Immigrant Social Policy in the American States: Race Politics and State TANF and Medicaid Eligibility Rules for Legal Permanent Residents

Abstract This article examines differences in the drivers of state Temporary Assistance for Needy Families (TANF) and Medicaid immigrant eligibility policies, determined in the wake of the 1996 Welfare Reform. The findings show that differences in the incentive structures of the two programs may affect the way race politics influence each. Specifically, race is a strong negative correlate for TANF inclusion of immigrants as states with large African American populations were more likely to exclude legal permanent residents from the program. In the case of Medicaid, the size of the immigrant population is a strong positive correlate for inclusion. The effect of the size of the black population, although negative, is small and not significant. The study confirms extant research findings that ideological factors play an important role in the formation of both policies.


Introduction
Since the formation of the United States, race politics has been at the heart of immigration and social welfare policies at the federal and the state levels. Over the past century, immigration policy and social welfare policy have served as the means to maintain systems of racially based preferential treatment and to discriminate against racial minorities (Brown 1999;Daniels 2004;Fording 2003;Fox-Piven and Cloward 1971;Gilens 1999;Haney-Lopez 2006;Hero 1998;Katz 1989;King 2002;Ngai 2004;Quadagno 1994;Tichenor 2002).
The explanation for these empirical findings lies in Key's (1949) "group threat" (which in this article, I call "race politics") hypothesis, which posits that in the American polity, adversarial relationships have developed between the dominant white majority and the country's racial minorities. In recent decades, this conflict has centered on multicultural policies (e.g., bilingual education, affirmative action) and social policies that benefit large numbers of minorities and immigrants. Key argued that fearing loss of power and privilege, whites in America sought to exclude and marginalize African Americans and other minorities socially, economically, and politically. This "group threat" hypothesis has been extensively tested and validated in the context of black-white relationships especially in research on attitudes (Blalock 1967;Blumer 1958;Bobo and Hutchings 1996;Glaser 1994;Taylor 1998;Tolbert and Hero 2001). Research in policy formation has sought to tie the theory of "group threat" to policy outcomes, on the premise that responsive politicians (themselves of the majority racial group) will cater to the demands of the threatened majority. Indeed, a number of studies show that in immigration policy and welfare policy, race politics is implicated in policy outcomes (Campbell, Wong, and Citrin 2006;Fording 2003;Hero and Preuhs 2007;M. Johnson 2003;Schildkraut 2001;Soss, Fording, and Schram 2008;Soss, Schram, and Fording 2003;Tolbert and Hero 1996).
The influence of race politics on immigration and welfare policies is well documented; however, the effects of race politics at the intersection of the two policy domains remain relatively unexplored. The two policy areas intersected in 1996 when Congress passed the most extensive welfare reform in a generation. The Personal Responsibility and Work Opportunity Reconciliation Act of 1996 (PRWORA or Welfare Reform Act) introduced fundamental change to America's social welfare system by eliminating the welfare entitlement program that was established by the 1935 Social Security Act, and by expanding the autonomy of states in the establishment of welfare eligibility rules (Mettler 2000;Soss, Schram, and Fording 2003). PRWORA was also a ground-breaking reform of American immigration law. The new legislation decentralized authority to states, allowing them to decide on immigrant access to federally funded low-income support programs. As a result of the Act, states assumed responsibility for setting eligibility rules for legal permanent residents (LPRs) for the Temporary Assistance for Needy Families (TANF) program and for Medicaid. Despite concerns about the potential for wholesale immigrant exclusion from these programs (Fix and Tumlin 1996), states, in fact, produced varied responses: Some states chose to incorporate almost all LPRs into their public benefits and health care programs; other states were far more selective and restrictive in terms of eligibility and inclusion. 1 Some states chose to exclude LPRs from TANF only, others from both programs.
This article seeks to explain variation in state responses to the devolution of authority in the immigrant social policy domain, and to evaluate the role that race politics played in state policy decisions. In contrast to previous research that has either focused exclusively on TANF (Graefe et al. 2008) or on a combination of social programs (Hero and Preuhs 2007), this study analyzes state immigrant eligibility rules for TANF and Medicaid separately. PRWORA invited states to change their immigrant eligibility rules not only for TANF but also for Medicaid. Yet, little attention has been paid by political science to this health care policy change. This is a major omission because, as will be discussed, the two programs have very different incentive structures set by the federal government. In the case of TANF, states have an incentive to exclude, as predicted by the "group threat" hypothesis. However, this is not true in Medicaid where the entitlement nature of the program and the funding formula employed along with the federal requirement of universal access to emergency care may make the costs of exclusion higher than the political benefits. By looking separately at TANF and Medicaid, the findings break from previous research to show that the two policies are quite different, and to caution against analyzing "welfare reform" for immigrants as a unidimensional phenomenon.

PRWORA and the Devolution of Immigrant Social Policy
The Welfare Act represented a substantial shift in American welfare and immigration federalism. PRWORA dismantled the Aid to Families with Dependent Children (AFDC) entitlement program and replaced it with the TANF block grant. TANF was also decoupled from Medicaid, the health insurance program for the poor, which continued life as a means-tested federal entitlement program.
