Inequality and Competition in State Redistributive Systems: Evidence From Welfare and Health

Abstract

When determining their redistributive budgets, states must strike a subtle balance—to provide for their needy residents without becoming a “welfare magnet” and attracting poor individuals from neighboring states. We examine the competing incentives that state politicians face in federal systems and their effects on program accessibility and redistributive spending across U.S. states between 2005 and 2011. Comparing two redistributive programs under state control—Medicaid and Temporary Assistance for Needy Families (TANF)—we find strong evidence of interstate competition in the case of cash assistance programs, but less evidence in the case of health care. Yet our data show that states do not alter their policies in response to rising inequality, that is, when the median voter becomes poorer than the average voter. Moreover, the Great Recession had a greater impact on TANF than Medicaid. We attribute these differential effects to different funding mechanisms used by the federal government to finance the two state-administered programs.

Keywords

race to the bottom median voter TANF medicaid great recession

Introduction

When determining their redistributive budgets, state politicians face competing incentives. On one hand, they are responsible for addressing the social needs of their states by providing public assistance to people in need. Ultimately, redistributive policies aim to decrease economic inequality by taking from the haves and giving to the have-nots. On the other hand, states seek to meet the needs of low-income populations without becoming a “welfare magnet” (Peterson, 1995) that attracts poor people from neighboring states. This article investigates these competing incentives and how they affect redistributive programs’ accessibility and spending patterns across U.S. states.

Prior research has examined whether state redistributive systems compete in a “race to the bottom” fashion but produced mixed results. Some studies have reported strong evidence of negative interstate competition (e.g., Bailey & Rom, 2004; Peterson, 1995; Rom, Peterson, & Scheve, 1998). Others (e.g., Berry, Fording, & Hanson, 2003) have registered a statistically significant but not substantial effect of peers’ welfare policies. A third group of scholars (e.g., Volden, 2002) has argued that states actually intend to increase benefits but hesitate to do so before their neighbors do. Moreover, most studies have focused on the former cash assistance program, the Aid to Families With Dependent Children (AFDC), which was abolished in 1996. Our study uses data on Temporary Assistance for Needy Families (TANF) programs, which replaced AFDC. Following Bailey and Rom (2004), we extend the analysis beyond welfare and compare the effects with another redistributive program under state control—Medicaid—which provides health care services to low-income individuals.

Specifically, our study seeks to address two interrelated research questions. First, it examines if a state’s redistributive policies reflect social conditions in this state. In other words, when inequality in a state widens, do redistributive expenditures grow as well? Are state politicians responsive to changes in the income level of the median voter, as predicted by the median voter theorem? The extant literature reports inconclusive findings on the relationship between economic inequality and redistributive spending (Chernick & Reschovsky, 2003; de Mello & Tiongson, 2006; Freitas, 2012; Karabarbounis, 2011; Mohl & Pamp, 2009; Persson & Tabellini, 1994; Rodrigìuez, 1999). One group of scholars finds support for the classic model proposed by Meltzer and Richard (1981), which relates rising inequality to greater redistribution. Another group of scholars provides strong evidence for the opposite effect: widening inequality drives government redistribution down. Second, we test Peterson’s (1995) “race to the bottom” thesis and ask if a state’s redistributive policies compete with the policies of neighboring states in terms of eligibility requirements for recipients and the proportion of state budget devoted to redistribution.

To answer these questions, we gather data on state Medicaid and TANF programs between 2005 and 2011. These data also allow us to explore the effect of the 2008 financial crisis on redistributive expenditures. The Great Recession presents an intriguing case. In times of strained public budgets, redistributive spending seems most vulnerable. Conversely, these are conditions in which poverty and inequality are likely to rise along with demand for public assistance. And indeed, in response to the economic recession of 2008, the federal government implemented various policy initiatives, chief among which were the American Recovery and Reinvestment Act (ARRA) of 2009 and the Patient Protection and Affordable Care Act (ACA) of 2010. Under ARRA, the federal and state governments increased the funding for vulnerable social groups; respectively, the ACA considerably expanded the coverage of Medicaid. Yet, did the states respond in a similar manner by increasing the amount of funding, given that they are subject of stringent balanced budget requirements and greater competitive pressures compared with the national government?

Our results show that states do compete, albeit some more than others, and generally attempt to maintain redistributive programs that are less generous than those in neighboring states. When states set eligibility rules, redistributive programs can be designed to ease or to limit access to benefits. To understand if interstate competition applies to program access, we analyze the eligibility rules for welfare across states and find that states behave in a manner consistent with the “race to the bottom” hypothesis (Peterson, 1995) for both TANF and Medicaid. The results of our regression analysis also provide support for the existence of interstate competition between states’ redistributive systems, yet the competition is more pronounced in welfare programs than in health programs. In regard to state responsiveness to social conditions, we find that policy makers do not alter redistributive policies when the median voter becomes poorer than the average voter (indicating rising inequality). Our analysis shows that the coefficient of inequality has a negative sign, suggesting an inverse relationship with redistributive spending, but the effect fails to achieve statistical significance.

The article contributes to the literature by examining and comparing two popular theories—the median voter and the race to the bottom—and testing whether they can explain why some states opt for more generous redistributive policies than others. Prior studies have only examined these two hypotheses separately. This approach allows us to test the two theories of federalism, advanced by Peterson (1995): functional federalism versus legislative federalism. We study in detail the eligibility rules of Medicaid and TANF, and assess how they relate to the amount of spending devoted to each program. Furthermore, we analyze the link between inequality and redistributive spending in a period of economic downturn—when redistributive expenditures are highly vulnerable. We also compare the effects of devolution across two redistributive programs, Medicaid and TANF, which allows us to gauge the respective effects of different funding systems: entitlements versus block grants.

This article proceeds as follows. The next section provides the theoretical framework and reviews prior research used to devise the hypotheses. We then describe our data and the results of both the eligibility rules analysis and regression analysis. The last section discusses the implications of our findings and outlines directions for future research.

Redistribution and American Federalism

Functional and Legislative Theories of Federalism

Federalism refers to the division of power between the federal government and the states. Some of these powers are specified in the U.S. Constitution, particularly in the Tenth Amendment, but many are not. Although the Supremacy Clause implies a hierarchical relationship between the levels of government, in one popular metaphor, federalism in the United States is characterized as a “marble cake” rather than a “layered cake” (Grodzins, 1966). Indeed, state and federal policy-making responsibilities are not neatly distinguished; rather, they are intertwined, subject to judicial interpretation, and constantly evolving. In many policy areas, the provision of goods and services involves various levels of government that can, in turn, work in a cooperative or competitive way. In this sense, policies at the state level must strike a fine balance between multiple factors, including the needs of state residents for particular goods and services, the level of federal supply/funding for these goods and services, the policies of neighboring states, and the needs of politicians making those policies.

Peterson (1995) advances two theories to explain the division of power between the federal government and the state governments: functional theory and legislative theory. According to functional theory of federalism, each level of government focuses on the areas in which it is most competent and can handle more efficiently. Peterson distinguishes between two broad types of programs: developmental and redistributive.

Developmental programs provide the physical and social infrastructure necessary to facilitate a country’s economic growth. The physical infrastructure includes roads, mass transit systems, sanitation systems, public parks, and a vast array of other basic utilities. The social infrastructure includes institutions that protect persons and property from unlawful activity, guard against conflagrations, protect the public health, and educate the next generation. Redistributive programs reallocate societal resources from the “haves” to the “have-nots.” They transfer economic resources from those who have gained the most from economic development to those who have gained the least: the elderly, the disabled, the unemployed, the sick, the poor, families headed by single parents, and others lacking in material resources. (Peterson, 1995, p. 17)

Given that state and local governments are subject to market forces and compete for business and affluent residents, they are more efficient in handling developmental policies. The federal government is not subject to such market forces; thus, the federal government could more efficiently implement redistribution. Yet, in the mid-1990s, the federal government delegated the responsibility for welfare programs to the states, by replacing the former system, AFDC, with state-controlled TANF. When redistribution is left to the states, they manage it in a competitive manner—as with any other program. Thus, functional theory postulates that states with higher poverty rates will have lower redistributive expenditures,¹ but higher developmental expenditures (Peterson, 1995). Policy makers in such states prefer to keep redistributive spending down to avoid becoming a “welfare magnet” and attracting poor people from neighboring states.

