program eligibility rules change, the measure of the potentially needy stays the same. An additional benefit of using this approach is that it allows us to have a more consistent population across programs, so we can examine whether a larger proportion of this group receives one kind of assistance or another.
3. Child-care indicators are available starting in 1998. The most recent year of data for the health insurance indicators is 2012, and it is 2013 for targeted work assistance.
4. In the case of direct services, including health insurance and child care, variation in local market prices may also cause the generosity of benefits to vary across states even when state discretion over financing is limited.
5. In the bootstrapping process, we resample pairs of observations by state (as opposed to resampling based on year values). This leads to much lower variation in the bootstrap estimates because state values are highly correlated over years. To determine when a change in COV is significant statistically, we rely on a bootstrapping estimation procedure that tests the difference in COVs across 2 years using a cutoff of p < .05. A bootstrapping method similar to what we employ here is used by Kenworthy (1999).
6. Average tax liabilities at the poverty line are reverse coded to capture state tax benefits. States with no income tax—that rely on more regressive sales taxes—are not included in these measures.
7. States can supplement the SSI benefits, but currently 18 states do not supplement the federal benefit for children, and 4 states only supplement benefits for specific types of disabilities. The supplements that the remaining states do give are relatively small (Social Security Administration 2016). Albritton (1989) also found less variation in programs with more federal financing in several programs, including SSI for the aged, blind, and disabled and cash and food assistance from the late 1960s to the mid-1980s.
8. The eligibility for children in Medicaid/CHIP varies greatly in terms of the income eligibility levels. The federal government mandates coverage of children under 100 percent of the federal poverty line (FPL), and all states have chosen to expand coverage to children above this minimum. The vast majority of states have eligibility levels between 200 and 300 percent of the poverty line, and only three states fall below this threshold (Kaiser Family Foundation 2016).
9. One concern might be that these differences in the generosity of benefits are due to cost-of-living differences across states. To assess this possibility, we adjust the generosity measures using the state-specific all-items Regional Price Parity measures from the Bureau of Economic Analysis. In the case of cash assistance, using the adjusted measure results in the 10th percentile increasing to $2,111 and the 90th percentile decreasing to $5,748—a $3,637 or a 63 percent difference. As this example demonstrates, there are not large differences between the adjusted and unadjusted measures in either the range of values or the magnitude of variation for any of the programs. See the appendix for a fuller analysis and description of these differences.
10. Households with children and gross incomes up to 130 percent of the FPL are generally eligible for food assistance (SNAP) as long as they meet other resource and asset tests; therefore, the denominator for the food assistance inclusion measure is 130 percent of the FPL. States can get federal CHIP matching funds for child coverage up to 300 percent of the FPL; therefore, the denominator for the child health insurance inclusion measure is 300 percent of the FPL. Fully 46 percent of states cover children above 200 percent of the FPL (CMS 2016).
Sarah K. Bruch is an assistant professor in the Department of Sociology and director of the Social and Education Policy Research Program at the Public Policy Center at the University of Iowa. Her research focuses broadly on social stratification and public policy. In particular, she focuses on integrating theoretical insights from relational and social theorists into the empirical study of inequalities. She brings this approach to the study of social policy, education, race, politics, and citizenship.
Marcia K. Meyers is emeritus professor of social work at the University of Washington and founding director of the West Coast Poverty Center. Her scholarship examines issues of poverty and inequality, US social policy, gender, and welfare state structures in advanced capitalist nations.
Janet Gornick is professor of political science and sociology at the Graduate Center of the City University of New York. She also serves as director of the James M. and Cathleen D. Stone Center on Socio-Economic Inequality and as director of the US Office of LIS (the cross-national data archive). Most of her research is comparative, across countries or across the American states, and concerns public policies and their impact on gender disparities in the labor market and on income inequality.