With PRWORA, welfare reform and devolution were tied to immigration reform and devolution. States, for the first time, had to decide whether to exclude immigrants from social programs. Although Medicaid eligibility rules did not change for citizen recipients, for immigrants, the new law specified that the same eligibility rules that applied for TANF also applied for Medicaid. According to PRWORA, states could choose to include or exclude from TANF and/or Medicaid any or all the three groups of immigrants that the law created. 2 If states chose to incorporate immigrants in their programs, Washington matched funds only for preenactment immigrants and those who had been in LPR status for more than 5 years. New immigrants (those in their first 5 years of residency) were excluded from federal TANF and Medicaid funding. If states wanted to extend any benefits to this group they had to do so proactively by establishing state-funded initiatives.

Race Politics at the Intersection of Two Policy Domains
Immigration and social welfare policies have been analyzed from the perspective of race relationships and racial threat (Dixon and Rosenbaum 2004;Fellowes and Rowe 2004;Fording 2003;Fraga et al. 2010;Hero and Preuhs 2007;Hero and Tolbert 1996;M. Johnson 2003;Soss et al. 2001;Soss, Fording, and Schram 2008;Soss, Schram, and Fording 2003;Tolbert and Hero 1996;Wright 1976). In both cases, research has shown that the presence of large minority populations tends to lead to more restrictive and harsher policy preferences among voters and more restrictive policy outcomes. Although we tend to treat immigration and social welfare policies as two separate domains, race serves as an important link between the two. Alone among areas of law, immigration policy is not subject to the same strict scrutiny that the Supreme Court has applied to cases involving race and minorities. The federal government can and does continue to use race as a factor in its immigration policy decision making (Chin 1998). Especially threatening to the white majority were those people at the intersection of race and class: poor non-white immigrants who were thought to have a higher likelihood of becoming wards of the state and a social burden (Katz 1989). Concerns over welfare, or the likelihood that an immigrant may become a "public charge," serve as the basis for exclusion from immigration to the United States as well as a limitation for naturalization. 3 Since PRWORA, immigration status serves as the basis for exclusion from welfare programs.

Race Politics and Immigration Policy
Americans (and white Americans especially) have been quite hostile to immigrants and immigration. R. J. Simon and Alexander (1993) have shown that American public opinion has consistently favored immigration restriction for decades. Research has documented the existence of high levels of white animosity toward Latinos in many settings across the United States (Campbell, Wong, and Citrin 2006;Hero and Tolbert 1996;Tolbert and Hero 1996). Especially in states that have institutionalized direct democracy, the enactment of restrictive policies targeting immigrants and Latinos is more likely because native voters do not have to go through the moderating filter of party politics in legislatures (Schildkraut 2001). Negative media cues about the costs of immigration tend to increase negative views of non-white immigrants among white natives (Brader, Valentino, and Suhay 2008). In areas experiencing high immigration growth rates, especially of undocumented immigration, perceptions of economic and social threat are magnified (Berg 2009).
Recent developments in state immigration laws have been discussed in the context of racial threat. From California's Proposition 187 in 1984 to recent laws in Arizona, Alabama, Georgia, and other states have pitted lawmakers against immigrants and Latinos who argue that these laws promote racial profiling and discrimination. Evidence exists that race politics has played a role in the shaping and promotion of these types of laws Provine and Chavez 2009). Recent findings also indicate that the growth of immigration when combined with a salient national discourse that depicts immigrants as a threat can lead to the introduction of restrictive legislation at the local level (Hopkins 2010).
States that experienced high immigration growth in the 1990s may have been more likely to exclude LPRs from social programs out of fear of becoming "immigrant welfare magnets" (Borjas 1999), that is, attracting large numbers of poor minority immigrants. During the Congressional debates on PRWORA, immigrants were portrayed as lazy, fraudulent, and undeserving of state support (Yoo 2008). The "dependency" argument that portrayed low-income immigrants as "likely public charges" found support in many quarters in spite of several studies that showed that the enrollment and use of public assistance and Medicaid services among immigrants were significantly lower than among the native-born population (Currie 1998;Tienda and Jensen 1986).
Interestingly, extant findings indicate that the size of the Latino or the foreign-born population of the state may not be a trigger for immigrant exclusion from social welfare, but the size of the African American population, especially as a proportion of welfare rolls, does correlate positively with immigrant exclusion (Graefe et al. 2008). Fellowes and Rowe (2004) present similar findings regarding Latinos in their analysis of welfare reform overall, including citizens. This presents two possible implications: Either the black/white conflict is spilling over to include immigrants and Latinos in areas with large African American concentrations, or this is an indication of intraminority conflict. Certainly, tensions exist between African Americans and Latinos because many blacks perceive of immigrants as economic competition (Gay 2006). A number of studies have documented social and political conflict between blacks and Latinos (Bobo and Massagli 2001;Vaca 2003). This intraminority competition hypothesis is especially relevant in the case of social policy if African Americans believe that welfare resources going to immigrants correspond to a reduction in the slice of the pie that goes to the black community. Should this be the case, we would expect to see more immigrant restrictions emerging in states with large black populations.

