The Politics of Welfare Exclusion: Immigration and Disparity in Medicaid Coverage

We Abstract The rapid growth of the immigrant population in the U.S., along with changes in the demographics and the political landscape, has often raised questions for understanding trends of inequality. Important issues that have received little scholarly attention thus far are excluding immigrants’ social rights through decisive policy choices and the distributive consequences of such exclusive policies. In this paper, we examine how immigration and state policies on immigrants’ access to safety net programs together influence social inequality in the context of health care. We analyze the combined effect of immigration population density and state immigrant Medicaid eligibility rules on the gap of Medicaid coverage rates between native- and foreign-born populations. When tracking inequality in Medicaid coverage and critical policy changes in the post-PRWORA era, we find that exclusive state policies widen the native-foreign Medicaid coverage gap. Moreover, the effect of state policies is conditional upon the size of the immigrant population in that state. Our findings suggest immigrants’ formal integration into the welfare system is crucial for understanding social inequality in the U.S. states. We do not detect troublesome VIF statistics. The mean VIF is 2.72. The VIF statistics associated with Immigration and Eligibility are 3.09 and 2.13, respectively. Government Liberalism is associated with the highest VIF score, 5.82. Including and excluding Government Liberalism do not alter findings pertaining to Immigration and Eligibility


Introduction
The United States stands alone from other industrialized democracies because of its longstanding political struggle over universal health care reforms (Starr, 2011). Among hundreds of thousands of workers who live without health insurance, America's newcomersthe immigrant population-face even more daunting situations. The Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA) of 1996, along with the surge of anti-immigration sentiment among the masses, made it more difficult for immigrants and their children to obtain health care. According to the Census Bureau, the proportion of the foreign-born population without health care coverage was more than double the amount of native-born citizens (DeNavas-Walt, Proctor and Smith, 2011, 25). Immigration status, as Ku and Matani (2001) substantiate, has become "an important component of racial and ethnic disparities in insurance coverage and access to care (247)." The ideological battle on the issue of immigration and immigrants' access to health care has its root in divided opinions about how the American democracy should integrate its newcomers. At the national level, political compartmentalization led to the Clinton welfare reform and substantial retrenchment of federal responsibility in providing health care to immigrant families. 1 As such, much of the political stake was left to the states' discretion (Hero and Preuhs, 2007). The state-level picture, nevertheless, shows mixed promises: a few states have been successful in pushing forward new generous policies to qualify immigrants for Medicaid provisions, while others have followed the federal government and tightened their health care provisions to immigrants. To this end, the health care disparities confronting 1 According to the PRWORA (1996), immigrants entering the US after August 22, 1996 are barred from Medicaid and most other federal-funded welfare programs for the first five years after their entry. Besides Medicaid, immigrants are also barred from federal-funded food stamps, Supplemental Security Income, Temporary Assistance for Needy Families (TANF) benefits, and services provided through the Social Services Block Grant (SSBG). immigrants create a major normative dilemma: on the one hand, immigrants are an integral part of American society; on the other hand, they face more political barriers to being incorporated into the American welfare system. This normative dilemma has motivated an important research agenda in welfare state politics that deals with both immigration and social inequality.
Numerous prior studies have examined the relationship between immigration and welfare policy, much of which focuses on either how attitudes toward ethnic minorities influence public support for welfare (Gilens, 2000;Hainmuller and Hiscox, 2010;Martin, 2001), or on how immigration and ethnic diversity influence the generosity of welfare states (Agrawal, 2008;Borjas, 1999;Hero and Preuhs, 2007;Nannestad, 2007). Research has increasingly theorized that immigration-induced ethnic diversity is a challenge for sustaining generous welfare states. Yet, much fewer studies chart the distributive consequences of state policies that define immigrants' legal access to safety net programs. In this study, we take a different prism by focusing on the link between immigration, state policies, and social inequality in the context of health care. Our primary goal is to uncover how states' policy decisions in welfare inclusion/exclusion of immigrants influence inequality in health care coverage between immigrants and their nativeborn counterparts, and how such an effect is conditional upon the state environment for immigrants.