Albritton, Robert B. 1989. “Impacts of Intergovernmental Financial Incentives on State Welfare Policymaking and Interstate Equity.” Publius 19 (2): 127–41. Allard, Scott. 2008. Out of Reach: Place, Poverty, and the New American Welfare State. New Haven, CT: Yale University Press. ASPE (Office of the Assistant Secretary for Planning and Evaluation). 2015. “The Child SSI Program and the Changing Safety Net.” ASPE Research Brief. Office of Human Services Policy, US Department of Health and Human Services, Washington, DC. Bailey, Michael A., and Mark Carl Rom. 2004. “A Wider Race? Interstate Competition across Health and Welfare Programs.” Journal of Politics 66 (2): 326–47. Barnett, W. Steven, Megan E. Carolan, James H. Squires, Kristy Clarke Brown, and Michelle Horowitz. 2015. “The State of Preschool 2014: State Preschool Yearbook.” National Institute for Early Education Research, New Brunswick, NJ. Beamer, Glenn. 1999. Creative Politics: Taxes and Public Goods in a Federal System. Ann Arbor: University of Michigan Press. Béland, Daniel, and François Vergniolle de Chantal. 2004. “Fighting ‘Big Government’: Frames, Federalism, and Social Policy Reform in the United States.” Canadian Journal of Sociology 29 (2): 241–64. Berry, Frances Stokes, and William D. Berry. 1990. “State Lottery Adoptions as Policy Innovations: An Event History Analysis.” American Political Science Review 84 (2): 395–415. Blank, Rebecca. 1997. It Takes a Nation: A New Agenda for Fighting Poverty. Princeton, NJ: Princeton University Press. Broadway, Robin, and Anwar Shah. 2011. Fiscal Federalism: Principles and Practice of Multiorder Governance. Cambridge: Cambridge University Press. Campbell, Andrea Louise. 2014. Trapped in America’s Safety Net: One Family’s Struggle. Chicago: University of Chicago Press. CMS (Centers for Medicare and Medicaid Services). 2016. “Eligibility.” Medicaid.gov. https://www.medicaid.gov/chip/eligibility-standards/chip-eligibility-standards.html. DeParle, Jason. 2009. “For Victims of Recession, Patchwork State Aid.” New York Times, May 9. http://www.nytimes.com/2009/05/10/us/10safetynet.html. Dobbin, Frank, Beth Simmons, and Geoffrey Garrett. 2007. “The Global Diffusion of Public Policies: Social Construction, Coercion, Competition, or Learning?” American Review of Sociology 33:449–72. Fender, Lynne, Signe-Mary McKernan, and Jenny Bernstein. 2002. “Linking State TANF and Related Policies to Outcomes: Preliminary Typologies and Analysis.” Final report prepared for US Department of Health and Human Services’ Office of the Assistant Secretary for Planning and Evaluation. Finegold, Kenneth. 2005. “The United States: Federalism and Its Counter-Factuals.” In Obinger, Castles, and Leibfried 2005, 138–78. Floyd, Ife, LaDonna Pavetti, and Liz Schott. 2017. “TANF Reaching Few Poor Families.” Center on Budget and Policy Priorities, Washington, DC. Freeman, Richard B., and Joel Rogers. 2007. “The Promise of Progressive Federalism.” 205–27 in Remaking America: Democracy and Public Policy in an Age of Inequality, edited by Joe Soss, Jacob S. Hacker, and Suzanne Mettler. New York: Russell Sage. Hahn, Heather, Eleanor Pratt, Eva H. Allen, Genevieve M. Kenney, Diane K. Levy, and Elaine Waxman. 2017. “Work Requirements in Social Safety Net Programs: A Status Report of Work Requirements in TANF, SNAP, Housing Assistance, and Medicaid.” Urban Institute, Washington, DC. Halpern-Meekin, Sarah, Kathy Edin, Laura Tach, and Jennifer Sykes. 2015. It’s Not like I’m Poor: How Working Families Make Ends Meet in a Post-Welfare World. Oakland: University of California Press. Haskins, Ron. 2016. “TANF at Age 20: Work Still Works.” Journal of Policy Analysis and Management 35 (1): 224–31. Heichel, Stephan, Jessica Pape, and Thomas Sommerer. 2005. “Is There Convergence in Convergence Research? An Overview of Empirical Studies on Policy Convergence.” Journal of European Public Policy 12 (5): 817–40. Holzinger, Katharina, and Christoph Knill. 2005. “Causes and Conditions of Cross-National Policy Convergence.” Journal of European Public Policy 12 (5): 775–96. Howard, Christopher. 1999. “The American Welfare State, or States?” Political Research Quarterly 52 (2): 421–42. Kaiser Family Foundation. 2016. “Where Are the States Today? Medicaid and CHIP Eligibility Levels for Adults, Children, and Pregnant Women.” Fact sheet. Kaiser Family Foundation, Menlo Park, CA. https://www.kff.org/medicaid/fact-sheet/where-are-states-today-medicaid-and-chip/. Keiser, Lael R. 2001. “Street-Level Bureaucrats, Administrative Power and the Manipulation of Federal Social Security Disability Programs.” State Politics and Policy Quarterly 1 (2): 144–64. Kenworthy, Lane. 1999. “Economic Integration and Convergence: A Look at the U.S. States.” Social Science Quarterly 80 (4): 858–69. Knill, Christoph. 2005. “Introduction: Cross-National Policy Convergence: Concepts, Approaches, and Explanatory Factors.” Journal of European Public Policy 12 (5): 764–74. Kubik, Jeffrey D. 2003. “Fiscal Federalism and Welfare Policy: The Role of States in the Growth of Child SSI.” National Tax Journal 56:61–79. Lobao, Linda, P. Wilner Jeanty, Mark Partridge, and David Kraybill. 