Race Politics and Welfare Policy
Race politics has played an equally important role in the development of the country's social welfare system at the federal and the state levels, largely because white public opinion and political leaders have resisted inclusion of minorities in the social welfare net (Brown 1999;Fording 2003;Fox-Piven and Cloward 1971;Gilens 1999;Hero 1998;Katz 1989;Quadagno 1994). State and federal welfare provisions have consistently discriminated against racial minorities: State mothers' pension programs excluded black mothers (Bell 1965;Gooden 1999), whereas the Social Security Act of 1935 barred most blacks through its exclusion of agricultural workers (Brown 1999).
Studies have shown that the presence of large numbers of racial and ethnic minorities among welfare beneficiaries has led those in the dominant white group to object to social programs on the basis that they are racially driven and preferential in nature (Soss et al. 2001;Taylor 1998). Furthermore, the relationship between race and policy outcomes is theorized to be the result of negative perceptions and negative media discourses that paint welfare recipients as "undeserving," "dependent," or "lazy." Analyses of survey data also show that American public opinion has been quite hostile to welfare since the late 1960s (Gilens 1999), and there is a strong tendency to view welfare recipients as "undeserving" (Gilens 1999;Katz 1989). This negative frame of "undeservedness" that had been primarily associated with African Americans was extended to Latinos and immigrants during the period of the welfare reform debate (Gilens 1999;Yoo 2008).
Studies of AFDC and its successor program TANF have documented substantial differences in the generosity and stringency of the programs, which scholars have attributed to racial factors (Soss, Schram, and Fording 2003). Among the earliest to discuss the role of race, Wright (1976) showed that AFDC benefit levels were strongly correlated with the racial composition of the state's population. The conclusion directly tied to Key's (1949) "group threat" theory. Post-PRWORA research confirms the importance of racial threat as a motivator of social policy restriction: Soss et al. (2001) found that stringency was higher in states where the welfare rolls included more minorities. Howard (1999), M. Johnson (2001), and Keiser, Mueser, and Choi (2004) have documented similar findings, including differences at the policy implementation stage. Furthermore, there is strong evidence that prior to PRWORA, states with large black populations were more likely to request waivers from the federal government that allowed them to implement more stringent welfare rules, and more likely to implement more stringent eligibility rules after PROWRA (Fellowes and Rowe 2004;Soss et al. 2001;Zylan and Soule 2000). However, as is the case with some studies of immigration attitudes (Dixon and Rosenbaum 2004;Fox 2004), these findings did not extend to Latinos (Fellowes and Rowe 2004).
As mentioned earlier, analyses of immigrant welfare eligibility at the state level have produced mixed results on the strength of the "group threat" hypothesis. Graefe et al. (2008) found no statistically significant relationship between the percentage of Latino TANF cases or the size of the Latino population and welfare stringency, but the study indicates that states that are traditional immigration destinations and have a long history of immigration are more likely to extend TANF eligibility to immigrants than are new destination states. Hero and Preuhs (2007) found no statistically significant relationship between the overall size of minority populations and immigrant eligibility rules.

Race Politics and Medicaid Policy
Although access to health services, health insurance and resultant health outcomes are all correlated with race, Medicaid has not been a key focus of researchers studying race politics in the context of policy formation. The focus has generally been policy outcomes. In part, this is because the program has not changed eligibility and funding rules for decades. Even after PRWORA, studies of welfare reform did not pay attention to Medicaid because it remained an entitlement program. Most changes to the program have occurred on the supply side and not the demand side.
Today, 47 million people in the United States are uninsured, including 20% of all blacks and Asians and 30% of Latinos. Studies of health insurance coverage show the existence of significant differences in access and coverage between whites and blacks going back to the 1950s and persisting to the current time (Katz-Olson 2010;Thomasson 2006). Similar differences exist between natives and immigrants. According to Currie (1998), in the 1990s immigrants were less likely than natives to have access to private health insurance through employers. Furthermore, although a substantial number of immigrants and their children met the income eligibility criteria for Medicaid, rates of enrollment were significantly lower than for natives. Usage of Medicaid services was also lower among immigrants than among natives. 4 The establishment of Medicaid in 1965 addressed some of the existing racial disparities in health service access by providing medical insurance coverage to lowincome families that met the criteria for AFDC. Over the next two decades, the scope of Medicaid was expanded to include a variety of non-AFDC eligible low-income populations. Although concerns about "undeservedness" have on occasion surfaced in relation to Medicaid, and the origins of welfare medicine are steeped in theories of individual responsibility for disease and punishment for sinful excess (Katz 1986;Stevens and Stevens 2004), 5 the debate over deservedness has been limited in the post-World War II era (Katz-Olson 2010;Stevens and Stevens 2004). Medicaid has received very little media attention except in cases of scandal and fraud (Olson 2010). Furthermore, health care policy in recent decades has tended to be framed around issues of cost, access, and public health rather than "deservedness" (Vilardich 2009). 6 Until recent years, these frames did not encourage the growth of a majority/minority conflict over health care for the poor.