Our exploration of immigrants' welfare rights and the implications on social inequality hinges on a political exclusion perspective. We contend that the exclusion of immigrants from the welfare system at the subnational level enlarges the Medicaid coverage gap between the native-and foreign-born populations for two reasons. First, states use restrictive eligibility policies to formally exclude some immigrants from safety-net programs (Ku, 2009a;Zimmermann and Fix, 1998). Second, exclusive policies create an icy policy environment that sends negative signals about the role of government, thus these policies produce negative social constructions for targeted populations and closely associated groups (Campbell, 2012;Schneider and Ingram, 1993;Soss and Schram, 2007) which discourages eligible immigrants from participating in safety net programs (Fix, 2009). Extending previous studies that only compare state-level immigrant welfare eligibilities in one year (Hero and Preuhs, 2007;Filindra, 2013), we offer the first systematic comparison of immigrant Medicaid eligibility rules across states and multiple years in the post-PRWORA era. We also use immigrant network theory to endorse a conditional effect of state immigrant welfare policies and the state-level immigrant population density on social inequality. We contend that the Medicaid coverage gap between immigrants and their native-born counterparts is larger in states with more exclusive policies, and this positive relationship between state policy exclusiveness and social inequality is strengthened in states with lower levels of immigrant population density.
We then empirically examine the combined effects of immigrant population density and state-level immigrant Medicaid eligibility rules on the native-foreign Medicaid coverage gap.
The state-level panel data analysis shows support for both hypotheses. We find that states with inclusive Medicaid policies have lower levels of health care inequality than states with exclusive Medicaid policies. In addition, immigrant population density is also found to condition the relationship between state policies and disparities in Medicaid coverage. The native-foreign Medicaid coverage gap is seen as the greatest among those states with a relatively sparser immigrant population and very exclusive Medicaid policies, yet the gap is negligible in states with a denser immigrant population and inclusive immigrant Medicaid policies. Our research suggests that social inequality in the U.S. states cannot be fully understood without considering the politics of exclusion in policymaking and immigrant social network at the subnational level.

Immigration, Political Exclusion, and Social Inequality in Health Care
The United States witnessed a substantial increase in immigration in the past few decades.
According to the Census Bureau, in 1970, the United States had a foreign-born population of approximately 9.5 million; however, the number increased to 38 million in 2007, almost quadrupling from 1970 (Census 1999;. 2 This rapid demographic change coincides with the rising public concern regarding immigrants' use of public services, such as education, social assistance, and health care (Burns and Gimpel, 2000;Fix, 2009;Hainmuller and Hiscox, 2010).
Exclusive welfare reforms at both the national and state level restricted immigrants' access to government funds that finance health insurance plans for low-income families. As a result, the gap in health insurance coverage between native-and foreign-born populations continues to grow (LaVeist, 2005). The nexus between immigration and social inequality in health care access has recently gained scholarly attention and offered a new basis for studying the implications of immigration on social equity. Various theoretical frameworks have been developed to probe the link between immigration and social inequality, and we focus on twothe politics of exclusion and the immigrant social network.

Political Exclusion through Welfare Reform
A growing body of literature finds that immigration has increased racial and ethnic complexity in American states and raised new challenges to sustaining generous social policy provisions (Miguel and Gugerty, 2005;Hero, 2010). The increasing race and ethnic diversity may dissolve social cohesion and reduce the generosity of safety net programs (Alesina, Glaeser and Sacerdote, 2001;Burns and Gimpel, 2000;Hero, 1998;Hero and Tolbert, 1998;Rowthorn, 2008;Stichnothe and Van der Straete, 2011). Consistent with this social erosion argument is the group-competition and political exclusion argument. For example, Esses et al. (2001) assert that immigration in North America, to a large extent, has triggered perceived threats and groupcompetition among native-born populations. With such a symbolic threat, an in-group (e.g. native-born citizens) is likely to demand policies that restrict an out-group (e.g. immigrants) from accessing their material resources.
Indeed, the 1996 federal welfare reform was driven by a wave of strong anti-immigrant sentiment along with the resurgence of nativism after the passage of Proposition 187 in California (Agrawal 2008, Alvarez andButterfield 2000). From 1982 to 1992, the United States witnessed the number of immigrant applicants for Supplemental Security Income (SSI) double in size. By 1992, the number of immigrant recipients rose to over 600,000 and accounted for more than 25% of the total number of recipients (House of Representatives Committee on Ways and Means, 1998). Both the American public and the federal government were concerned about such sharp increases in the volume of immigrant welfare recipients, as well as the possibility that immigrants might "bring in their parents …with the intention of supporting them by taking advantage of the welfare benefits" (Committee on Ways andMeans, 1998, 1398). Facing the increasing concerns that immigrants' consumption of social assistance may lead to a slew of problems threatening the resource pool for low-income citizens, Congress passed the PRWORA in 1996, which restricted immigrants from receiving federal-funded welfare benefits including Medicaid in the first five years after their entry. One negative consequence of excluding immigrants' welfare rights under the PRWORA is that it led to a substantial reduction in immigrants' participation in various safety net programs, including Medicaid. Ku and Papademtriou (2007) report that, since the enactment of the PRWORA, low-income non-citizens have had much lower Medicaid coverage rates than low-income citizens.