The legislative theory of federalism focuses on political needs of elected officials at the national and subnational levels. This perspective assumes that members of Congress and those of state legislatures are in a constant struggle for reelection. Therefore, they seek to support policies, for which they can claim credit before the electorate. In other words, policy makers are in the business of promoting popular policies and avoiding blame for unpopular ones. In regard to redistributive policies in states with high poverty rates, the credit-claiming behavior of public officials suggests an opposite expectation than the one extended under functional theory. According to legislative theory, representatives in a state that has higher percentages of low-income individuals would be confronted with “the political pressure from those representing poor people” (Peterson, 1995, p. 92). Because elected officials attempt to appear responsive, redistributive expenditures in states with higher poverty will increase, and not decrease, as predicted by functional theory.

Given the conflicting expectations of legislative and functional theories of federalism about the behavior of politicians in regard to state redistributive programs, our study attempts to test the “race to the bottom” thesis and the responsiveness of politicians to electoral preferences.

Interstate Competition: Race to the Bottom

The “race to the bottom” thesis posits that states compete with surrounding states and each state seeks to set the levels of its social welfare lower than its neighbors (Peterson, 1995). The theory builds on the assumption that states worry about designing policies that are too generous: If a state’s welfare policies are more generous than those of its neighbors, then this state risks becoming a “welfare magnet” and attracting poor people from other states. Furthermore, it also assumes that poor people possess information about benefit levels in different states as well as the means to relocate to a state that offers better welfare.

In effect, the theory produces two hypotheses, as Berry et al. (2003) accurately point out: the migration hypothesis and the benefit competition hypothesis. First, the migration hypothesis implies that poor people follow the welfare benefit and make their residential decisions based on the amount of state welfare they can collect. Furthermore, if migration in fact occurs, this would lead to a higher concentration of poor people in that state. In other words, if a state becomes a welfare magnet, its poverty rate should rise (Berry et al., 2003; Peterson, 1995; Rom et al., 1998). Thus, research testing this hypothesis focuses on whether migration actually occurs and whether the states with more generous welfare programs experience higher poverty rates.

If the migration hypothesis refers to the behavior of the poor, the benefit competition hypothesis pertains to the behavior of state politicians, who in turn perceive the migration hypothesis as being true. Fearing that higher welfare benefits might lead to an influx of less desirable populations, state officials act to prevent this from happening. Thus, they compare their levels of benefits with those of adjacent states and set their levels lower than those of their neighbors (Berry et al., 2003; Rom et al., 1998). In this sense, the competition between states influences the design of the redistributive system in each state, including redistributive budgets (Berry et al., 2003; Fellowes & Rowe, 2004). From the point of view of state legislators, a higher concentration of poor people is undesirable, because it requires even higher welfare spending and could potentially obstruct states’ efforts to attract desirable populations such as firms, investors, and affluent residents (Berry et al., 2003; Figlio, Kolpin, & Reid, 1999; Rom et al., 1998; Saavedra, 2000). The ability to attract such populations is important to states to maintain stable tax bases. Therefore, the benefit competition hypothesis concentrates on the variation in state redistributive systems. The present study tests this hypothesis.²

Most prior evidence shows that states are influenced by the policy choices of their neighbors (e.g., Figlio et al., 1999; Rom et al., 1998). Several studies analyzing AFDC—the program for cash assistance to low-income families which was in effect prior to the passage of TANF in 1996—reveal that a state’s AFDC spending was affected by adjacent states’ levels of spending on cash assistance. Bailey and Rom (2004) compare multiple programs under state control, such as AFDC, Medicaid, and Supplemental Security Income, with programs under federal control, such as Medicare, and find evidence for the “race to the bottom” hypothesis. Their study demonstrates that interstate competition lowered generosity levels of state-controlled programs but had no effect on programs controlled by the federal government. A study by Berry et al. (2003) registers a statistically significant but not substantial effect of peers’ AFDC levels. Yet there have been fewer studies on the post–welfare reform period. The few studies that have investigated TANF either used only 1-year cross-sectional data (e.g., Hero & Preuhs, 2007), compared states’ benefit levels without controlling for other plausible explanations (e.g., Albert & Catlin, 2002), or did not examine how the economic downturn of the late 2000s affected redistributive spending across U.S. states (Sheely, 2012).³ Given the prevailing evidence of interstate competition in welfare programs controlled by the states and the vast policy discretion delegated to states under TANF, we expect that a state will maintain its redistributive programs at levels that are less generous than those of neighboring states.

Policy Responsiveness: Median Voter Theorem

According to the legislative theory of federalism, elected officials strive to be responsive—or at least appear responsive—to their constituents to secure reelection, which is their main objective. The median voter theorem, in turn, indicates to whom politicians will be responsive. The theorem asserts that, in the aggregate, the outcomes in majoritarian voting systems will reflect the preferences of the median voter rather than those of the average voter.

Several studies have examined the link between redistributive policies and inequality through the lens of the median voter theorem (e.g., Bergstrom, Rubinfeld, & Shapiro, 1982; de Mello & Tiongson, 2006; Gouveia & Masia, 1998; Gramlich & Rubinfeld, 1982; Mohl & Pamp, 2009; Rodríguez, 2004). Similar to public choice theories that explain public policy with economic models, the median voter theorem suggests that the median voter is the decisive group that shapes public policies in a majority system. In other words, the majority voting equilibrium coincides with the preferences of the median voter.

Gramlich and Rubinfeld (1982) contend that higher income individuals do not have greater preference for public spending than lower income individuals, which means that the median income group exerts a greater effect on public policies than that of the higher income group. In other words, median voter income determines the demand for public goods and services rather than the mean voter income (Mueller, 2003). Such a public choice approach uses the ratio of mean to median income to gauge economic inequality across different social groups. As Mueller (2003) explains it, the ratio shows “if there are different degrees of skewness across communities, and if these differences in skewness are important in determining the demand for public goods” (p. 244).

Prior research has examined the effect of inequality on redistributive policies. The classic model belongs to Meltzer and Richard (1981). The model suggests that as income inequality (measured as the ratio of mean to median income) increases, government size (measured as the percentage of income devoted to government redistribution) grows as well. Stated differently, if the median voter becomes poorer than the mean voter in political systems with strict majority rule, redistribution should increase to compensate for this effect. A follow-up study (Meltzer & Richard, 1983), with U.S. data from 1937 to 1977, finds further support for this prediction. Since the two studies by Meltzer and Richard, the field has been split on the issue, with some studies finding evidence consistent with their model and others reporting the opposite relationship. Employing panel data from the Organisation for Economic Co-operation and Development (OECD) countries from 1975 to 2001, Karabarbounis (2011) documents that redistributive expenditures are negatively associated with the ratio of median to mean income, a result that supports the Meltzer and Richard (1981). thesis of a positive association between the ratio of mean to median income and redistribution expenditures. Findings from Gouveia and Masia (1998), however, run contrary to the argument advanced by Meltzer and Richard (1981). Analyzing panel data from the 50 U.S. states from 1979 to 1991 with a fixed effect model, Gouveia and Masia report that the ratio exerts a negative effect on redistributive⁴ expenditures, but positive effect on public supply of private goods.⁵ Rodrigìuez (1999) also finds no support for the Meltzer–Richard argument: Using data from U.S. states between 1984 and 1994, the study indicates no significant relationship between distributive skewness (the ratio of mean to median income) and welfare spending. More recent studies on the effect of inequality on redistributive policy have also arrived at findings contrary to the Meltzer–Richard idea. Using land-ownership related measures, Ramcharan (2010) finds a large negative impact of inequality on redistribution.

We contribute to this debate by testing the relationship between inequality and redistributive expenditures against newer data spanning across two redistributive programs. If the medium voter theorem—which asserts that state policy makers are responsive to preferences of the median voter—is correct, then growing income inequality will require higher redistribution.