The implementation of PRWORA imposed limits on immigrant access to social programs and devolved to states the decision on immigrant eligibility for Medicaid. However, the program remained a jointly run and funded entitlement without spending caps. The only major change in Medicaid rules related to immigrant eligibility. The immediate result of PRWORA was a substantial decline in the rates of public health insurance coverage among immigrants either because of ineligibility or out of fear and misinformation (Hagan et al. 2003;Kandula et al. 2004).

Constraints and Incentives in TANF and Medicaid
Race politics is only one part of the story. Understanding differences in the politics of policy formation between welfare and health care programs for immigrants requires an analysis of the incentive structures that prevail in each program. TANF and Medicaid operate under very different constraints and incentive structures that may enhance or counter the effects of race politics. Specifically, in the case of TANF, individuals and families bare the cost of exclusion from welfare programs, which makes it easier for the state to yield to dynamics resulting from race politics. However, in the case of Medicaid, the mandate to provide expensive emergency health care without regard to ability to pay means that the program often pays for care ex poste, often at a higher rate. As the state will have to absorb a substantial portion of the costs at a higher price, this creates an incentive for inclusion of more people in Medicaid.
In the case of TANF, the social costs of excluding legal immigrants from TANF are mostly borne by the immigrant families themselves and by the communities within which they lived, but these costs are less visible to the general public. Some of the cost may be incurred by the state in the form of urban blight, higher crime, and the need for additional police in certain poor neighborhoods, but the median voter is less aware of this association (Weir 1997). In addition, white voters (and many minority voters as well) endorse "tough on crime" policies and spending for law enforcement (J. Simon 2007). Policy makers could thus score political points by claiming that immigrant exclusion from TANF encouraged immigrant self-sufficiency and prevented immigrants from becoming "dependent" on the state (Yoo 2008).
In the case of Medicaid, the incentives for states operate quite differently. First, the program has been set up to create incentives for high-poverty states to include more people. Medicaid operates with a formula that is based on a state's poverty level. Thus, states with higher poverty levels receive a higher match from the federal government for each individual they cover through Medicaid. Federal rate of reimbursement ranges from 50% to 80% depending on the state's poverty level. It is thus not surprising that, even as states experimented with ways to reduce AFDC rolls and limit their welfare spending, Medicaid spending increased through the 1980s and 1990s, and states' cost reduction efforts focused on changing reimbursement mechanisms, doctor pay rates, and service delivery systems (Schneider 1998). The goal was not to exclude people from the program; rather, new populations were added, especially pregnant women and young children, as well as elderly and disabled individuals (Schneider 1998). Thus, the incentives to exclude legal immigrants from Medicaid were lower than those for TANF because of the differences in funding structure. Exclusion also requires documentation and systems to process and determine eligibility based on citizenshiprelated criteria. These systems can add to the cost burden for states and citizens covered under the program. For example, recent research has shown that compliance with the additional administrative and compliance requirements of screening out ineligible noncitizens have come at a cost of $600 million (Sommers 2010).
Second, federal law requires states to provide emergency care to all without taking into account the ability to pay. Immigrant families without access to Medicaid may forego cheaper routine care and wait until faced with an acute health crisis to go to the emergency room. Emergency care and hospitalization is far more expensive than routine care. By reducing the use of routine care but not affecting or even increasing the use of emergency and hospital services, states may actually increase their overall health care costs related to immigrant care (Currie 1998). The Congressional Budget Office (1995) in its assessment of the PRWORA legislation warned that immigrant exclusion could shift the immigrant health care burden from the federal government to local emergency rooms and clinics, both the exclusive responsibility of subnational governments. 7 Third, states recognized that the cost of immigrant exclusion could be high for American citizens as well: Health problems associated with the lack of prenatal care are particularly acute and expensive. Citizen children born to uninsured immigrant parents are more likely to be born prematurely, be underweight, and require expensive neonatal care. In North Carolina, 80% of immigrant-related health care costs borne by the state involved childbirth and postnatal care (Committee on the Future of Emergency Care in the U.S. Health System 2007). As the children born to immigrant parents are American citizens by birthright and thus immediately eligible for state benefits, states have an incentive to provide preventive care to immigrants to minimize the costs associated with neonatal services.
Additional evidence based on a study of undocumented immigrants, a population that is eligible only for emergency benefits, indicates that exclusion of LPRs from Medicaid does lead to higher health care costs. Specifically, a comparison between California, which in the 1990s provided some funding for nonemergency care of undocumented immigrants, and the other states that did not offer similar programs shows that the per capita health care cost of undocumented immigrants in California was 60% of the average resident spending, whereas in the rest of the country it was 223% of the average (Ku and Kessler 1997).