The federal-level reform gave states discretion to make complementary state welfare policies. For example, by using state funds, states can determine (1) whether or not to give legal immigrants who entered the United States before August 22, 1996 access to Medicaid, (2) whether or not immigrants who entered the United States after August 22, 1996 are eligible for Medicaid during the five-year bar, (3) whether or not immigrants who entered the United States after August 22, 1996 are eligible for Medicaid after the five-year bar, (4) whether or not to provide Medicaid to unqualified immigrants in the first five years after their entry, and (5) whether or not to have state funded health insurance programs for immigrants (Bitler and Hoynes, 2013). States differ in the making of their own immigrant Medicaid policies, with some strictly following the federal law without providing immigrants with any additional Medicaid coverage, and others using state funds to provide Medicaid coverage in all five aspects. For example, ever since 1996, Wyoming has strictly enforced the federal regulations and excluded immigrants from Medicaid. On the contrary, states like California, Massachusetts and Pennsylvania decided, immediately after the 1996 PRWORA, that they would use their own state funds to fund immigrants Medicaid in all of the above-mentioned aspects. Ever since then, they have been providing Medicaid to immigrants and treating them just like citizens.
More tellingly, exclusive state policies not only disqualify non-eligible immigrants' access to safety net programs, but also discourage eligible immigrants from participating in social programs by producing negative policy feedback. Social construction theory (Schneider and Ingram 1993) suggests that specific public policy designs create positive or negative social constructions for the targeted populations. Such social constructions can directly influence the government's role in providing social welfare, and the targeted population's behavior of welfare participation. For instance, exclusive state policies create an icy policy environment for immigrants by attaching them with negative social labels, such as "being lazy", "undeserving", and "welfare magnet." Through such stigmatization, exclusive policies send negative signals about the role of government in the lives targeted, depressing eligible immigrants' participation in these policy programs.
As Hacker (2006) explains, in health care and other social policy areas, government often hesitates to directly provide social protection to the underclass, which sends "unwelcoming" signals to those who are negatively socially constructed. Other empirical studies show that subnational policies that are exclusive (or punitive) toward undocumented immigrants can have negative policy feedback (Campbell, 2012;Soss and Schram 2007) or the so-called "chilling effect" (Fix, 2010;Waston, 2014) on eligible immigrants' participation in safety-net programs.
In their study of health-care services in immigrant communities in five metropolitan counties in Texas, Hagan et al. (2003) find that many eligible immigrants voluntarily withdrew from Medicaid after Texas followed the federal welfare reform and initiated exclusive immigrant eligibility rules for various state safety net programs. Other scholars who study welfare reform and immigrants' Medicaid enrollment report a similar "chilling effect" on non-citizens' Medicaid participation in more than one state (Bilter, Gelbach and Hoynes, 2005;Ellwood and Ku, 1998;Ku, 2009a). As Ku and Matani (2001, 247) describe, changes in welfare program eligibility rules constitute "an important component of racial and ethnic disparities in insurance coverage and access to care." In sum, the underlying mechanism of the "chilling effect," explained by Hook (2003, 614), is that "because of immigrants' particularly vulnerable legal and social status, the immigrant-specific provisions of welfare reform may have increased immigrants' confusion about their eligibility for welfare benefits and heightened their distrust of the U.S. government." Tightening eligibility is conceptualized as an important source of the icy policy climate for immigrants (Fix, 2009;Potocky-Pripodi, 2004;Ku, 2009a,b). Considering that exclusive state welfare policies not only set legal barriers for immigrants' welfare participation but also cause a "chilling effect" and depress eligible immigrants' welfare participation, we expect the nativeforeign difference in Medicaid coverage to be larger in states with exclusive immigrant Medicaid eligibility rules than that in states with inclusive policies.

Immigrant Population Density as a Conditional State Context
American states differ from one another in their stocks of immigrants as well as their immigration flows. Gateways such as California, New York, Florida, and Texas not only maintain a large foreign-born population, but also saw large amounts of immigration inflows in recent years. New destination states, such as Alabama, North Carolina and South Carolina, experienced large influxes of immigrants in the two most recent decades, although their foreignborn population stocks might not be as high. States such as West Virginia, Montana, Wyoming, Kentucky and Nebraska have exactly the opposite experience in immigration. Not only were their foreign-born population stocks low, immigration inflows to these states in recent years were also low. Both the existing immigrant population that resided in a state for a relatively long time period and the newcomers who arrived to a state recently formed important immigrant networks.