Funding Mechanisms of Redistributive Programs: Entitlements Versus Block Grants

The national grant-in-aid system operates generally through two types of grants: categorical grants and block grants (McFarlane & Meier, 1998; Mikesell, 2010). Categorical grants, including entitlements, usually serve a special purpose and come with restrictions specifying how the money can be spent. The national government attaches conditions to its funds to ensure that recipient governments will provide the level and the types of public services consistent with national interest (Mikesell, 2010). In contrast to categorical grants, block grants come with no stipulations and allow recipient governments to spend the money at their own discretion. Traditionally, the federal government has used categorical grants (entitlements) for redistributive policies and block grants for developmental policies (Peterson, 1995).

To understand whether the effects hypothesized above differ across programs, our study analyzes two redistributive programs that operate under different funding mechanisms: Medicaid and TANF. Medicaid is an entitlement program that provides recipients with government health insurance benefits. In other words, everyone who qualifies is entitled by law to receive benefits under Medicaid. Legislation also defines the eligibility requirements and benefit levels (Kettl, 2011). To be eligible, applicants should pass a test, which reviews their socioeconomic status. As a joint federal–state program, Medicaid does not allow state governments to exercise full discretion. States can set higher benefit levels than the federal standard, but they should adhere to federal guidelines to receive matching funds and grants.

TANF is a cash assistance program that operates through block grants. It was adopted as part of welfare reform under the Personal Responsibility and Work Opportunity Reconciliation Act of 1996. Specifically, TANF replaced the previous public assistance program, AFDC, which had been criticized for causing enormous federal spending. Under TANF, each state has the authority to determine eligibility criteria and benefit levels. Specifically, states set the income level for families to qualify for cash assistance (i.e., “welfare”) and determine the amount of assistance. Under TANF, unlike AFDC, recipients are required to seek work or job training, and face time limits on benefits. The rationale for TANF enactment is that state governments are better suited to handle welfare services than the federal government because they can better identify needs within their jurisdictions (Kim & Fording, 2010). Yet states and localities are subject to market forces and thus likely to engage their redistributive systems in a “race to the bottom,” as suggested by Peterson (1995).

Data and Method

Program Accessibility for Health and Welfare

Because states have discretion over eligibility rules and benefit levels, redistributive programs vary greatly in terms of generosity. Bailey and Rom (2004) distinguish among “dimensions of generosity.” For instance, distributive programs can be designed to broaden or limit the number of recipients. To understand whether interstate competition applies to program access, we begin with an analysis of eligibility rules for Medicaid and TANF.

Given that some states have more neighbors than others, we present in Table 1 each U.S. state along with its adjacent states. For example, Colorado has seven neighbors (AZ, KS, NE, NM, OK, UT, and WY), yet Florida has only two (AL and GA). Alaska and Hawaii were dropped from the analysis because they do not have any proximate states.

Table 1.

States Sharing Common Borders.

State	Neighboring states
AL	FL	GA	MS	TN
AK
AZ	CA	CO	NV	NM	UT
AR	LA	MS	MO	OK	TN	TX
CA	AZ	NV	OR
CO	AZ	KS	NE	NM	OK	UT	WY
CT	MA	NY	RI
DE	MD	NJ	PA
FL	AL	GA
GA	AL	FL	NC	SC	TN
HI
ID	MT	NV	OR	UT	WA	WY
IL	IN	IA	KY	MO	WI
IN	IL	KY	MI	OH
IA	IL	MN	MO	NE	SD	WI
KS	CO	MO	NE	OK

Medicaid eligibility rules and index

To analyze the eligibility rules for Medicaid, we use data from the annual 50-state Survey of Medicaid and Children’s Health Insurance Program (CHIP) eligibility rules conducted by Henry J. Kaiser Family Foundation (Heberlein, Brooks, & Guyer, 2012; Ross & Cox, 2005). Heberlein et al. (2012) argue that most states maintain stable Medicaid eligibility rules over time, or at least at a level similar to the prior year, even during the Great Recession of 2008, despite severe budget shortfalls. They attribute the “maintenance of eligibility” to the passage of the ACA, which instituted the same eligibility requirements as those for Medicaid. Below, we investigate changes in the eligibility rules by comparing 2 years of data.

The Center for Medicaid and CHIP Services describes Medicaid eligibility rules. Federal law requires states to cover certain population groups (mandatory eligibility groups) and gives them the flexibility to cover other population groups (optional eligibility groups).⁶ Even though Medicaid is an entitlement program, for which the federal government pays for all eligible recipients within mandatory eligibility groups, states can have different eligibility requirements for those in optional eligibility groups. The federal government establishes a standard coverage level for each mandatory eligibility group based on income calculated as percent of the federal poverty level (FPL). States must abide by FPL coverage rules but set their own rules for additional coverage they provide.

To analyze the Medicaid eligibility rules, we first categorize dependent populations. Our study focuses on three groups: children, parents, and pregnant women. Then, we develop an eligibility index based on the income limit rule by computing the average for the three groups and analyze which state has more stringent eligibility rules compared with neighboring states. The appendix contains the eligibility analysis for each of the three groups.

To construct the total eligibility index for Medicaid, we calculate the average for the three groups—children, parents, and pregnant women (see Table 2). The comparison analysis shows that states in the Rocky Mountain region (e.g., Idaho and Utah), Plains (e.g., North Dakota and Kansas), Southeast (e.g., Alabama and Virginia), and Southwest (e.g., Texas and Oklahoma) tend to have more stringent requirements relative to other states. Moreover, the eligibility index values of the first 20 states are greater than neighboring states’ index values, indicating that these states have more restrictive policies than their neighbors. Finally, as Table 5 depicts, variation in ranking between 2005 and 2011 is very small, suggesting that high-ranking states have generally maintained restrictive eligibility rules. This result infers that states aim to keep their Medicaid benefit levels at least as low as surrounding states—a behavior consistent with the welfare magnet hypothesis.

Table 2.

Medicaid Eligibility Index: 2005 and 2011.

Ranking	2005				2011
Ranking	State	Region	Eligibility index	Neighbors’ eligibility index	State	Region	Eligibility index	Neighbors’ eligibility index
1	ND	Plains	113	174	ND	Plains	117	179
2	MT	Rocky Mt.	116	123	ID	Rocky Mt.	119	151
3	ID	Rocky Mt.	116	140	UT	Rocky Mt.	126	151
4	AL	Southeast	117	151	WY	Rocky Mt.	128	146
5	IN	Great Lakes	126	161	SD	Plains	128	171
6	VA	Southeast	127	155	AZ	Southwest	132	174
7	UT	Rocky Mt.	128	145	AR	Southeast	139	161
8	KS	Plains	129	151	NV	Far West	140	154
9	WV	Southeast	129	154	KS	Plains	140	166
10	SD	Plains	131	157	OK	Southwest	141	164
11	WY	Rocky Mt.	131	130	TX	Southwest	142	156
12	OK	Southwest	138	149	MS	Southeast	143	165
13	TX	Southwest	138	145	AK	Far West	144
14	NV	Far West	139	153	VA	Southeast	144	177
15	MS	Southeast	140	137	NC	Southeast	145	168
16	AR	Southeast	140	147	NE	PLAINS	147	165
17	LA	Southeast	140	139	FL	Southeast	148	157
18	NE	Plains	143	139	KY	Southeast	148	172
19	AK	Far West	144		MI	Great Lakes	149	161
20	CO	Rocky Mt.	146	144	MT	Rocky Mt.	152	123
41	ME	New England	186	181	CA	Far West	219	149
42	WI	Great Lakes	187	211	MD	Mid Atlantic	222	164
43	MD	Mid Atlantic	196	144	RI	New England	227	229
44	IL	Great Lakes	197	160	IA	Plains	227	195
45	NY	Mid Atlantic	200	198	VT	New England	230	213
46	CT	New England	214	203	CT	New England	247	229
47	NJ	Mid Atlantic	217	173	NY	Mid Atlantic	250	223
48	VT	New England	231	186	NJ	Mid Atlantic	250	200
49	RI	New England	231	196	MN	Plains	257	185
50	MN	Plains	277	148	WI	Great Lakes	267	227

Note. The index values reported here are based on the eligibility analysis of the three dependent groups (children, parents, and pregnant women) described in the appendix.