Finally, immigrant exclusion from Medicaid could result in significant spillover costs for public health. Unlike exclusion from TANF, which mostly affects the lives of individual immigrants and their families, immigrant exclusion from Medicaid can have a serious impact on the larger population. Public health programs covered through Medicaid, such as access to immunization or screening for contagious diseases such as HIV/AIDS, are beneficial to everyone in society. Not only are citizens and immigrants in increased danger of various communicable diseases but lower immunization rates have real economic costs as well. According to the U.S. Department of Health and Human Services (2000), every dollar spent on immunizations represents future costs savings of $27 in health care. Insured individuals are more likely to be vaccinated for contagious diseases, whereas a cluster of unvaccinated immigrants can have serious implications for public health.
If anything, these incentives suggest that higher immigrant populations should be positively correlated with Medicaid inclusion because states with larger such populations may end up with higher costs related to the uninsured. Furthermore, because African Americans and Latinos have the highest rates of poverty, states with larger such populations should be more likely to be inclusive rather than exclusive in their Medicaid policy. As discussed, this is not the case for TANF where the costs of exclusion are borne primarily by the individuals and their families, not by the state and the population as a whole.

Data and Method
As explained in section "PRWORA and the Devolution of Immigrant Social Policy," the federal government created three new categories of LPRs through PRWORA and enabled states to determine which of these groups, if any, they wish to include in their TANF and Medicaid programs. The analytical goal of this article is twofold: (1) to determine the factors that affected state decisions to include immigrants in TANF or Medicaid and (2) to identify differences between the two programs in terms of influential drivers of inclusion. To achieve these goals, a series of binary variables were coded as "1" when the state offers a program to immigrants and "0" when the state does not include immigrants in the program (see Appendix A for details). The data for the construction of these variables are from the Urban Institute (Zimmerman, Tumlin, and Ost 1999). Continuous dependent variables (indices of inclusion) were created for the TANF and Medicaid models by extracting the principal component from the set of these binary variables. 8 Principal components analysis (PCA) with Varimax rotation was conducted to assess the underlying structure of each set of three variables. This method was used in consistence with Hero and Preuhs (2007) and Graefe et al. (2008) because it allows for the assessment of the degree to which states included immigrants in each of the two programs, instead of, simply, the odds that a state may include a certain portion of the immigrant population in TANF or Medicaid.
Unlike Graefe et al. (2008), this study focuses on the differences in the way states treat LPRs in the context of TANF and Medicaid policy regardless of the person's specific immigration classification (e.g., refugees, asylees). 9 Furthermore, Hero and Preuhs (2007) combined the welfare and health care programs together with several other programs to compute a single-factor score. However, the incentive structures and the politics of each of the two types of policies (welfare, health care) differ; therefore, I contend that treating them separately leads to a better understanding of their social correlates. For that reason, TANF and Medicaid were analyzed as two different linear models and, correspondingly, two indices of inclusion were developed.
The first dependent variable is labeled TANF Eligibility Score and it combines the three binary variables associated with TANF. The TANF Eligibility Score ranges from a minimum of −2.70 to a maximum of 1.93. The higher the index score indicates more inclusive state eligibility rules for TANF. Analysis of eigenvalues shows that these policies can be combined into a single dimension: The eigenvalue of the first principal component has a numerical value of 1.4 that meets Kaiser's criterion for inclusion. Scree plots confirm that the variables can be reduced to a single component. Tests of independent sampling, normality, and linearity were conducted to further establish the strength of the dependent variable. The Kaiser-Meyer-Olkin (KMO) test of sampling adequacy (.649) is within acceptable levels, and the Bartlett test of sphericity is statistically significant (p < .006) indicating that the variables are highly correlated. The Cronbach's alpha for this index score is .72, further indicating that the three variables do stand as a single factor. The total variance explained by this index is 49%.
The second dependent variable is labeled Medicaid Eligibility Score and it includes the three binary variables associated with Medicaid eligibility. The Medicaid Eligibility Score ranges from −3.52 to a maximum of 1.00. The higher factor score indicates more inclusive state eligibility rules for Medicaid. Analysis of eigenvalues shows that these policies can be combined into a single dimension: The eigenvalue of the principal component has a numerical value of 1.5, which also meets the Kaiser standard. Scree plots confirm that the variables can be reduced to a single component. The KMO test of sampling adequacy (.654) is within acceptable levels, and the Bartlett test of sphericity is statistically significant (p < .021) indicating that the variables are highly correlated. The Cronbach's alpha for this factor is .69, which indicates that the score is somewhat less correlated but within acceptable limits. The total variance explained by this index is 50%. Descriptive statistics and source information on the dependent variables can be found in Appendix A.
The TANF score and the Medicaid score models were analyzed using ordinary least squares (OLS) regression. A total of seven independent variables were included in the base model. Demographic and economic variables were lagged. Three predictors related to the racial/ethnic composition of the state population: (1) percentage of population that is black, (2) percentage of population that is Latino, and (3) percentage of population that is foreign born. The population data were derived from U.S. Census sources. Also included are two lagged measures of the state's economic conditions, unemployment and percentage of population under poverty. The sources for the economic variables were the U.S. Statistical Abstracts and the U.S. Bureau of Labor Statistics. Finally, the models include two measures of the political environment of the state. These are Erickson, Wright, and McIver's (1993) measure of public opinion liberalism, and Rom, Peterson, and Scheve's (1999) measure of Democratic Party control updated to include data from 1996 to 1997. Table 1 presents descriptive statistics for all the variables included in the equations.