These immigrant networks play an important role in welfare participation through two pathways: (1) reducing the stigma of welfare participation, and (2) information spillovers (Banerjee, 1992;Bikhchandani, Hirschleifer and Welch, 1992;Borjas, 1995;Case and Katz, 1991;Fudenberg, 1993, 1995). Simply put, socializing with other immigrant welfare recipients will make individuals feel less shameful of taking welfare benefits. Meanwhile, immigrants are more likely to obtain more information about welfare programs in communities with strong immigrant social networks than in places where they are isolated.
Information spillovers through immigrants' social networks provide an important mechanism for immigrants to learn about welfare programs and eligibility rules. Such a mechanism is crucial for immigrants' welfare participation, because it is very common for immigrants not to know if they are eligible for welfare programs such as Medicaid in their state. Such blindness is due to a combination of factors. First, state welfare eligibility rules are often quite complicated and contain multiple aspects that could influence the eligibility of a particular immigrant. For example, immigration status, age, length of stay, and whether the first entry was before or after August 22, 1996 could all potentially influence an immigrant's eligibility for Medicaid in one way or another. Besides the complexity, eligibility rules also change over time and vary substantially across states. Many states have changed their immigrant eligibility rules more than once since 1996. Mastering the knowledge of the eligibility rules and keeping up with the changes is a challenging job to any immigrant individual. Last but not least, information on these immigrant welfare eligibility rules, supposedly all public information, is ironically publicly unavailable in a vast majority of the states. Indeed, we discover in our endeavor of data collection that few states publicize information on immigrant welfare eligibility rules on a public web site. We also discover from our email correspondence and phone conversations that officials in state Medicaid agencies who deal with Medicaid recipients on a daily basis do not always know the eligibility rules for immigrants in their own state.
The fact that immigrant welfare eligibility rules are complex and relevant public information is not readily available invites us to consider other factors in the policy-disparity mechanism. Even though welfare exclusion of immigrants plays an important role in state level native-foreign Medicaid coverage gaps, the working of such exclusion should depend upon the immigrant network. In states with a denser immigrant population, there is a much higher chance for an average immigrant to use the social network among fellow immigrants to learn about the eligibility rules and participate in welfare programs. In contrast, in states with a sparse immigrant population, immigrants struggle to obtain information from a social network of their own about how to participate in various welfare programs. Imagine if an individual migrates from the Philippines to the United States and wishes to participate in its welfare programs. Both California and Alaska have quite generous welfare policies toward immigrants, yet, the Filipino might have a much better chance to have contacts with his/her fellow immigrants, learn about the eligibility rules from the large immigrant population in California and jump on Medicaid as soon as conditions mature. In contrast, if the Filipino arrives to Alaska where there are few immigrants, it will possibly take him/her a much longer time to learn about the welfare eligibility rules or (s)he will never learn about them. Indeed, by using a micro sample from the Census data, Bertrand, Luttmer and Mullainathan (2000) find strong empirical evidence that immigrants who have more contacts with other immigrants, especially immigrants with knowledge of welfare programs, are much more likely to participate in welfare programs themselves. Aizer and Currie (2003) also find that the use of publicly funded prenatal care in California is highly correlated within race/ethnicity groups and neighborhoods (2574).
Alternatively, the size of immigrant population is also an important demographic factor that is indicative of immigrants' (especially Latinos') political mobilization and their influence in the policymaking process (Leighley 2001, Sanchez 2006. A large and strong immigrant population can positively influence how state-level political institutions deal with immigration-related social policy issues. Indeed, public opinion research shows Latinos are more attentive to immigration policies and have more liberal policy attitudes than non-Latino voters. Whether or not immigrants' (or Latinos') policy preferences are reflected in the actual policy making process depends on the level of their acculturation and how political elites respond to their policy interests (Branton 2007). In fact, previous research shows that political elites (such as legislators) are more likely to be responsive to immigrants' demands in states with a large immigrant (or Latino) population than states with a small immigrant population (Casellas 2009).
Considering that a large immigrant population will better provide a necessary social network for immigrant welfare participation and promote immigrants' influence on the policymaking process, we argue that the size of the immigrant population will condition the

Data and Methods
We devise a cross-section-time-series design by pooling state-level data of the native-foreign (1) Figure 1 presents the cross-state and cross-year variation of the inequality measure.