TANF eligibility rules and index

To analyze the eligibility rules for the TANF program across the states, we collect data from the Welfare Rules Databook (WRD; Kassabian, Whitesell, & Huber, 2012; Rowe, Murphy, & Williamson, 2006).⁷ Prior research has extensively utilized these data. For instance, Fellowes and Rowe (2004) develop TANF eligibility rule indicators based on 24 questions from the WRD. The appendix explains the procedures used to construct the TANF eligibility index for our study and analyze the patterns among the states. Specifically, we code as 1 policies that limit access to TANF, and 0 otherwise.

Table 3 reports total eligibility index at two points: 2005 and 2011.⁸ The table shows the states ranked in terms of their TANF accessibility—with the most restrictive states listed first—and depicts states’ relative position within their region. Overall, the Southeast region seems to be stricter than others: it includes Mississippi and Georgia, which are among the five states with the most restrictive TANF policies. Moreover, the eligibility index values of the first 20 states are greater than neighboring states’ index values, indicating that these states adopted policies that are more stringent than those of their neighbors. Finally, there is little variation in the rankings of states in 2005 and 2011, which suggests that high-ranking states tend to keep their restrictive policies over the years.

Table 3.

Temporary Assistance for Needy Families Eligibility Index: 2005 and 2011.

Rank	2005				2011
Rank	State	Region	Eligibility index	Neighbors’ eligibility index	State	Region	Eligibility index	Neighbors’ eligibility index
1	IN	Great_Lakes	17	11	MS	Southeast	18	10
2	MS	Southeast	17	12	GA	Southeast	16	11
3	TN	Southeast	16	13	OK	Southwest	16	11
4	GA	Southeast	16	12	IN	Great_Lakes	15	11
5	NV	Far_West	15	11	NH	New_England	15	11
6	MO	Plains	15	12	ME	New_England	14	15
7	NH	New_England	15	12	NV	Far_West	14	10
8	ME	New_England	14	15	IA	Plains	14	10
9	TX	Southwest	14	11	KY	Southeast	14	12
10	KY	Southeast	14	13	MO	Plains	14	12
11	DC		14		TX	Southwest	13	11
12	ND	Plains	14	9	MA	New_England	13	11
13	MA	New_England	13	11	TN	Southeast	13	12
14	SC	Southeast	13	12	AK	Far_West	12
15	OK	Southwest	13	11	CT	New_England	12	12
16	FL	Southeast	12	14	FL	Southeast	12	13
17	IA	Plains	12	10	HI	Far_West	12
18	LA	Southeast	12	14	NY	Mid_Atlantic	12	10
19	NY	Mid_Atlantic	12	11	SC	Southeast	12	12
20	OH	Great_Lakes	12	12	DE	Mid_Atlantic	11	10
21	PA	Mid_Atlantic	12	11	ID	Rocky Mt.	11	11
22	UT	Rocky Mt.	11	11	MT	Rocky Mt.	11	10
23	AK	Far_West	11		CA	Far_West	11	11
24	CT	New_England	11	10	OH	Great_Lakes	11	12
25	DE	Mid_Atlantic	11	11	PA	Mid_Atlantic	11	10
26	ID	Rocky	11	11	RI	New_England	10	13
27	AL	Southeast	11	15	UT	Rocky	10	10
28	IL	Great_Lakes	11	14	AZ	Southwest	10	10
29	CA	Far_West	11	12	AR	Southeast	10	14
30	OR	Far_West	10	11	OR	Far_West	10	11
31	AR	Southeast	10	15	SD	Plains	10	10
32	CO	Rocky	10	10	IL	Great_Lakes	10	13
33	MT	Rocky Mt.	10	11	KS	Plains	10	12
34	SD	Plains	10	10	ND	Plains	10	10
35	VT	New_England	10	13	VA	Southeast	10	10
36	WI	Great_Lakes	10	10	WY	Rocky	10	10
37	AZ	Southwest	10	11	DC		10
38	KS	Plains	10	11	NJ	Mid_Atlantic	10	11
39	VA	Southeast	10	11	MI	Great_Lakes	9	13
40	MD	Mid_Atlantic	10	11	WV	Southeast	9	11
41	NJ	Mid_Atlantic	10	12	CO	Rocky	9	10
42	WY	Rocky	10	10	LA	Southeast	9	14
43	HI	Far_West	9		AL	Southeast	9	15
44	WV	Southeast	9	12	MD	Mid_Atlantic	9	10
45	NM	Southwest	8	12	NM	Southwest	8	12
46	WA	Far_West	8	11	WA	Far_West	8	11
47	MI	Great_Lakes	8	15	WI	Great_Lakes	8	11
48	NC	Southeast	8	14	MN	Plains	8	11
49	NE	Plains	7	10	NE	Plains	7	11
50	MN	Plains	7	12	NC	Southeast	7	13
51	RI	New_England	6	12	VT	New_England	5	13

Figure 1 presents graphically the regional variation in the stringency of TANF eligibility rules. The analysis of state rules for TANF eligibility provides evidence that states do compete with neighboring states to avoid becoming a welfare magnet and attracting low-income individuals from the region.

Figure 1.

Comparing TANF eligibility rule stringency across U.S. States in 2005 and 2011 (higher numbers and darker color correspond to greater stringency).

Modeling State Redistributive Spending

Data and unit of analysis

To conduct regression analyses, we collect data on state Medicaid and TANF programs. Medicaid data come from the National Association of State Budget Officers (NASBO). TANF data are available from the Administration for Children and Families. We also use data from the U.S. Census Bureau, the U.S. Bureau of Economic Analysis, the U.S. Bureau of Labor Statistics, the National Conference of State Legislatures, and the National Governors Association.

We pooled data for all U.S. states from 2005 to 2011. Consequently, three states—Alaska, Hawaii, and Nebraska—were dropped from the analysis for different reasons. As previously mentioned, Alaska and Hawaii were dropped because they do not have surrounding states, which limits our ability to assess interstate competition. Nebraska is a unique case with its nonpartisan and unicameral legislature. This leaves us with 47 states and 282 state-year observations for each of the two redistributive programs, which drops to 235 after accounting for a 1-year lag for some variables.

Dependent variables

Our analysis utilizes two dependent variables: the state expenditures for each program, TANF and Medicaid, normalized by state population and adjusted to real 2012 dollars using consumer price index (TANF Expenditures per Capita and Medicaid Expenditures per Capita).⁹ Table 4 presents all variables used in the models along with their descriptions and data sources.

Table 4.

Descripton of Variables in the Models and Data Sources.

Variables	Definition	Sources
Per capita TANF	TANF and maintenance-of-effort (MOE) expenditures / state population	Administration for Children and Families
Per capita Medicaid	General fund expenditures for Medicaid / state population	State Expenditure Report by National Association of State Budget Officers (NASBO)
Ratio of mean to median income	Mean household income/median household income	U.S. Census Bureau, computed by authors
Neighbors’ per capita TANF	Mean of neighboring states’ TANF expenditure per capita	Computed by authors
Neighbors’ per capita Medicaid	Mean of neighboring states’ Medicaid expenditure per capita	Computed by authors
Neighbors’ TANF policy	Eligibility Index for TANF	Welfare Rules Databook (WRD)
Neighbors’ Medicaid policy	Eligibility Index for Medicaid	Annual 50-state survey of Medicaid and CHIP eligibility rules
Democrats in legislature	Proportion of Democrats in state legislature	National Conference of State Legislatures
Democratic governor	Governor’s party affiliation (Democratic = 1)	National Governors Association (NGA)
Per capita income	Per capita income (in thousands)	U.S. Bureau of Economic Analysis (BEA)
Unemployment rate	Annual unemployment rate	U.S. Bureau of Labor Statistics (BLS)
Population density	State population / state land area (sq miles) / 100	Computed by authors
Minority	Percent of non-White state residents	U.S. Census Bureau
Older adults	Percent of state residents older than 65 years	U.S. Census Bureau
Poverty	Percent of state residents in poverty (all ages)	U.S. Census Bureau
Public opinion	Public opinion on government spending for welfare (too little = 1, about right = 0, too much = −1)	General Social Survey (GSS)
Recession	Great Recession (2008-2009 = 1 and 0 otherwise)	U.S. National Bureau of Economic Research

Note. All expenditure figures are expressed in constant 2012 dollars. TANF = Temporary Assistance for Needy Families; CHIP = Children’s Health Insurance Program.