2012. “Poverty and Place across the United States: Do County Governments Matter to the Distribution of Economic Disparities?” International Regional Science Review 35 (2): 158–87. Lobao, Linda, and David Kraybill. 2009. “Poverty and Local Governments: Economic Development and Community Service Provision in an Era of Decentralization.” Growth and Change 40 (3): 418–51. Marshall, T. H. 1949. “Citizenship and Social Class.” In Class, Citizenship, and Social Development: Essays by T. H. Marshall, edited by Seymour Martin Lipset. Chicago: University of Chicago Press. Mashaw, Jerry L., and Dylan S. Calsyn. 1996. “Block Grants, Entitlements, and Federalism: A Conceptual Map of Contested Terrain.” Yale Law and Policy Review 14 (2): 297–324. Meyers, Marcia K., Janet C. Gornick, and Laura R. Peck. 2001. “Packaging Support for Low-Income Families: Policy Variation across the United States.” Journal of Policy Analysis and Management 20 (3): 457–83. Miller, Susan M., and Lael R. Keiser. 2013. “State Governments as Entrepreneurs in Securing Federal Benefits for Their Citizens.” Publius 43 (4): 497–526. Newman, Katherine S., and Rourke O’Brien. 2011. Taxing the Poor: Doing Damage to the Truly Disadvantaged. Oakland: University of California Press. Nicholson-Crotty, Jill. 2007. “Politics, Policy, and the Motivations for Advocacy in Nonprofit Reproductive Health and Family Planning Providers.” Nonprofit and Voluntary Sector Quarterly 36 (1): 5–21. Obinger, Hebert, Francis G. Castles, and Stephan Leibfried, eds. 2005. Federalism and the Welfare State: New World and European Experiences. New York: Cambridge University Press. Peck, Jamie. 2002. “Political Economies of Scale: Fast Policy, Interscalar Relations, and Neoliberal Workfare.” Economic Geography 78 (3): 331–60. Peterson, Paul E., and Mark C. Rom. 1990. Welfare Magnets: A New Case for a National Standard. Washington, DC: Brookings Institution. Pierson, Paul. 1994. Dismantling the Welfare State? Reagan, Thatcher, and the Politics of Retrenchment. Cambridge: Cambridge University Press. ———. 1995. “Fragmented Welfare States: Federal Institutions and the Development of Social Policy.” Governance: An International Journal of Policy and Administration 8 (4): 449–78. ———. 2000. “Increasing Returns, Path Dependence, and the Study of Politics.” American Political Science Review 94 (2): 251–67. Ratcliffe, Caroline, Signe-Mary McKernan, and Kenneth Finegold. 2008. “Effect of Food Stamp and TANF Policies on Food Stamp Receipt.” Social Service Review 82 (2): 291–334. Schmidt, Lucie, and Purvi Sevak. 2004. “AFDC, SSI, and Welfare Reform Aggressiveness: Caseload Reductions versus Caseload Shifting.” Journal of Human Resources 39 (3): 792–812. Schneider, Anne L. 2012. “Punishment Policy in the American States from 1890 to 2008: Convergence, Divergence, Synchronous Change, and Feed-Forward Effects.” Policy Studies Journal 40 (2): 193–210. Schott, Liz, LaDonna Pavetti, and Ife Floyd. 2015. “How States Use Federal and State Funds under the TANF Block Grant.” Report. Center on Budget and Policy Priorities, Washington, DC. https://www.cbpp.org/research/family-income-support/how-states-use-federal-and-state-funds-under-the-tanf-block-grant. Schram, Sanford F., and Joe Soss. 1998. “Making Something out of Nothing: Welfare Reform and a New Race to the Bottom.” Publius 28 (3): 67–88. Shipan, Charles R., and Craig Volden. 2012. “Policy Diffusion: Seven Lessons for Scholars and Practitioners.” Public Administration Review 72 (6): 788–96. Social Security Administration. 2016. “Understanding Supplemental Security Income SSI for Children: 2016 Edition.” Social Security Administration, Baltimore. https://www.ssa.gov/ssi/text-child-ussi.htm. Soss, Joe, Richard C. Fording, and Sanford F. Schram. 2011. Disciplining the Poor: Neoliberal Paternalism and the Persistent Power of Race. Chicago: University of Chicago Press. Soss, Joe, and Lael R. Keiser. 2006. “The Political Roots of Disability Claims: How State Environments and Policies Shape Citizen Demands.” Political Research Quarterly 59 (1): 133–48. Starke, Peter, Herbert Obinger, and Francis G. Castles. 2008. “Convergence towards Where: In What Ways, If Any, Are Welfare States Becoming More Similar?” Journal of European Public Policy 15 (7): 975–1000. Terman, Jessica. 2015. “Performance Goal Achievement in Fiscal Federalism: The Influence of State Partisan Environments and Regulatory Regimes.” Policy Studies Journal 43 (3): 333–54. Volden, Craig. 2002. “The Politics of Competitive Federalism: A Race to the Bottom in Welfare Benefits?” American Journal of Political Science 46 (2): 352–63. ———. 2006. “States as Policy Laboratories: Emulating Success in the Children’s Health Insurance Program.” American Journal of Political Science 50 (2): 294–312. Volden, Craig, Michael Ting, and Daniel Carpenter. 2008. “A Formal Model of Learning and Policy Diffusion.” American Political Science Review 102 (3): 319–32. Wildavsky, Aaron. 1985. “Federalism Means Inequality.” Society 22 (2): 42–49. Wolfe, Barbara, and Scott Scrivner. 2005. “The Devils May Be in the Details: How the Characteristics of SCHIP Programs Affect Take-Up.” Journal of Policy Analysis and Management 24 (3): 499–522.