Overall, the net-difference between native-born and foreign-born individuals' Medicaid coverage does not change dramatically within states, but the between-state comparison is much sharper than within state differences. Figure  Immigrant Medicaid Eligibility Score. Our first key explanatory variable "Immigrant Medicaid Eligibility Score" is an index measure of states' immigrant Medicaid eligibility rules in five specific areas. Specifically, we code the following five major eligibility rules for each state: (1) whether or not states grant Medicaid coverage to pre-PRWORA immigrants (i.e. immigrants who settled in the U.S. before August 22, 1996), (2)  provision of Medicaid to certain unqualified immigrants for non-emergency medical care; (5) whether or not immigrants are eligible for state-only health insurance program for nonemergency medical care. A few prior studies provide snapshots of immigrants' Medicaid eligibilities (Bitler and Hoynes, 2013;Fortuny and Chaudry, 2011;Tumlin, Zimmermann and Ost, 1999), but not in all the years from 1998 to 2010. Our annual eligibility score measure is a compilation of these prior policy snapshots (Tumlin, Zimmermann and Ost, 1999) and our originally collected policy data through email and phone surveys of the 50 state Medicaid agencies. 4 We code the first three eligibility items as "2" if all immigrants are eligible for Medicaid coverage, "1" if some immigrants are eligible, and "0" otherwise. We code the last two eligibility items as "1" if state-provided coverage is available to immigrants and "0" otherwise.
Because we code each eligibility item as a categorical variable, with arbitrarily assigned values (e.g. 0, 1, 2, etc.), an additive scale will result in the first three eligibility items carrying more weight. Moreover, in order to capture more nuanced differences in states policies, we code the first three eligibility indicators by using a three-category ordinal scale, but the last two  Table 2). To cross validate policy information from multiple sources and to track policy changes in years between 1998 and 2010, we conducted a structured email/phone survey over the fifty state agencies (see Table 1). We sent email inquiries to each state agency, followed by one or multiple phone calls if we did not receive any email response from a state agency. In our structured email/phone survey, we asked state agencies to verify their current Medicaid immigrant eligibilities along the aforementioned five areas. We then asked state agencies to provide information about any policy change between 1998 and 2010 ( i.e. whether there were any changes in immigrant Medicaid eligibility between 1998 and 2010; if so, what changes and when they took place).
indicators are dichotomous. In other words, the five eligibility items are scaled differently, which is a problem for generating an additive score. Therefore, we adopt the Bayesian measurement approach developed by Kevin Quinn (2004) to deal with such mixed multivariate responses.
Compared with a simple additive scale or the standard factor analysis, Quinn's Bayesian factor analysis approach has two advantages. First, the Bayesian factor analytical model produces a Racial/Ethnic Diversity. In our empirical models, we include the racial/ethnic diversity of state population as a control variable. According to the "group competition and exclusion" thesis mentioned in section 2, racial diversity triggers perceived threat and group-competition among native-born citizens (Esses et al., 2001;Hero and Preuhs, 2007). As a consequence, native-born citizens might demand policies that restrict immigrants' access to public health care. Therefore, racial diversity is expected to be positively associated with the native-foreign difference in Medicaid coverage. We measure racial and ethnic diversity based on the Blau Index (Blau, 1977;Hero, 1998;Tolbert and Hero, 2001): In equation (2), i and t index a specific state-year observation, j indexes a particular racial and ethnic group, and p denotes the proportion of group j as a share in the total population. We accounted for five racial groups (white, Black, Hispanic, Asian, and others) and scale the diversity measure from perfect homogeneity (0) to perfect heterogeneity (100). 6 Macroeconomic Factors. We include Unemployment and Poverty 7 as two macroeconomic factors that affect government redistribution in general. Moreover, Union Density is considered as another labor-market factor that influences state-level redistributive politics, including inequality in health care. Considering union's pro-immigrant attitudes in recent history, we argue that the union should reduce social inequality between immigrants and native-born citizens. This variable measures the percentage of wage and salary employees who are labor union members. Data for all three socioeconomic variables are drawn from the U.S.

Census Bureau's Current Population Surveys.