Appendix B shows the intercorrelations between these variables. Some collinearity exists across variables, but collinearity tolerance tests for the TANF model indicate that all tolerances are well above the 1 − R 2 limit (.38). Collinearity is more of a concern for the Medicaid model where two variables do not meet this standard. Specifically, public opinion liberalism (.53) and percentage under poverty (.57) are slightly below the limit (.61). Because of collinearity concerns, models are developed that exclude the public opinion variable. 10 The final model was formalized as follows: TANF/Medicaid Eligibility Score i = a + β 1 (percentage of population black) + β 2 (percentage of population Latino) + β 3 (percentage foreign born) + β 4 (public opinion liberalism) + β 5 (Democratic Party control) + β 6 (percentage of population below poverty) + β 7 (unemployment rate) + Error i The TANF and the Medicaid models were examined for normality, collinearity, and heteroskedasticity, using visual inspection of the data and standard formal tests. The TANF model shows weak violation of normality (at the 10% level, using the Shapiro-Wilk test), whereas skewness and kurtosis are well within the acceptable norms for weak deviation from normality. No equivalent deviations appear when testing the residuals in the Medicaid model. The model was reexamined excluding possible outlier states (California, Rhode Island, Utah, and Maine) that resulted in approximately 6% change in the coefficient of determination and similarly small changes in the other diagnostic metrics. On the basis of these results, all states were included in the model. As explained above, collinearity diagnostics also fell within acceptable levels (with the exceptions noted) although there is no significant evidence of heteroskedasticity from the residual scatterplots in the TANF model (Breusch-Pagan test p value is .99). There is, however, evidence of heteroskedasticity in the Medicaid model (Breusch-Pagan test p value is .11). The variance inflation factor (VIF) ranges from 1.23 to 1.86 for the TANF and Medicaid models. Both models have the same design matrix whose condition number is 9.43 with a square-root value of 3.07. For reasons of consistency, both models were run using robust standard errors procedures (Huber-White). Furthermore, using the feasible generalized least squares (FGLS) method for seemingly unrelated regression (SUR), to test the hypothesis that the two equations are correlated, no significant evidence of correlation in the error terms of the two regressions was found. Specifically, the correlation between the two regressions was .15, whereas the covariance between the two regressions was 0.086. Table 2 summarizes the effects I expect to see in the models based on the theoretical discussion. As noted earlier, I expect that the size of various minority populations and of the population under poverty correlates negatively with immigrant inclusion in TANF. In the case of Medicaid, however, I expect a positive correlation between the size of these populations and immigrant inclusion. The condition of the economy is expected to be a negative correlate for inclusion of immigrants in both policies. An economic downturn puts pressure on budgets, forcing states to cut down on social services. I expect that states deciding on immigrant inclusion under conditions of economic hardship (measured in terms of unemployment rates) are less likely to be inclusive. However, as noted earlier, I expect that for Medicaid, higher poverty rates should be positively correlated with inclusion because of the different incentive structure of the program. In the case of TANF, poverty rates should be negatively correlated with inclusion. The two political variables, public opinion liberalism and Democratic Party control, are expected to correlate positively for both policies as states with more liberal and Democratic legislators are expected to be more inclusive. Table 3 presents the results of the regressions for TANF and for Medicaid immigrant inclusion. A total of four models are presented, with public opinion liberalism excluded from two of the models. The standard errors for each term are included in parentheses.

Results
It can be seen that the models explain very different proportions of variance. Adjusted R 2 is .479 and .554 for the TANF models, but it is only about half of that (R 2 = .151   and .287) for the Medicaid models. The beta weights for the TANF models indicate that the size of the African American population and the poverty level both contributed significantly and negatively to predicting inclusion. Specifically, when adding 1% point to the African American population, the TANF inclusion score declines by .24 in the comprehensive model (Model 2). The effect of poverty is even greater: Adding 1% point to the poverty level leads to a reduction of the TANF inclusion score by .35 in the comprehensive model. In Model 1, which excludes the effects of political ideology (public opinion liberalism), the effects of these two variables are even larger: .31% for African American and .53% for poverty, respectively. This is consistent with the findings of Graefe et al. (2008). The beta weight for unemployment is positive and statistically significant in Model 1 (.27), but the size of the effect declines substantially and loses statistical significance in Model 2 when political ideology is included. Also consistent with previous research is the significance of political ideology. Model 2 shows that the addition of political ideology to the model increases the variance explained by more than 7% points. Public opinion liberalism is strongly positively correlated (p < .01) with the TANF inclusion score. The size of the Latino and the size of the foreign-born populations are positively correlated with the TANF inclusion score in both models, but these results are not statistically significant and the null hypothesis cannot be rejected. The beta weights for the foreign-born population are quite small in the TANF models.