There is considerable variation in the amounts spent on the two programs, with Medicaid expenditures far exceeding TANF expenditures. For example, in FY 2011, 50 states spent US$397,986 million on Medicaid, which is 23.8% of total expenditures (US$1,672,045 million), or on average of 16.5% of a state’s General Fund that were dispersed for Medicaid. In comparison, only US$30,374 million was used for TANF in 50 states, excluding transferred funds to states’ social programs.¹⁰ The TANF budget share amounted to 1.82% of states’ total expenditures.

States also differ in the amount of money they devoted to each program. In 2011, New York had the largest spending on TANF per resident, with US$259.30 expressed in constant 2012 dollars, and Hawaii had the second largest TANF spending, with US$235 per capita. In contrast, Idaho and Texas spent the least relative to other states, spending US$16.70 and US$32.30 per capita, respectively. In regard to Medicaid, in 2011, Connecticut and Ohio disbursed the most funds among the states: US$1,592 and US$941 per capita. States with the lowest Medicare expenditures were Mississippi and Alabama, with US$60.3 and US$85 per capita, respectively. Table 5 reports the summary statistics of variables in the models.

Table 5.

Descriptive Statistics of Variables Used in the Models.

Variable	M	SD	Minimum	Maximum
Per capita TANF	74.872	49.554	16.684	290.195
Per capita Medicaid	360.287	245.715	49.292	1,592.204
Ratio of mean to median	1.324	0.045	1.220	1.463
Neighbors’ per capita TANF	70.151	36.598	24.451	200.994
Neighbors’ per capita Medicaid	3.703	1.862	1.248	12.912
Neighbors’ TANF policy	11.514	1.262	9	15.25
Neighbors’ Medicaid policy	1.671	0.261	1.221	2.293
Democratic governor	0.550	0.498	0	1
Democrats in legislature	0.530	0.147	0.186	0.92
Per capita income	37.381	5.783	26.443	56.959
Unemployment rate	6.069	2.368	2.6	13.8
Population density	1.988	2.606	0.055	11.869
Minority	17.507	9.385	3.648	39.748
Older adults	13.011	1.510	8.657	17.375
Poverty	13.474	3.088	7.3	22.4
Public opinion	−0.147	0.101	−0.353	0.044
Recession	0.333	0.472	0	1

Note. TANF = Temporary Assistance for Needy Families.

Main explanatory variables

Our study tests the explanatory power of two popular theories about the factors shaping redistributive policies in American political system: the median voter theorem and the race to the bottom. Specifically, we examine how helpful they are in explaining the level of redistributive spending in the states. If the median voter theorem is correct, redistributive spending should reflect the spending preferences of the median voter. Therefore, when income inequality widens (i.e., the median voter becomes poorer than the average voter), the government should respond by increasing redistributive spending to compensate for this. We use the ratio of mean to median income to operationalize the extent of inequality in each state.¹¹ We denote m as inequity (i.e., the ratio of mean to median income), where m = (mean income) ⁄ (median income). We assume redistributive expenditures (Y) to be an increasing function of the ratio m, ΔY⁄Δm > 0. In our data set, New York is the state with the highest inequality, m = 1.45 in 2011, followed by Florida with m = 1.42. In contrast, Alaska and Wyoming had the lowest inequality, with ratios of 1.219 and 1.223, respectively.

To test the “race to the bottom” thesis advanced by Peterson (1995), we use the Neighbors’ Expenditures on TANF and Medicare. Besides spending, states can compete in limiting access to redistributive programs, as our prior analysis suggests. To account for differences in the eligibility rules, we include in the model the variables developed in our earlier analysis—the eligibility indices for TANF and Medicaid (neighbors’ policies).

Control variables

We also include a number of controls to account for other plausible explanations of the amount of money states devote to redistribution. We control for the political, socioeconomic, and demographic context in the states; public opinion; and the impact of economic crisis of 2008.

Because redistribution is associated with more liberal policies, we control for the proportion of Democrats in Legislature. We also include to the models a dummy variable with a value of 1, if a state has a Democratic governor. Based on prior research (Peterson, 1995; Rom et al., 1998), states headed by Democratic governors and those with more Democrats in legislature should spend more on redistribution. Admittedly, political parties across states differ in their policy preferences and representation of the poor. A study by Rigby and Wright (2013) reports that similar to Republicans, Democrats tend to cater to preferences of more affluent citizens. Studies mapping ideological stances of state legislators (e.g., Shor & McCarty, 2011; Wright & Birkhead, 2014) find that, for instance, the Democratic party in Mississippi is more conservative than the relatively liberal Republican parties of Connecticut, Delaware, Hawaii, Illinois, Massachusetts, New Jersey, New York, and Rhode Island. Generally, relatively conservative Democratic parties are typical for states like Alabama, Arkansas, Kentucky, Louisiana, North Dakota, Oklahoma, South Carolina, South Dakota, and West Virginia. Yet, as Rigby and Wright (2013) point out, “it is clear that the Democrats are a good deal more liberal and many of their policies favor redistribution and regulations that would improve the material circumstances of those with lower incomes” (p. 563).

To account for the broader socioeconomic and demographic profile of a state, we include a number of economic indicators such as per capita income, unemployment rate, poverty rate, population density, percent minority residents, and percent older adult residents. Higher per capita income is associated with more taxable resources. The more resources a state has, the more it can spend on redistribution. Thus, per capita income is expected to have a positive effect on redistributive spending (Peterson, 1995). Unemployment Rate is another important control because it directly relates to the demand of welfare programs: the more the unemployed residents, the higher the demand for public assistance. Thus, we expected that an increase in unemployment rate would result in higher redistributive expenditures. Population Density indicates the degree of urbanization and is usually associated with more developmental expenditures (Choi, Bae, Kwon, & Feiock, 2010). In this sense, we expect population density to have a negative effect on redistributive spending. We also account for size of the racial/ethnic Minority population, measured as the percent of non-White residents. According to the legislative theory of federalism (Peterson, 1995), higher minority population will negatively affect redistributive expenditures because minorities are underrepresented in state legislatures and generally perceived as undeserving (Barrilleaux, Holbrook, & Langer, 2002). Furthermore, the model includes a variable, called Older Adults, operationalized as the percent of residents older than 65 years in each state, which is expected to have a positive effect on redistributive spending (Barrilleaux et al., 2002). Poverty indicates the percent of state residents of all ages in poverty. This variable is hypothesized to be positively associated with higher redistribution because of federal mandates and political pressures to ameliorate poverty (Figlio et al., 1999; Peterson, 1995).¹² In our data set, states with the highest percentage of poor people are Mississippi (22.8%) and New Mexico (20.9%), while the lowest levels of impoverishment are registered in New Hampshire (9%) and Maryland (10.2%).

The model also controls for Public Opinion about social welfare. We used a question from the General Social Survey (GSS) inquiring whether the respondent thinks the government is spending too much, too little, or about the right amount on social welfare.¹³ A vast body of literature argues that politicians are, on average, responsive to public preferences (e.g., Page & Shapiro, 1983; Wlezien, 1995). As Stimson MacKuen, and Erikson (1995) write, “Public sentiment shifts. Political actors sense the shift. And they alter their policy behavior at the margin” (p. 543). Therefore, greater public support for welfare is likely to boost state expenditures for Medicaid and TANF. Not surprisingly, the average value of the public opinion variable is negative, suggesting that the average American opposes greater redistribution and thinks that the poor are already given too much. Prior research has well recognized this pattern and offered various explanations for the American contempt for welfare (see Gilens, 2009, 2012).