Political Contexts. We include a set of state-level political variables to control for 6 Because the Diversity index is computed by counting states' Hispanic and Asian population, it has a positive correlation with the Immigration variable. To make sure that our key result pertaining to Immigration is not affected by the correlation between these two variables, we run a robustness check by replacing Diversity with % Black Population. We obtain comparable results in models using % Black Population. Moreover, with the consideration that different ethnic minority groups may have different preferences on social and immigration issues, we reestimate the empirical models by replacing the Diversity index by % Black, Hispanic, and Asian populations. This alternative model specification does not alter the substantive findings regarding how Immigration and Eligibility interactively affect the native-foreign Medicaid coverage gap. See more details in the Supporting Information. 7 Because we focus on comparing the native-and foreign-born Medicaid coverage rates, it is conceivable that the relative poverty rates between these two groups rather than the overall poverty rates may have an impact. In the Supporting Information, we re-estimate the empirical models using two native-foreign relative poverty measures to replace the overall Poverty variable reported in the manuscript. The first relative poverty measure captures the net difference between poverty rates for foreign-and native-born population. The second relative poverty measure is the ratio of foreign-native poverty rates. Using these two relative poverty measures does not alter our key substantive findings. different political contexts. First, we control for mass liberalism, because prior studies suggest voters' liberal-conservative orientation affects the politics of immigration (Monogan, 2013), welfare generosity toward immigrants (Hero and Preuhs, 2007), and the overall welfare generosity (Erikson, Wright and McIver, 1993). We expect to see a negative association between mass liberalism and the native-foreign difference in Medicaid coverage. The Mass Liberalism variable is the Pacheco (2011)  Second, we control for governor's partisanship. According to Bartels (2008), the partisanship of political executives has an influence on inequality. More specifically, Democratic presidents tend to prevent inequality from growing, while Republican presidents do not care about inequality as much, and therefore, inequality levels tend to increase under Republican presidents. Since our analysis is at the state level, we borrow Bartels' "political executives' partisanship and inequality" thesis, and argue that the partisanship of a political executive also influences social inequality levels at the state level. Therefore, the gap of Medicaid coverage between native-and foreign-born should be smaller in states with Democratic governors compared to states with Republican governors.
Third, we include the percentage of Democrats in state legislatures. Numerous scholars have examined the link between the partisan balance of state legislatures with welfare generosity, and they often connect left-wing partisanship with more redistribution, because left-wing parties mainly draw their support from the working class, who favor generous welfare spending (Hibbs, 1977;Tufte, 1980;Bradley et al., 2003;Bartels, 2008). On the other hand, right wing parties are often times linked with low levels of support for welfare spending and high levels of inequality (Hibbs, 1977;Tufte, 1980;Bradley et al., 2003;Bartels, 2008). Based on this contention, we argue that the percentage of Democrats in state legislatures is negatively associated with inequality in health care between native-and foreign-born populations. Data on these two variables are collected from Kapeluck and Garand (2011). 8, 9 Lastly, we include a dummy variable for southern states, because southern states have unique historical, political and cultural characteristics that differentiate them from other states (Key, 1949). Including the southern-dummy variable also helps to control for other unobserved policy factors, such as the emerging trend of adopting anti-immigration laws and aggressive local immigration enforcement in southern states (Rocha et al. 2014). We expect to see southern states to have greater social inequality.
Model Specification. Because we pooled data from fifty states and fifteen years, we consider both cross-state heterogeneity and time dependence in the pooled CSTS analysis (Beck and Katz, 1996;Beck, 2001). To deal with both heterogeneity and contemporaneous correlation across states, we implemented the Panel-Corrected -Standard-Error procedure (PCSE) proposed by Beck and Katz (1996). In addition, an AR(1) error specification is applied to the panel model to correct for serially auto-correlated disturbance terms. Based on the analysis of residuals, we identified a handful of state-year cases that provide unreliable data on the foreign-born Medicaid coverage rates. When mapping these state-year cases into the CPS sample, we see that they are all state-year cases, whereby the CPS sample suffers from small-population sampling errors for 8 Nebraska has a non-partisan state legislature, we proximate the Democratic Seat Share variable for Nebraska using Census Bureau's biannual data on vote cast for US Representatives by major political parties. As such, Nebraska is not excluded from the empirical analysis. 9 Both the immigrant population density variable and the set of state-level political variables included in our model might be correlated with state Medicaid eligibility rules for immigrants. To make sure that multicollinearity is not a concern, we check the variance inflation factor (VIF) statistics after estimating an OLS baseline model including all the explanatory variables. We do not detect troublesome VIF statistics. The mean VIF is 2.72. The VIF statistics associated with Immigration and Eligibility are 3.09 and 2.13, respectively. Government Liberalism is associated with the highest VIF score, 5.82. Including and excluding Government Liberalism do not alter findings pertaining to Immigration and Eligibility.
the foreign-born population. We dropped these unreliable state-year cases from the empirical models reported in the paper. 10 The potential endogenous relationship between the size of the foreign-born population and native-foreign differences in Medicaid coverage is another important issue that we are concerned with. Prior studies have reported that immigrants tend to cluster in states with generous welfare benefits (Borjas, Bronars and Trejo, 1992;Borjas, 1999;Frey et al., 1996), 11 or in states where their relative economic opportunity and access to welfare resources are better (Hero, 1998). If the endogenous selection presents, the relationship between immigration and native-foreign differences in Medicaid coverage may not be static.