In the case of the Medicaid inclusion score, of the three population variables, only percentage foreign born is statistically significant (p < .01 in Model 3 and p < .05 in Model 4). The positive beta coefficient indicates that an increase of 1% point in the size of the foreign-born population leads to a .22 increase in inclusiveness in Model 4. In Model 3, which excludes the political ideology variable, the effect of the foreignborn population is even greater (.26). The remaining population variables (percentage African American and percentage Latino) are not statistically significant, and the beta weights are quite small. Interestingly, the coefficient for percentage African American is positive, but the coefficient for percentage Latino is negative in Models 3 and 4. The size of the population under the poverty line is statistically significant in the case of Medicaid but only when political ideology is excluded from the model (Model 3). The beta coefficients for poverty are negative but much smaller than in the case of TANF. Specifically, in Model 3, increasing the poverty rate by 1% point leads to a decline in the Medicaid inclusion score by .28 (p < .01), but in Model 4, the decline is only .4 and not statistically significant. For the political environment variables, Democratic Party control is statistically significant (p < .01) and positively correlated with the Medicaid inclusion score in Model 3, but the relationship weakens and loses statistical significance in Model 4 when political ideology is included. As is the case with TANF, the addition of political ideology (public opinion liberalism) has a substantial impact on the model, adding almost 14% points to the explained variance. Political ideology is positively correlated with Medicaid inclusion (β = 5.217; p < .01).

Discussion
The expectation of Key's (1949) "group threat" theory is that the white majority would resist immigrant inclusion in social programs in states with large minority and immigrant populations. Threatened by the expansion of multiculturalism and social policies benefiting poor minorities, white majorities and their legislative leaders in the states would opt for restriction. The data point to a more complex reality. In the case of TANF, the results indicate that restriction was more likely to occur in states with large African American populations. The size of the Latino and foreign-born population correlates positively with inclusion, although the relationship is not statistically significant and the coefficients are quite small. Other studies have found similar effects (Graefe et al. 2008). The negative impact of the size of the African American population on TANF immigrant inclusion is consistent with more than one explanation. First, in these states, most of which are in the South, the issue of immigration and especially Latino immigration has been linked to the long-standing conflict over African Americans' civil rights. In a sense, the majority/minority conflict of a biracial society has spilled over to include immigrants. Certainly, recent developments in Alabama, Georgia, Tennessee, and Arizona involving undocumented immigrants point to that direction. In the past year, Alabama passed the country's most stringent law against "illegal immigration," exceeding in severity Arizona's Senate Bill (SB) 1070. Critics say that these laws legalize and promote racial profiling, which is detrimental to all racial minorities not just Latinos (Immigration Policy Center 2011).
But this analysis cannot preclude the possibility of immigrant exclusion being the result of intraminority conflict. In states where they are present in large numbers, African Americans also have a stronger presence in the legislature. Immigration has often been portrayed as a threat to African Americans because immigrants compete with blacks for jobs (Borjas 1999). The argument was also made that immigrants are "welfare magnets," which implicitly means that excluding immigrants from welfare programs can increase the share of the funding available for poor African American families (Borjas 1999). Research has shown that economic competition certainly exists between African Americans and Latinos (Gay 2006). My own research in Rhode Island revealed tensions between the two communities over police profiling legislation and policy priorities. However, little research exists on black-Latino relationships in the 1990s. Evidence from recent years indicates that in spite of differences, major African American organizations have supported immigrant and Latino groups. The National Association for the Advancement of Colored People (NAACP) has joined forces with major Latino organizations such as League of United Latin American Citizens (LULAC) and National Council of La Raza (NCLR), and it has taken a strong position against the Arizona law and similar laws elsewhere. It has also fought against the repatriation of Haitians in the 1990s. The organization has even filed amicus briefs in Supreme Court cases relating to these issues (K. R. Johnson 1998). The long-standing position of the NAACP on immigrant rights suggests that welfare was probably not an issue of intraminority conflict in the 1990s but rather one of cooperation. Certainly, more historical analysis is needed on this topic.
In the case of Medicaid, the results indicate that the incentive structure imposed by federal conditionality may mitigate the possible effects of race politics. As mentioned earlier, not only does Medicaid reimburse states with higher poverty rates at a higher rate but federal law also requires universal access to emergency care regardless of ability to pay. The incentive to be inclusive is suggested in the data. The size of the foreign-born population is positively correlated with Medicaid inclusion, showing that states with larger immigrant populations where the burden of ex poste and unreimbursed emergency health services is greater are more likely to include immigrants. Even the effects of poverty are relatively small in Medicaid, and in the comprehensive model they are not statistically significant. This indicates that the incentives built-in to the program lead to higher levels of inclusion across the board.
In TANF and Medicaid, the political environment and political ideology specifically play an important role. Public opinion liberalism explains a substantial portion of the variance in each model and mitigates the effects of race variables. The population coefficients decline markedly in both models when ideology is included. This is consistent with previous research that has indicated that ideology plays an important role in the development of immigration policies (Graefe et al. 2008;Hero and Preuhs 2007;Provine and Chavez 2009). Liberal states are significantly more likely to include immigrants in their TANF and Medicaid programs than are conservative states. It is interesting that the strength of the Democratic Party in state politics is not a statistically significant indicator in three of the four models, and the coefficients are generally quite low (however, all are positive). This is consistent with Tichenor's (2002) supposition that in fact the Democratic Party has traditionally been split on immigration issues, with "blue dog" Democrats in the South likely to support restrictions. Even in some states in the Northeast, often associated with liberal Democrats, recent efforts at immigration restriction have come out of Democrats. Specifically, Arizona-style legislation has been proposed in Rhode Island by conservative Democrats and supported by Republicans. In Pennsylvania too, some Democrats have supported restrictions.