Furthermore, we create a variable that accounts for the effect of the economic downturn of 2008. Recession is operationalized as a dummy with a value of 1 for the years 2008 and 2009, and 0 otherwise.¹⁴ The Great Recession brought enormous fiscal stress to state governments, by lowering revenues and raising demand for public services. Concerned with balancing their budgets and sustaining the level of essential services, state politicians might spend less to care for the needy. Thus, we anticipate redistributive spending to decrease during the recession.

Finally, we enter in the model the prior year’s expenditures on TANF and Medicare. We do so to account for the incremental nature of the budget process: “the largest determining factor of this year’s budget is last year’s” (Wildavsky & Caiden, 2001, p. 47).

Estimation routine

We estimate a dynamic panel data (DPD) model with the log of state redistributive expenditures for each program¹⁵ as the dependent variable and the lagged value of the dependent variable on the right-hand side of the equation. The DPD technique is known as an efficient method to estimate models that include the lagged dependent variable among explanatory variables. The regression equation is as follows:

\ln Y_{it} = β_{1} X_{it - 1} + β_{2} W_{it - 1} + u_{i} + e_{it - 1}, where

$X_{it}$ are strictly exogenous regressors, and $W_{it}$ are predetermined regressors and endogenous regressors that include $Y_{it - 1}$ so that the dynamic panel model uses $Y_{it - 1}$ as an instrument (Baum, 2006).¹⁶ Because this is a log-linear model, coefficients indicate the percentage change of dependent variable per additional one unit change of X or W and the effects are calculated by using derivatives. That means the explanatory variables affect the rate of expenditure change, not the unit change.¹⁷

Results: Explaining state expenditures for health and welfare

Table 6 presents regression results.¹⁸ Models 1 and 3 use TANF expenditures per capita as the dependent variable, and Models 2 and 4 use Medicaid expenditures per capita as the dependent variable. Models 3 and 4 seek to capture the effect of the Great Recession and include a dummy variable that is 1 for the years of recession. Wald chi-square tests indicate good fit: All regression models are statistically significant, with p values less than .001.

Table 6.

Effects on Medicaid and TANF Expenditures (2005-2011).

	Model 1 (TANF)		Model 2 (Medicaid)		Model 3 (TANF)		Model 4 (Medicaid)
	β	SE	β	SE	β	SE	β	SE
Lagged dependent variable	0.111	0.120	0.263**	0.113	0.096	0.123	0.255*	0.130
Ratio of mean to median income_t–1	−1.692	1.147	−0.735	0.692	−0.959	1.000	−0.679	1.705
Neighbors’ expenditures_t–1	−0.005**	0.002	−0.023	0.018	−0.004*	0.002	−0.027	0.022
Neighbors’ policies_t–1	−0.073*	0.043	−0.054	0.225	−0.062	0.049	−0.048	0.274
Democrat governor_t–1	0.081	0.062	−0.009	0.046	0.110*	0.060	−0.008	0.048
Democrats in legislature_t–1	0.019	0.502	0.867**	0.370	−0.028	0.520	0.783*	0.432
Per capita income_t–1	0.022*	0.013	−0.013	0.009	0.027**	0.011	−0.013	0.008
Unemployment rate_t–1	0.099***	0.015	−0.071***	0.013	0.089***	0.017	−0.071***	0.014
Population density_t–1	−1.669	1.133	−0.606	0.796	−1.660	1.171	−0.412	0.889
Minority_t–1	0.057	0.087	0.083	0.052	0.059	0.089	0.073	0.062
Older adults_t–1	−0.164	0.104	−0.034	0.109	−0.107	0.118	−0.018	0.114
Poverty_t–1	−0.026	0.031	0.075***	0.023	−0.044	0.033	0.071***	0.026
Public opinion_t–1	0.218*	0.122	−0.023	0.277	0.278**	0.136	−0.012	0.296
Recession					−0.062*	0.034	−0.003	0.046
Number of observations	235		235		235		235
Number of states	47		47		47		47
Wald χ² (11)	171.00***		133.61***		118.66***		114.49***

Note. The dependent variable is the logarithm of TANF expenditure per capita in Models 1 and 3 and the logarithm of Medicaid expenditure per capita in Models 2 and 4. Time Series = 2005-2011; cross section = 47 states (Alaska, Hawaii, and Nebraska are omitted). TANF = Temporary Assistance for Needy Families.

p < .1. **p < .05. ***p < .01.

We start with results for the two main hypotheses. To test the median voter theorem, we used the ratio of mean to median as the key explanatory variable. Yet its coefficient failed to reach statistical significance in all four models. Thus, our data do not support the expectation that state redistributive policies respond to rising income inequality in the U.S. states.

Second, we find that a state’s redistributive budget is affected by the policy choices of neighboring states. The coefficient on neighbors’ expenditure for both TANF and Medicaid is negative, indicating the presence of competition. Specifically, if neighboring states increase their redistributive spending by one unit per capita, a state’s expenditures would decrease by 0.49% in per capita for TANF and 2.27% in per capita for Medicaid,¹⁹ ceteris paribus. Yet the coefficient for Medicaid is not statistically significant, so the competition effect of neighbor states’ expenditure is meaningful only in the case of TANF. This result indicates that the interstate competition hypothesis predicting that states compete in reducing their welfare to avoid becoming welfare magnets has explanatory power only for the TANF program, which is fully state-controlled and funded through a block grant mechanism. The other key independent variable, Neighbors’ Policies, was also significant only in Model 1 explaining TANF spending.

Among the socioeconomic variables, our analysis shows that state poverty rate affects the spending level for Medicaid but not for TANF. In Models 2 and 4, the coefficients on poverty rate are 0.075 and 0.071, which means that 1% change of poverty rate increases per capita Medicaid expenditure by 7.79% and 7.36%, respectively. This result could also be attributed to the different funding mechanisms through which the two programs operate. Specifically, state governments do not have much discretion on Medicaid expenditures because it is an entitlement program. In contrast, for TANF, states have the authority to set their own eligibility rules and benefit levels. The results from our analysis imply that policy makers are not likely to consider the poverty level in their states when deciding on the level of TANF spending. However, state poverty rate is a key component for Medicaid eligibility, so it was expected that the poverty variable would be positive and statistically significant in the model explaining state Medicaid expenditures. The data also show that more resourceful states can afford spending more for welfare. Per capita income speaks to the size of a state’s revenues source. The variable exhibits a positive and significant effect on TANF expenditures. Specifically, the coefficients in Models 1 and 3 are 0.022 and 0.027, which indicate that US$1 change of per capita income results in an increase in per capita TANF expenditure by 2.2% and 2.7%, respectively. This result can also be interpreted in connection with the funding mechanisms used by the two programs. As per the Federal Medical Assistance Percentage (FMAP), Medicaid expenditures are backed by the federal government. Therefore, even if a state decides to expand its current Medicaid coverage, its fiscal burden is relatively small compared with TANF.

Moving to political context variables, as predicted, higher percentage of Democrats in state legislature is associated with more money spent on Medicaid. The coefficients on Democrats in Legislature in Models 2 and 4 are 0.876 and 0.783. Both coefficients are statistically significant. In other words, 1% increase in Democrats in Legislature induces 0.87% and 0.79% increase in Medicaid expenditures per capita.²⁰ This finding is consistent with prior research showing that Democrats favor greater redistribution than Republicans (Fellowes & Rowe, 2004; Peterson, 1995; Rigby & Wright, 2013).

Furthermore, the coefficient on political opinion is significant in Models 1 and 3. The results indicate that public opinion on welfare spending is positively associated with TANF spending. In other words, states whose residents show less support for welfare programs are likely to spend less for redistribution than states whose residents are more supportive of social welfare.

Finally, we examined the effect of the Great Recession on state redistributive expenditures. Models 3 and 4 include an indicator that is coded 1 for 2008 and 2009, the years of the recession. The estimations revealed that the economic downturn negatively affected TANF spending but not Medicaid spending. Again, this differential effect can be attributed to the difference in funding mechanism. The recession strained state budgets and states cut expenditures that were under their control. This result also indicates that programs financed through block grants are more vulnerable during economic downturns than entitlement programs.