Instead, there could be a long-run relationship between the two variables. With our panel data, we do observe a weakly endogenous relationship between immigration and relative Medicaid inclusion of immigrants. Regressing Immigrationt on Inequalityt produces a negative and significant coefficient for Inequalityt. When regressing Immigrationt on Immigrationt-1 and Inequalityt-1, we obtain an insignificant coefficient for Immigrationt-1. When regressing Inequalityt on Inequalityt-1 and Immigrationt-1, we obtain a significant coefficient for Immigrationt-1. The more complex dynamics suggest that the stock of immigrants and the level of inequality in Medicaid coverage may share a long-run equilibrium relationship. In other words, changes in immigration and changes in the native-foreign Medicaid coverage gap may adjust to each other over time. Other scholars, who use alternative datasets to study immigration and welfare provision, also find a similar dynamic relationship (Lipsmeyer and Zhu 2011).
To depict the long-run dynamic relationship, we added an error-correction specification in our model, following the econometric theories contributed by Engle and Granger (1991) and Banerjee et al. (1999). For the sake of parsimony, we specified the generalized one-step error correction model (De Boef, 2001;De Boef and Keele, 2008). The dynamic component is written as equation (3), in which i and t index state and year; β denotes the vector of coefficients corresponding to all the control variables; and X denotes the vector of control variables. In the subsequent section, we report both the static and dynamic models and discuss our key findings primarily based on the error correction model.
Despite the biased estimations in Model (1), both models report consistent signs of all explanatory variables. In both models, the linear term of Eligibilityt, Immigrant Medicaid Eligibility Score has a negative and significant coefficient (b= -2.116, SE=0.510 in Model (1); b= -1.273, SE= 0.365 in Model (2)). The interaction terms in both models are positive and significant. Overall, we find support for the hypotheses that immigrant population density and states' immigrant Medicaid eligibility rules interactively shape the native-foreign Medicaid coverage gap. 12 [ Table 1 About Here] The significant coefficients in both Models (1) and (2) mean that the effect of states' immigrant eligibility policies is conditional upon the immigrant population density within that state. Because coefficients in an interaction model are difficult for direct interpretation, we use Figure 4 to show the marginal effects of Immigrant Medicaid Eligibility Score conditional upon the two immigration variables (Brambor, Clark and Golder, 2006). Since the error-correction model is more appropriate than the static model in depicting the relationship between state policy, immigrant population density and inequality in Medicaid coverage, we generate Figure 4 based on Model (2).
[ Figure 4 About Here] Figure 4a shows the marginal effect of Immigrant Medicaid Eligibility Score conditional upon the immigrant population density in the previous year. We observe that in states with sparser immigrant populations, state Medicaid eligibility restrictions have negative and significant marginal effects on the native-foreign difference in Medicaid coverage. This negative marginal effect, however, is attenuated as the immigrant population density increases and cannot be statistically differentiated from zero in states with a very dense immigrant population (approximately, Immigrationt-1 ≥ 20%). 13 Figure 4b shows a consistent pattern that the effect of coverage gap between citizens and non-citizens, and the Medicaid coverage gap between native-born citizens and naturalized citizens. The comparison between native-born citizens and naturalized citizens would be a stricter test of our hypotheses, because naturalized citizens enjoy the similar citizen privilege as native-born citizens under PRWORA. Using these two alternative "gap" measures, we still reach similar substantive conclusions. We report these additional analyses in the Supporting Information. Ideally, we would also like to analyze Medicaid coverage rates for undocumented immigrants. To do so, we not only need reliable state-level estimation of the undocumented immigrant population, but also need individual-level records for how many undocumented immigrants are enrolled in Medicaid. The CPS data we use in this paper does not provide information on immigrants' legal status, therefore, we cannot separate undocumented immigrants from other immigrants in our analysis. 13 Our data show that states such as California, New York and New Jersey ( (Berry, Golder and Milton, 2012). Therefore, to gauge empirical evidence for H2, we further examine whether both the short-run and long-run effects of immigration differ across the value of Immigrant Medicaid Eligibility Score in Figure 5 and Figure 6.