Conclusion
Since at least the early 1990s, Republicans have supported the block granting of Medicaid. In fact, the first bills that proposed welfare reform sought to turn AFDC and Medicaid into block grants, but then President Clinton successfully resisted the change. The issue has come up again in recent years as part of the discussion over health care reform and the recession. Republicans have argued that turning Medicaid into a block grant will free states from a variety of restrictions and save money. Weeks before his exit from office, President Bush granted Rhode Island a waiver that allows the state to circumvent federal regulations, effectively treating the program as a block grant.
How will block granting of Medicaid affect immigrants? The answer is certainly complex. This analysis indicates that political pressure could come to bear and more conservative states may exclude immigrants from the program in the name of "deservedness," even at a significant economic cost. Certainly, states will face significant resistance from ethnic groups but also from hospital associations, medical associations, and other professional groups. In the 1990s, the American Medical Association, the American Hospital Association, and senior citizen groups argued for expansion of Medicaid services and, in the process, shifted the debate away from inclusion to issues such as the desirability of managed care, the type of payment schemes for doctors, and the reimbursement schedules. Given the current political mood and the public's hostility to social policy benefiting minorities and immigrants, it is unlikely that a technocratic debate will prevail. It is thus an open question whether the change of some of the incentives included in Medicaid will turn the program more like TANF where race politics have played more of a role.

Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding
The author received no financial support for the research, authorship, and/or publication of this article.

Notes
1. States could not offer differential programs to legal permanent residents (LPRs); for example, they could not provide smaller benefits or less comprehensive insurance. Only full inclusion or full exclusion was at play with Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA). 2. The law created three new categories of LPRs: those who were granted permanent residency prior to August 22, 1996 (preenactment LPRs), those who were granted LPR status after August 22, 1996, and have been LPRs for fewer than 5 years (new immigrants), and those who received residency status after August 22, 1996, but have been in LPR status for more than 5 years. 3. In the early part of the twentieth century, courts used expansive and creative notions of being a "public charge" to deport immigrants (Alpert 1939). For example, having spent time in county jail for a minor concern was interpreted as having received public support, which could then be used by the federal authorities as grounds for deportation of undesirables as public charges. In the 1930s, states and localities used public benefits as a guise in the "repatriation" of more than half a million Mexican immigrants and Mexican Americans. Internal studies conducted by the U.S. consular services as recently as the late 1970s, noted that visa adjudicators used racial tropes in public charge exclusions: Mexicans and others from Central and South America were more likely to be excluded on the basis of becoming a likely public charge than were Canadians or Chinese immigrants (Anderson and Gifford 1978;K. R. Johnson 1998). In general, public charge exclusions, which tie directly to the fear that immigrants will become "dependent" on the American welfare system, have had a disproportionate impact on minority applicants for permanent residence (K. R. Johnson 1998). 4. Moreover, studies of health outcomes have documented significant discrepancies across racial groups, especially blacks, and these differences are to a great extent the result of differential access to quality health care services (Agency for Healthcare Research and Quality 2009; Katz-Olson 2010). In the case of immigrants, health outcomes vary by ethnic/race group, income, and length of presence in the United States. Health experts have documented the "paradox of assimilation": consistent findings that show deterioration of immigrant health over time as they stay in the United States (see Fennelly 2006 for detailed literature review). 5. Historical accounts indicate that public health programs systematically excluded racial minorities and immigrants based on beliefs that the diseases afflicting these groups were the result of behavioral choices or heredity (Irvin Painter 2010; Katz 1986). 6. In academia, too, much of the analysis has been on the supply side, focusing on issues of managed care, fee-for-service schedules, reimbursement schemes, physician incentives and disincentives, service and drug tiering, and other administrative considerations. 7. The Congressional Budget Office (CBO 1995) estimated that federal savings from immigrant exclusion from Medicaid would be approximately $7.7 billion between 1996 and 2000 but the costs would shift down burdening states and local governments. More recently, the CBO issued similar warnings when arguments surfaced for the exclusion of legal immigrants from the Obama health care reform law (Capps, Rosenblum, and Fix 2009). 8. The approach that I use is conceptually identical to the one used by Hero and Preuhs (2007) and Graefe et al. (2008). I used the dimension reduction/factor command on SPSS v.17. 9. Graefe et al. (2008) created factors based on primary and secondary legal classifications of immigrants (e.g., asylum seekers, battered spouses, new LPRs) on the assumption that states would treat various immigrant groups differently. The focus here is with LPRs in the context of Temporary Assistance for Needy Families (TANF) and Medicaid. 10. I thank the anonymous reviewers for bringing up this issue.