Conclusion

Redistributive programs are public programs designed with the aim to help the poor and alleviate poverty and economic inequality. Although no one would disagree about the need for redistribution on moral and human dignity grounds, the extent of redistribution remains highly contested and varies greatly across societies. These controversies become even more pronounced in federal systems like the United States, where the services for the poor are delivered through joint financing between the national and state governments and through various mechanisms that allow for more or less discretion on the part of subnational actors.

We argue here that state politicians face conflicting incentives and must strike a subtle balance when designing their welfare systems. As elected representatives, they must respond to the needs of their constituents and ensure that all residents have adequate living standards regardless of their socioeconomic status. Yet states must respond to the needs of the poor but in a way that is less generous than neighboring states. Peterson (1995) argued that states compete in a “race to the bottom” and seek to keep redistribution down to avoid becoming a welfare magnet. Consequently, our study tested two main hypotheses—whether state redistributive expenditures vary systematically as a function of economic inequality and whether state redistributive programs engage in a “race to the bottom.” To run the analyses, we used data from two state redistributive programs—Medicaid and TANF—which receive funding from the national government through two different mechanisms. Medicaid is an entitlement program that provides health care insurance to low-income individuals, and states exercise little discretion in determining recipients’ eligibility and benefit levels. TANF is a cash assistance program that operates through block grants from the federal government, which allows states to set their own eligibility rules and benefit levels.

Our analysis shows that the median voter theorem offers little help in explaining why the level of redistributive spending varies among states. Our data indicate that state politicians do not alter their policy choices when the median voter’s income falls sharply below the income of the average voter. In other words, widening inequality in a state cannot serve as a predictor of the level of redistribution in this state. Yet we find evidence of interstate competition in redistributive programs, as predicted by Peterson (1995). Both the analysis of program accessibility and regression analysis of state redistributive spending offered support for the race to the bottom hypothesis. The effects, however, are stronger for TANF compared with Medicaid. Moreover, the data demonstrate that the Great Recession of 2008 had greater impact on TANF expenditures than Medicaid expenditures. States lowered their spending for cash assistance to the poor, but health services to the poor were not significantly affected. We attribute these differential effects to differences in funding mechanism. Programs funded through block grants are subject to negative interstate competition and vulnerable when the economy is weak.

Our analysis also raises questions for future investigations. First, this article analyzed only two redistributive programs: TANF and Medicaid. Yet Peterson (1995) defines redistributive programs as encompassing a range of services and argues that “a race to the bottom” should occur in all of them. Second, we find that politicians do not respond to shifts in social conditions in a state, especially in the case of TANF, but they are responsive to public opinion—a result that is in line with prior research. One caveat of our analysis is that the public opinion data do not differentiate between redistributive programs. It might be that policy makers are in fact responsive, but public support for different programs is uneven, with more favorable public opinion toward health care services for the poor as opposed to cash assistance. Third, our study analyzes redistributive spending from the perspective of sponsoring governments, that is, state elected and appointed public officials. Yet the differential effects for Medicaid and TANF registered here might be linked to recipients’ understanding and use of these benefits. Given that TANF is more of a cash-in-hand benefit, it may be easier for recipients to understand its benefits. On the contrary, the monetary value of Medicaid benefits is not that obvious and might vary depending on the services provided, and thus, it may be more difficult for recipients to comprehend its benefit. Future research could examine the perceived utility of the two programs from the view point of users, whether and how these benefits are utilized by those in need and whether the extent of use affects the governments’ decision about the size of redistributive spending in a state.

The results of our study have strong implications about how redistribution works in majoritarian democracies with federal systems of government. We focus on two main implications. First, we find that states do not respond to economic inequality within their jurisdictions, despite the fact that economic inequality in the United States has dramatically risen over the last three decades. Such finding is especially indicative of the state of our democratic governance, given the evidence in the current literature relating economic inequality to the notion of unequal democracy (e.g., Bartels, 2016; Solt, 2008). Inequality suppresses political engagement of economically disadvantaged citizens, which combined with the general public sentiment against redistribution serves to effectively take the discussion about the poor off the political agenda. As Solt (2008) contends, “[t]he declining political engagement of nonaffluent citizens with rising inequality suggests that issues on which a consensus exists among richer individuals, such as redistribution, become increasingly unlikely even to be debated within the political process” (p. 57). Furthermore, our data reveal that the amount of TANF expenditures does not reflect a state’s poverty level. If the goal of redistributive policies is to help poor populations, these policies should presumably reflect social conditions within the states. Such finding underscores once again the economic development priorities of lower level governments over social equality concerns.

Second, our analysis indicates that the negative competition applies mostly to programs like TANF, where states have the authority to set the eligibility rules and the federal funding comes through block grants. Given the controversial nature of redistributive policies and the fact that states will always be subject to market forces and balanced budget rules, it seems that the federal government might need to assume a greater role in redistribution, as suggested by the functional theory of federalism. The federal government is in a better position to fund redistributive programs because of its ability to tax the rich and impose a progressive structure of taxation. In contrast, state governments barely use any progressive taxes, fearing that such taxes would discourage businesses and affluent residents to stay in the state. Thus, as argued by Peterson (1995), the federal government is a more competent and capable agent of redistribution compared with state and local governments. Even more, keeping the decisions about redistributive policies at the state level can potentially deepen social inequality, exacerbate political exclusion of economically disadvantaged populations, and make democracy a privilege of the rich.

Footnotes

Appendix

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by the Ministry of Education of the Republic of Korea and the National Research Foundation of Korea (NRF-2016S1A3A2924563; NRF-2017S1A3A2065838).

Notes

ORCID iD

Milena I. Neshkova

Author Biographies

NakHyeok Choi is a research fellow in the Community Wellbeing Research Center at Seoul National University, South Korea. He received his Ph.D. from Florida International University. His research interests focus on government capacity, public finance and budgeting, infrastructure, and transportation.

Milena I. Neshkova is an associate professor of public administration at Florida International University. She is interested in the issues of responsiveness, representation, and accountability of public administrators in both domestic and international settings. Other research interests include the adoption and implementation of administrative reforms in new democracies.

References

Albert

Catlin

S. N.

(2002). Strategic interaction among the states: An in-depth look at the welfare “race to the bottom.” Social Work Research, 26, 199-216. doi:10.1093/swr/26.4.199

Bailey

M. A.

Rom

M. C.

(2004). A wider race? Interstate competition across health and welfare programs. Journal of Politics, 66, 326-347. doi:10.1111/j.1468-2508.2004.00154.x

Barrilleaux

Holbrook

Langer

(2002). Electoral competition, legislative balance, and American state welfare policy. American Journal of Political Science, 46, 415-427. doi:10.2307/3088385

Bartels

L. M.

(2016). Unequal democracy: The political economy of the new gilded age. New York, NY: Princeton University Press.

Baum

C. F.

(2006). An introduction to modern econometrics using stata. College Station, TX: Stata Press.

Bergstrom

T. C.

Rubinfeld

D. L.

Shapiro

(1982). Micro-based estimates of demand functions for local school expenditures. Econometrica, 50, 1183-1205. doi:10.2307/1911869

Berry

W. D.

Baybeck

(2005). Using geographic information systems to study interstate competition. The American Political Science Review, 99, 505-519.

Berry

W. D.

Fording

R. C.

Hanson

R. L.

(2003). Reassessing the “race to the bottom” in state welfare policy. The Journal of Politics, 65, 327-349. doi:10.1111/1468-2508.t01-2-00003

Chernick

Reschovsky

(2003). State fiscal responses to welfare reform during recessions: Lessons for the future. Public Budgeting & Finance, 23(3), 3-21. doi:10.1111/1540-5850.2303001

10.

Choi

S. O.

Bae

S.-S.

Kwon

S.-W.

Feiock

(2010). County limits: Policy types and expenditure priorities. The American Review of Public Administration, 40, 29-45. doi:10.1177/0275074008328171

11.

de Mello

Tiongson

E. R

. (2006). Income inequality and redistributive government spending. Public Finance Review, 34, 282-305. doi:10.1177/1091142105284894

12.

Fellowes

M. C.

Rowe

(2004). Politics and the new American welfare states. American Journal of Political Science, 48, 362-373. doi:10.1111/j.0092-5853.2004.00075.x

13.