[ Figure 5 About Here] In equation (3), the coefficient of Immigrationt-1 (α2) represents the immediate changes in inequality associated with a one-unit change in immigration in year t-1, when Immigrant Medicaid Eligibility Score equals zero. The coefficient for Immigrationt-1 × Eligibility, α5, represents how the immediate impact of Immigrationt-1 is conditional upon Immigrant Medicaid Eligibility Score. We substantively illustrate this conditional effect in Figure 5a. This interaction figure is generated for two different policy scenarios: exclusive and inclusive state Medicaid policy, by setting Eligibilityt at its 10th (exclusive states) and 90th (inclusive states) percentiles.
We use the Clarify program (Tomz, Wittenberg and King, 2003) to simulate the mean predicted based on other survey samples could be slightly different from ours. For example, using survey samples from the American Community Survey (ACS), the Pew Research Center estimated that, in 2012, the foreign-born population in Florida and Nevada was around 19% and 19.5%,respectively. changes in the native-foreign Medicaid coverage gap across the full range of values observed for Immigrationt-1, holding all the other control variables constant. Figure 5a shows that the effect of Immigrationt−1 differs in states with inclusive and exclusive Medicaid policies. When the percentage of foreign-born population is between 0% and 11% in the previous year, predicted changes in the native-foreign Medicaid coverage gap is always positive in exclusive states, but equal to zero in inclusive states. Overall, when the immigrant population density is high, immigration is associated with diminishing inequality.
Using the same method, Figure 5b shows the predicted changes in inequality along the full range of ∆Immigrationt, in exclusive and inclusive states. This conditional short-run effect of ∆Immigrationt is reflected by α3 and α6 in equation (3). In Figure 5b, we observe a negative and significant relationship between ∆Inequalityt and ∆Immigrationt in states with exclusive immigrant Medicaid eligibility policies, but not in states with inclusive policies. Figure 5 provides consistent evidence supporting H2.
The error-correction model also specifies a long-run equilibrium relationship between immigration and inequality, conditional upon states' Medicaid policies. This long-run relationship means that an increase in immigration disrupts the underlying equilibrium, causing inequality to be too high. Therefore, when Immigrant Medicaid Eligibility Score equals zero, the inequality measure will respond by decreasing a total of 0.262 points (i.e. This long-run effect is statistically differentiable from zero (shown in Figure 6a), and distributed across subsequent four years at a rate of 61.4% per year. This means that a 1% increase in foreign-born population leads inequality to decrease a total of 0.334 points in a five-year period.
When Immigrant Medicaid Eligibility Score is high, however, the mean estimated total long-run effect of immigration becomes much smaller (approximately -0.129), but its 95% confidence intervals overlap with zero. It is only distributed through the subsequent two years. We observe different long-run dynamics in exclusive and inclusive states, which provides additional support for H2. social networks and condition the relationship between state policies and social inequality. A natural extension of our research would be to further explore alternative approaches to operationalize the concept of immigrant social network along these lines.
Our research shows that states are important stakeholders when it comes to health care equality. As the Supreme Court has upheld the Patient Protection and Affordable Care Act of 2010, states will again be considered as pivotal stakeholders for policy implementation of health care reform in the near future. How could the American health care system be transformed to better incorporate its newcomers? Perhaps, the answer is more about bottom-up reforms from the states and less about the polarized political struggle at the national level.
To conclude, focusing on the social inequality aspect, we have explored the intersection of two problematic domains of the American democracy-immigration and health care. Our findings point toward a more complex relationship between immigration and social inequality in public health care provision. We show that the vulnerable group's (immigrants) relative wellbeing in a plural society hinges on a complex set of factors including its own group size, policy setups that define who are the "deserving" constituents, and the connection between socio-economic and political factors. Although providing everyone who is pursuing the "American dream" in the United States equal access to health care remains to be an "American struggle," the state-level picture presented here seems to shed some light on a future promise. So far, about half of the states have provided solely state funded health coverage to foreign-born residents and different strategies have been used to reduce eligibility restrictions to immigrants' access to health care. These inclusive policies help close the health care coverage gap between vulnerable immigrants and their citizen counterparts. Surprisingly, our research is one of the few systematic studies examining state-level immigrant welfare eligibility rules and its effect on social outcomes over time. Of course, Medicaid is only one of many welfare programs that are co-sponsored by the federal and state governments. Given that such policies have important implications on social equity in relation to over 40 million immigrants in the United States, scholars are encouraged to explore various state-level immigrant welfare policies, as well as their social and political implications on the quality of American democracy.