Figlio

D. N.

Kolpin

V. W.

Reid

W. E.

(1999). Do states play welfare games? Journal of Urban Economics, 46, 437-454. doi:10.1006/juec.1999.2131

14.

Freitas

(2012). Inequality, the politics of redistribution and the tax mix. Public Choice, 151, 611-630. doi:10.1007/s11127-010-9761-z

15.

Gilens

(2009). Why Americans hate welfare: Race, media, and the politics of antipoverty policy. Chicago, IL: The University of Chicago Press.

16.

Gilens

(2012). Affluence and influence: Economic inequality and political power in America. Princeton, NJ: Princeton University Press.

17.

Gouveia

Masia

(1998). Does the median voter model explain the size of government? Evidence from the states. Public Choice, 97, 159-177. doi:10.1023/A:1004973610506

18.

Gramlich

E. M.

Rubinfeld

D. L.

(1982). Micro estimates of public spending demand functions and tests of the tiebout and median-voter hypotheses. Journal of Political Economy, 90, 536-560.

19.

Grodzins

(1966). The American system: A new view of government in the United States. Transaction Publishers.

20.

Heberlein

Brooks

Guyer

(2012). Performing under pressure: Annual findings of a 50-state survey of eligibility, enrollment, renewal, and cost-sharing policies in Medicaid and CHIP, 2011-2012. Washington, DC: Kaiser Commission on Medicaid and the Uninsured.

21.

Hero

R. E.

Preuhs

R. R.

(2007). Immigration and the evolving American welfare state: Examining policies in the U.S. states. American Journal of Political Science, 51, 498-517. doi:10.1111/j.1540-5907.2007.00264.x

22.

Karabarbounis

(2011). One dollar, one vote. The Economic Journal, 121, 621-651. doi:10.1111/j.1468-0297.2010.02406.x

23.

Kassabian

Whitesell

Huber

(2012). Welfare rules databook: State TANF policies as of July 2011 (OPRE Report No. 2012-57). Washington, DC: The Urban Institute.

24.

Kettl

D. F.

(2011). The politics of the administrative process. Washington, DC: CQ Press.

25.

Kim

Fording

R. C.

(2010). Second-order devolution and the implementation of TANF in the U.S. states. State Politics & Policy Quarterly, 10, 341-367. doi:10.1177/153244001001000404

26.

McFarlane

D. R.

Meier

K. J.

(1998). Do different funding mechanisms produce different results? The implications of family planning for fiscal federalism. Journal of Health Politics, Policy and Law, 23, 423-454. doi:10.1215/03616878-23-3-423

27.

Meltzer

A. H.

Richard

S. F.

(1981). A rational theory of the size of government. Journal of Political Economy, 89, 914-927. doi:10.2307/1830813

28.

Meltzer

A. H.

Richard

S. F.

(1983). Tests of a rational theory of the size of government. Public Choice, 41, 403-418.

29.

Mikesell

J. L.

(2010). Fiscal administration. Boston, MA: Cengage Learning.

30.

Mohl

Pamp

(2009). Income inequality and redistributional spending: An empirical investigation of competing theories. Public Finance and Management, 9(2), 1-56.

31.

Mueller

D. C.

(2003). Public choice III. Cambridge, UK: Cambridge University Press.

32.

Page

Shapiro

(1983). The effects of public opinion on policy. American Political Science Review, 65, 389-400.

33.

Persson

Tabellini

(1994). Is inequality harmful for growth? The American Economic Review, 84, 600-621. doi:10.2307/2118070

34.

Peterson

P. E.

(1995). The price of federalism. Washington, DC: Brookings Institution Press.

35.

Ramcharan

(2010). Inequality and redistribution: Evidence from US counties and states, 1890–1930. The Review of Economics and Statistics, 92, 729-744.

36.

Rigby

Wright

G. C.

(2013). Political parties and representation of the poor in the American states. American Journal of Political Science, 57, 552-565.

37.

Rodrigìuez

F. C.

(1999). Does distributional skewness lead to redistribution? Evidence from the United States. Economics & Politics, 11, 171-199. doi:10.1111/1468-0343.00057

38.

Rodríguez

(2004). Inequality, redistribution, and rent-seeking. Economics & Politics, 16, 287-320. doi:10.1111/j.1468-0343.2004.00141.x

39.

Rom

M. C.

Peterson

P. E.

Scheve

K. F.

(1998). Interstate competition and welfare policy. Publius: The Journal of Federalism, 28(3), 17-37.

40.

Ross

D. C.

Cox

(2005). In a time of growing need: State choices influence health coverage access for children and families: A 50 state update on eligibility rules, enrollment and renewal procedures, and cost-sharing practices in Medicaid and SCHIP for children and families. Washington, DC: Kaiser Commission on Medicaid and the Uninsured.

41.

Rowe

Murphy

Williamson

(2006). Welfare rules databook: State TANF policies as of July 2005. Washington, DC: The Urban Institute.

42.

Saavedra

L. A.

(2000). A model of welfare competition with evidence from AFDC. Journal of Urban Economics, 47, 248-279. doi:10.1006/juec.1999.2141

43.

Sheely

(2012). Devolution and welfare reform: Re-evaluating “success.” Social Work, 57, 321-331. doi:10.1093/sw/sws022

44.

Sheely

(2013). Second-order devolution and administrative exclusion in the temporary assistance for needy families program. Policy Studies Journal, 41, 54-69.

45.

Shor

McCarty

(2011). The ideological mapping of American legislatures. American Political Science Review, 105, 530-551.

46.

Solt

(2008). Economic inequality and democratic political engagement. American Journal of Political Science, 52, 48-60.

47.

Stimson

J. A.

MacKuen

M. B.

Erikson

R. S.

(1995). Dynamic representation. American Political Science Review, 89, 543-565.

48.

Volden

(2002). The Politics of Competitive Federalism: A Race to the Bottom in Welfare Benefits? American Journal of Political Science, 46, 352-363.

49.

Wildavsky

A. B.

Caiden

(2001). The new politics of the budgetary process. New York, NY: Addison-Wesley.

50.

Wlezien

(1995). The public as Thermostat: Dynamics of preferences and spending. American Journal of Political Science, 39, 981-1000.

51.

Wright

G. C.

Birkhead

(2014). The macro sort of the state electorates. Political Research Quarterly, 67, 426-439.

State	Neighboring states
AL	FL	GA	MS	TN
AK
AZ	CA	CO	NV	NM	UT
AR	LA	MS	MO	OK	TN	TX
CA	AZ	NV	OR
CO	AZ	KS	NE	NM	OK	UT	WY
CT	MA	NY	RI
DE	MD	NJ	PA
FL	AL	GA
GA	AL	FL	NC	SC	TN
HI
ID	MT	NV	OR	UT	WA	WY
IL	IN	IA	KY	MO	WI
IN	IL	KY	MI	OH
IA	IL	MN	MO	NE	SD	WI
KS	CO	MO	NE	OK

State	Neighboring states
AL	FL	GA	MS	TN
AK
AZ	CA	CO	NV	NM	UT
AR	LA	MS	MO	OK	TN	TX
CA	AZ	NV	OR
CO	AZ	KS	NE	NM	OK	UT	WY
CT	MA	NY	RI
DE	MD	NJ	PA
FL	AL	GA
GA	AL	FL	NC	SC	TN
HI
ID	MT	NV	OR	UT	WA	WY
IL	IN	IA	KY	MO	WI
IN	IL	KY	MI	OH
IA	IL	MN	MO	NE	SD	WI
KS	CO	MO	NE	OK

State	Neighboring states
AL	FL	GA	MS	TN
AK
AZ	CA	CO	NV	NM	UT
AR	LA	MS	MO	OK	TN	TX
CA	AZ	NV	OR
CO	AZ	KS	NE	NM	OK	UT	WY
CT	MA	NY	RI
DE	MD	NJ	PA
FL	AL	GA
GA	AL	FL	NC	SC	TN
HI
ID	MT	NV	OR	UT	WA	WY
IL	IN	IA	KY	MO	WI
IN	IL	KY	MI	OH
IA	IL	MN	MO	NE	SD	WI
KS	CO	MO	NE	OK